JOURNAL OF GEOPHYSICAL RESEARCH, VOL. tOO, NO. Btt, PAGES 22,509-22,522, NOVEMBER t0, 1995 Quantitative analysisof abyssalhills in the Atlantic Ocean: A correlation between inferred crustal thickness and extensionalfaulting John A. Goff Universityof TexasInstitutefor Geophysics,Austin Brian E. Tucholke,JianLin, andGary E. Jaroslow WoodsHole Oceanographic Institution,WoodsHole, Massachusetts Martin C. Kleinrock VanderbiltUniversity,Nashville,Tennessee Abstract. A recentcruiseto the Office of Naval ResearchAtlanticNaturalLaboratoryobtained -100% Hydrosweepbathymetriccoverage,>200% Hawaii MR1 (HMR1) sidescancoverage, gravityandmagneticsover an areaspanningthreeridge segmentsalongaxis (--25ø25'Nto -27 ø10'N), andcrustalagesfrom0 to 26-30 Ma (-400 km) on thewestflank of theMid-Atlantic Ridge. This datasetrepresents a first opportunityfor an extensiveregionalanalysisof abyssalhill morphology createdat a slowspreading ridge. The primarypurposeof thiswork is to investigate the relationshipbetweenabyssalhill morphologyandthepropertiesof the ridgecrestat whichthey were formed. We applythe methodof Goffand Jordan[1988] for the estimationof twodimensionalstatisticalpropertiesof abyssalhill morphologyfrom the griddedHydrosweep bathymetry.Importantabyssalhill parameters derivedfrom thisanalysisincluderoot-meansquare(rms) height,characteristic width, andplan view aspectratio. The analysisis partitioned into two substudies: (1) analysisof near-axis(< 7 Ma) abyssalhills for eachof the threesegments and(2) analysisof temporalvariations(-2-29 Ma) in abyssalhill morphologyalongthe run of the southsegment.The resultsof thisanalysisare comparedandcorrelatedwith analysisof the gravitydataandpreliminarydetermination of faultingcharacteristics basedon HMR1 sidescan data. Principalresultsof thisstudyare:(1) Abyssalhill morphologywithinthe studyregionis stronglyinfluencedby theinside-outside cornergeometryof the mid-oceanridge segments; abyssalhills originatingat insidecomershavelargerrmsheightandcharacteristic width and smallerplan view aspectratiothanthoseoriginatingat outsidecorners.(2) The residualmantle Bouguergravityanomalyis positivelycorrelatedwith intersegment andalong-flow-linevariations in rmsheightandcharacteristic width,andit is negativelycorrelatedwith planview aspectratio. From thisresult,we infer thatlower-relief,narrower,andmoreelongatedabyssalhills are producedwhenthe crustbeinggenerated is thicker. (3) Intersegment variationsin near-axisrms heightnegativelycorrelatewith averagefault densityasdeterminedfrom analysisof HMR1 side scanimagery. 1. Introduction Neumannand Forsyth, 1995]. Many studieshave investigated The processof abyssalhill formationat mid-oceanridgecrests the geologic structureassociatedwith axial faulting and volcanof abyssalhill formation is a poorlyunderstood componentof the ridgesystem.Abyssal ism, which are the primarymechanisms (seeShawand Lin [1993] andTucholkeand Lin [1994] for up-tohills are complexand chaoticin natureand as suchpresenta uniquesetof challengesfor geologicand quantitativecharacteri- dateandthoroughreferencelistsfor Mid-AtlanticRidge(MAR) zation and for physicalmodelingof the processesthat create studies). Few studies,however,haveyet attemptedto formulate them. Quantitative characterization of abyssalhillshaslongbeen physicalmodelsof abyssalhill formationwhich satisfyquantitastudied,but it hasreceivedmuchmoreattentionin recentyears tive and/or geologicalobservationalconstraints[Schoutenand [e.g., Krause and Menard, 1965; Menard and Mammerickx, Denham, 1983; Malinverno and Gilbert 1989; Malinverno and Cowie, 1993; Shaw and Lin, 1993]. This is due in part to the 1967; McDonald and Katz, 1969; Bell, 1975, 1978; Fox and Hayes, 1985; Gilbert and Malinverna,1988;Gaff and Jordan, complicatednatureof abyssalhill faulting and volcanismand in 1988; Malinverno and Packalny, 1990; Gaff, 1991, 1992; part to the dearth, until recently, of high-resolutionoff-axis Malinverna,1991;Hayes and Kane, 1991; Gaff et al., 1993; bathymetryand geophysicaldata which help constrainthe relationshipbetweenabyssalhills and axial conditionsat the time of accretion. This paper continuesthe work of quantitativecharacCopyright1995by theAmericanGeophysical Union. terizationof abyssalhill morphologyandprovidesconstraints for physicalmodelsof abyssalhill formation.Previouscontributions Papernumber95JB02510. 0148-0227/95/95 JB-02510505.00 have relatedabyssalhill morphologyto spreadingrate variations 22,509 22,510 GOFFET AL.: ABYSSALHILLS IN THE ATLANTICOCEAN 49øW 48øW 47øW 46øW 45øW 44øW 49øW 48øW 47øW 46øW 45øW 44øW Figure 1. Simplifiedbathymetry(0.8-km contourinterval)of the surveyarea,with seafloorisochronsshownat 5m.y. intervals. Isochronsare brokenat segmentboundaries,whichare shownin Figure 2. [Menard, 1967; Malinverno, 1991; Hayes and Kane, 1991;Goff, 1991, 1992] andrelatedabyssalhill morphologyto ridgesegmentation at fast spreadingrates [Goff, 1991; Goffet al. 1993]. Our studyexpandson this work by relatingabyssalhill morphologyto ridge segmentationand temporal variability at slow spreading rates. Our principal data set consists of Hydrosweep swath bathymetry[Chayeset al., 1991] collectedon a recentgeophysical cruise to the Office of Naval Research (ONR) Atlantic Natural Laboratory [Tucholke et al., 1992] aboardR/V Maurice Ewing. This surveyobtainednearly 100% Hydrosweepbathymetric coverageas well as gravity, magnetic,and>200% Hawaii MR1 (HMR1) side scansonarcoveragein a region extending from nearthe Mid-Atlantic Ridge (MAR) axiswestward-400 km Near-Axis 49 øW 48 øW 47 øW to 26-30 Ma crust. The ridge-flankcrustin thisregionconsistsof threemain segments(south,center,and north segments;Figures 1 and 2) which have developedbetweenlong-lived, small-offset (_<30 km) ridge axis discontinuities. The southsegmentis the longestlived (> 29 m.y.) and exhibitsa continuousrecordof relatively stableevolution. The centerand north segmentshave had shorterlives (_<20 m.y.), their boundingdiscontinuitieshave small (to zero) offsets(Figure 1), and they are tectonicallymore complexthan the southsegment. We usethe methodologyof Goffand Jordan[1988, 1989] and Goff [ 1990] to estimatestatisticalparameters of abyssalhill morphology in the survey area through formalized inversion of bathymetric data. Principal estimatedparametersinclude the second-order statistical parameters root-mean-square (rms) height, characteristicwidth, and plan view aspectratio. The Data 46 øW SouthSegmentData 45 øW 49øW 48øW 47øW 46øW 45øW 27øN 27 øN 26øN ••t Segment • ,•: / 26øN 25 øN Figure 2a. Map of simplified segmentboundariessurveyedduring the R/V Ewing cruise EW9208. Locationsof the near-axis swathsusedfor parameterestimationare indicatedby line segmentsterminatedby dots. The swathdata setsare sampledfrom gridded bathymetricdata and are 7.5 km in width. The dashed line in the westernpart of the southernsegmentlocatesa smalloffset,secondarydiscontinuity.The long-shortdashedlinesindicatelikely pseudofaults createdby propagatingridges. 25øN Figure 2b. Locations of off-axis swathsused within the south segment(see Figure 2a for details). With minor, short-termexceptions, all offsets at the discontinuitiesboundingthe south segmentare right lateral (as is the offsetat the secondarydiscontinuity at 29-21 Ma); thus inside corner crust is in the southern part of the segment. GOFF ET AL.' ABYSSAL HILLS IN THE ATLANTIC OCEAN betweenthe two mid-oceanridgesgo well beyondsimplescaling (althoughit remainsto be determinedpreciselywhat thoseaddi- analysisis partitionedintotwo substudies (Figure2): (1) analysis of near-axis(< 7 Ma) bathymetricdatafor eachof the three segments,enablinga comparison betweenthe present-dayridgeaxis and the most recently createdabyssalhills and (2) analysisof bathymetricdata within the long-livedsouthsegment,enabling an analysisof temporalvariability(from ~2-29 Ma) in abyssal tional differences are). A significantamountof variation in abyssalhill morphology not attributableto spreadingrate variationalso occurson a regional scale (i.e., tens to hundredsof kilometers),. Studiesby Goff [1991] and Goff et al. [1993] on multibeamdata from the hill morphology. Resultsof the abyssalhill parameterestimationare compared with other geophysicaldata from the survey area, including residualmantle Bouguergravity anomaly (RMBA), and fault identificationsfrom side scansonardata. The RMBA provides an estimateof crustalthickness[e.g.,Kuo and Forsyth,1988;Lin et al., 1990; Lin and PhippsMorgan, 1992; Tucholkeand Lin, 1994], andby inferencethe temperature structureof the spreading axis at the time of crustalaccretion[e.g., $uet al., 1994]. This informationhelps to identify changesin the axial environment that mighthavecausedchangesin abyssalhill morphology.The fault identificationsfrom side scandata provide an independent measureof faulting,oneof the two principalabyssalhill forming flanks of the EPR show that much of this variation is controlled by segmentationof the mid-ocean ridge crest. In particular, abyssalhill rms heighttendsto be largesttowardthe endsof a segment(boundedby first- or second-order discontinuities)and smallesttoward the middle. This patternis well correlatedwith what is interpretedfrom morphologyandgravityconstraintsto be focusedmantleupwellingbeneaththe ridge axis [e.g., Cochranet al., 1993; Scheirer and Macdonald, 1993; Wang and Cochran, 1993], implying an associationof larger-throwfaults with colder axial mantletemperaturesnear segmentends[Goff, 1991; Goff et al., 1993]. The present study demonstratesthat MAR intrasegment abyssalhill variationis primarilycorrelatedto inside-outside cor- processes. ner location [Tucholke et al., 1992; Lin et al., 1992; Tucholkeand Quantitativeanalysisof abyssalhill morphologyhasbegunto place observationalconstraintson the processof generationof seafloorand crustat mid-oceanridge systems.Severalprevious studies[Menard, 1967; Malinverno, 1991; Hayes and Kane; 1991; Goff, 1991, 1992] showedthat abyssalhill rms heightsare generallylargerat slowerspreading rates.Thiscorrelationcanbe attributed,at least in part, to the spreadingrate dependenceof axial thermalstructure[PhippsMorgan et al., 1987; Shawand Lin, 1994] andthe consequent effecton lithosphericstrengthand faulting(seediscussion by Goff [ 1991]). The dependence of rms heighton spreadingrate is, however,only one small part of the relationshipbetweenabyssalhills andthe ridge systemat which they were formed. For example,Goff [ 1991] found that abyssal hills formedat the fast spreadingEastPacificRise (EPR) possess shapeandtexturecharacteristics that are fundamentallydifferent from thoseproducedat the slowspreadingMAR. This observation suggeststhat differencesin abyssalhill forming processes Lin, 1994] ratherthan to focusedmantleupwellinginferredfrom along-axis variations in crustal thickness[e.g., Detrick et al., 1995] or dynamicmodeling[e.g.,Sparkset al., 1993]. Variations in MAR abyssal hill morphology along the run of segments (Figure3) andbetweensegments arefoundto be large,implying strongtemporaland segment-to-segment variations,respectively, in axial processes.These variationsin abyssalhill properties (most notably rms height) are correlatedto variations in the RMBA (consistentwith earlierfindingsby Shawand Lin [ 1993] andNeumannand Forsyth[ 1995]) and,wherenotundulymasked by sedimentcover, to fault densityas interpretedfrom HMR1 side scansonardata. We infer from this a correlationof abyssal hill faulting with the temporal and spatialvariationsin crustal thicknessand/or mantle temperaturestructureat the ridge axis. Both this interpretationandearlieronesfrom EPR studiessuggest that the strengthandthicknessof the lithosphereare the dominant physicalfactorswhich controlthe formationof abyssalhills. • ß,. / 22,511 / I i / / ,% %, / / \ Second-order Discontinuity I \ I .. '• \ X // I / X l \ oc Second-order \ /---I / •1 \ \ / -% ---I / %, / \ Transform Discontinuity \ Figure 3. Schematicsketchdefininggeometryof insidecomercrust(IC, stippledregion), outsidecomer crust (OC), segmentlength and run for ridge segmentsboundedby transformand nontransformdiscontinuitiesin slowspreading crust[from Tucholkeand Lin, 1994]. 22,512 2. Stochastic GOFFETAL.' ABYSSALHILLSIN THEATLANTICOCEAN Characterization theseparameters cannotdo at presentis tell us, for example,that a certainregion is dominatedeither by extensionaltectonismor volcanism. Theseprocesses are dissimilar,andit tics of abyssalhill morphologyand the procedurefor inverting constructional to expectthat the morphologies they produce modelparametersfrom multibeambathymetricdataare detailed is not unreasonable will have characteristic differences which can be detected with by Goffand Jordan [1988, 1989] and Goff[1990]. The method takesas input a swathof multibeamdata,or the griddeddataset, the estimatedstochasticmodel parameters.However,to make and producesas outputestimationsof relevantphysicalparame- this distinctionwe will needto "ground-truth"the statistics,i.e., ters with estimateduncertainties.To obtainoptimaldataorienta- match statistical characteristicswith geologic characteristics. in part with the HMR1 sidescansonar tion relative to segmentboundaries,only gridded Hydrosweep This canbe accomplished data (griddedat 200-m spacing)were used for this analysis. data, from which can be identified steep,unsedimentedscarps Sectionsof the gridded data chosenfor analysisconsistedof interpretedas normal faults (G. Jaroslowet al., manuscriptin elongaterectangularareas7.5 km wide and 50-100 km long, ori- preparation,1995). In addition to the second-orderstatisticalparameterslisted entedsubparallelto plate flow lines (seeFigure2). above, estimatescanbe madeof higher-order(i.e., non-Gaussian) For simplicity,we refer to each elongated-boxsampleof the gridded data set as a "swath"of data (this term also emphasizes statistical parameters such as skewness, tilt, and kurtosis similarityto multibeamsampling). Two constraints are placedon (peakiness). In practice,however,for a regionalstudysuchas parametersrmsheight,linswathschosenfor input:(1) They are straightandlongenoughto presentedhere, only the second-order producewell-resolved estimationsof the statisticalparameters eamentazimuth, characteristicwidth and plan view aspectratio [Goff and Jordan, 1989], and (2) the morphologysampleddoes are well enoughresolvedthat their variationdueto geologicpronot includemajor seamountsor major segmentboundariesas de- cesseswithin the study area exceeds their resolution errors. terminedfrom bathymetry,gravity, and magneticdata. Because Hence this studyfocuseson only theseparameters(with the exceptionof lineamentazimuth,which is the subjectof a separate of differences in decorrelation distances (the distance between two points required for each to be consideredan independent study). The stochastic model used to characterize second-order statis- sample),longerdata recordsare requiredto make reasonableparameterestimationsin the Atlantic comparedto the Pacific. For example, whereasa 20- to 25-km-long swath of data might be sufficient for flanks of the EPR, a 60- to 75-km-long swath is necessaryfor the flanksof the MAR. Sampleswathsare orientedgenerallyalongflow line pathsfor two reasons:(1) This orientationmaximizesresolutionof alongisochronvariability associatedwith ridge segmentation,and (2) the along-flow-linedirectionrepresents the directionof shortest decorrelationdistanceand thereforemaximizesparameterresolution versussamplesize. Whereverheterogeneity existson scales shorterthanthe lengthof the swath,the resultsrepresentan average of the propertiessampled. For example,propagatorpseudofaults are contained within some of the swaths used for this anal- ysis (Figure 2) (B. Tucholke et al., manuscriptin preparation, 1995). These swathslikely containa mixtureof seafloortypes, includingpseudofaultmorphology,abyssalhills adjacentto the pseudofaultwhich are influencedor modifiedby the pseudofault, andabyssalhills unaffectedby the pseudofault.The spatialresolution of the statistical estimation is insufficient 3. Near-Axis Analysis Figure 2a showsthe pathsof individual swathsusedfor parameter estimation in the near-axis data set. These data extend from -1-2 Ma to -6-7 Ma crustal ages. We assumethat the abyssalhills sampledby theseswathswere formedby processes similarto thosethat are now active at zero age. Hencethesedata allow us to correlateabyssalhill characteristics with knownridge axis processes[Tucholke and Lin, 1994; B. Tucholke et al., manuscript in preparation,1995]. RMS Heights RMS heights estimatedfrom near-axis data (Figure 2a) are plottedwith 1-orerrorbarsin Figure4. There are two principal observations.First, variation in rms heightswithin each ridge segmentis asymmetric,with greaterrms heightsskewedtoward the southernparts of the segment. Each of thesesegmentsis to make these distinctions,so care must be taken in interpretationof results. The outputphysicalparameters includethe following: 1. The rmsheightis the averagevariationof bathymetryfrom the meandepth. 2. The lineamentazimuthis trendof abyssalhill lineaments. 3. The characteristic scaleis the dominantlengthof features alonga profile. It is definedby the width of the covariancefunction along the profile direction [Goff and Jordan, 1988]. Perpendicularto the grain, it is calledthe characteristic width of abyssalhills, andparallelto the grain,it is calledthe characteris- Near-Axis • Data NorthSegment ß CenterSegment 0 SouthSegment I tic length. 4. The aspectratio is the characteristic planview shapeof the abyssalhills, defined by the ratio betweencharacteristiclength and width. 5. The Hausdorff (fractal) dimensionis a measureof roughness. It is definedby the asymptoticpropertiesof the covariance functionat smalllag. An increasein fractaldimensionrepresents an increasein roughness, with the limiting casesof 2 and3 correspondingto a randomfield with continuousderivative(Euclidean surface)and one which is spacefilling (Peanosurface),respectively. Theseparametersand their uncertaintiesform an objectivebasis to distinguishdifferencesfrom one regionto another. What IC 25000 ' 25o20' OC 25o40' IC 26o00 ' OC 26o20 ' IC OC 26o40 ' 27000 ' Latitude,Deg N Figure 4. RMS heightof abyssalhills estimatedfrom the nearaxis swaths(Figure 2a), plottedas functionsof averagelatitude overthe lengthof the swath. Vertical barsrepresent1-orerrorestimates. "IC" and "OC" indicate locations of inner comers and outercomers,respectively. GOFF ET AL.: ABYSSALHILLS IN THE ATLANTIC OCEAN boundedby right-lateraldiscontinuities (Figure 1), so the southern portionsare insidecornercrest,and the northernportionsare outsidecomercrest(Figure3). This observationdiffersfrom that of Shaw [1992], who statedthat largerfaults (and consequently largerabyssalhills) tendto occurtowardbothendsof MAR segments. However, the Shaw [1992] study,which did not consider inside-outsidecornerasymmetry,demonstrated this observation by comparinga singleprofile throughthe centerof a segment with a singleprofile over an insidecomer,andthe latterwasused to demonstratethe behaviorof segmentendsin general. RMS heightsdo not increasemonotonically goingfrom outsideto insidecomers. Within the southsegment,rmsheighttendsto reach a minimumtowardthe middle of the segment,similarto observed variationsalongthe EPR flank [Goff, 1991; Goffet al., 1993]. In contrast,within the centersegment,maximum valuesare in the middle of the segment,and the sparselysamplednorth segment showsno clearpattern. The secondprincipalobservationis the large differencein rms heightsbetweenthe southsegmentandthe northandcentersegments. Neglectinginsideandoutsidecomerposition,rmsheights within the southsegmentare approximatelyhalf the rms heights within the center and north segments. This differenceis very striking;southsegmentrms heightsare amongthe lowestvalues seenin near-axis Atlantic abyssalhill morphology,while rms heightswithin the adjacentcentersegmentare amongthe largest [seeGoff, 1991]. Characteristic 22,513 Near-Axis • Data NorthSegment ß CenterSegment 0 SouthSegment IC 25000 ' 25020 ' 25040 ' 26o00 ' 26o20 ' 26o40 ' 2 7o00' Latitude,Deg N Figure 6. Plan view aspectratiosof abyssalhills estimatedfrom the near-axisswaths(Figure2a), plottedas a functionof average latitudeover the lengthof the swath. Vertical barsrepresent error estimates. "IC" and "OC" indicate locations of inner cor- nersandoutercomers,respectively. southsegment,which occursin the middleof the segment,is distinctlylessthanthatof the centerandnorthsegments. Width Characteristic widths estimated from the near-axis data are shownwith 1-(• error barsin Figure 5. This parameteris not as well resolvedas rmsheight,sointerpretation of variationare less clear. Nevertheless, we observethataswith rmsheights,characteristic widths display an inside-outsidecomer asymmetry,and the largest characteristicwidths tend to occur at inside comers (southernpart of segments). The center segmenthas a nearly monotonic increase in characteristic width from outside to inside Plan View Aspect Ratio Plan view aspectratios estimatedfrom the near-axisdata are shownwith 1-tyerror barsin Figure 6. Again, the variationin this parameteris dominatedby inside-outside cornerposition, with the lowestaspectratiostendingto be at insidecorners.For both the southand centersegments,the aspectratio nearthe inside comers approachesunity; i.e., there is no discerniblelineationto the fabric. Neglectingthe samplesnear the insidecorner,however,we find thataspectratiosattaintheirhighestvalues at the centerof the outsidecomerpartof the southsegment. corner. In contrast,the southsegmenthas a well-resolvedminimum in the middle of the segment. Unlike rms heights,thereis little contrastin averagecharacteristic width betweenthe southsegmentand the centerand north ComparisonWith Other GeophysicalConstraints segments. However, the minimum characteristicwidth for the To examinethe causesof variationsin abyssalhill parameters notedabove,here we considercorrelationswith other geophysical and geologicaldata from the samearea and with resultsand Near-Axis Data interpretationsfrom other studies. The mostconsistentvariation in all of our parametersis the inside-outsidecomer asymmetry - • NorthSegment (Figures 4-6), which is well recognizedfrom other studiesof ß CenterSegment 0 SouthSegment slow spreading ridges [e.g., Karson and Dick, 1983; Severinghausand Macdonald, 1988; Blackmanand Forsyth, 1991; Tucholke et al., 1992; Tucholke and Lin, 1994; Escartin and Lin, 1995]. Theseauthorsnote the following asymmetries: inside comer crest is more elevated than outside corner crest; IC 25000 ' OC IC OC I I I I 25o20 ' 25o40 ' 26000 ' 26020 ' IC OC 26040 ' 27o00 ' Latitude,Deg N Figure 5. Characteristicwidthsof abyssalhills estimatedfrom thenear-axisswaths(Figure2a), plottedasa functionof average latitudeoverthe lengthof the swath. Verticalbarsrepresent error estimates. "IC" and "OC" indicate locations of inner cor- nersand outercomers,respectively. faultingat insidecomersis of higheramplitudeandmoreirregular in patternthanat outsidecomers,wherefaultsgenerallyparallel the ridgeaxis;RMBAs are largeroverinsidecomercrestthan over outside corner crest, implying that inside corner crest is thinner; and basementsamplesof inside cornerscontainperidotitesand gabbrosmorefrequentlythanoutsidecomers. These observationssupporta model in which rifting toward segment endsis accommodated in a low-angledetachmentfault, with the insidecomer constitutingthe footwall and the outsidecomer the hangingwall [Dick et al., 1981; Karsonand Dick, 1983; Karson, 1990;Tucholkeand Lin, 1994]. In thisscenario,extrasiveserupting at the ridgeaxis are emplacedprincipallyon the outsidecorner hangingwall, whereasthe insidecomerfootwallis primarily deep crustal or mantle rocks [Karson, 1990; Tucholkeand Lin, 22,514 GOFF ET AL.: ABYSSAL HILLS IN THE ATLANTIC OCEAN 1994]. Both hangingwall and footwall are subjectto secondary, high-anglenormalfaulting. 30 A second primary observation is thelargecontrast in rms 25 heightsbetweenthe southsegmentandthe centerandnorthsegments(Figure4). Characterof the segments at the spreading axis shows nostrong correlation withthisobservation. Eachsegment • 20 isapproximately thesame length and has thesame sense ofoffset (Figures 1and 2).Offset length atboth ends ofthesouth and <m 15 center segments is~10-15 km,while offset length atthenorthern endof thenorthsegment is ~5 km. Depthprofilesof theseseg- ments along theMARaxisareshown inFigure 7. Thesouth and center segmentshave similar, asymmetricprofiles. The north segment profilereaches shallower depths andismoresymmetric. None of thesemorphologicvariationsconvincinglydistinguish the southsegmentfrom the centerandnorthsegments. The corresponding along-axisRMBA field for thesethree segments (J. Lin et al., manuscriptin preparation,1995)is shown in Figure 8. In theseprofiles,we do seemarkeddifferencesbetween the south segmentand the center and north segments. Over the southsegment,thereis a deeplow in the RMBA which is roughlycircularin two-dimensionalplanform(a "bull'seye" gravitylow [Kuo and Forsyth, 1988;Lin et al., 1990]). RMBA signalssuchas theseare interpretedto indicateeither focused mantle upwelling and/or thickenedcrustbeneaththe segment center [Kuoand Forsyth,1988;Lin et al., 1990;Lin andPhipps Morgan, 1992;Sparkset al., 1993]. Both are likely truebecause thickercrustin the middle of the segmentimpliesgreatermelt productionandthuselevatedmantletemperatures [e.g.,Sparkset al., 1993;Suet al., 1994]. In contrast to the southsegment, the 5 South 0 25ø30'N Center Noah Segment Segment Segment 25ø•5'N 26ø60'N 26'i5'N 26'•0'N 26'45'N Latitude Figure 8. ResidualmantleBougueranomaly(RMBA) (J. Lin et al., manuscriptin preparation,1995) sampledalongthe axis of theneovolcanic zoneof theMid-AtlanticRidgeat the easternend of the surveyarea. RMBA is derivedby subtractingfrom the free-airanomalythefollowingpredictivecomponents: (1) theeffects of water-crustand crust-mantleinterfaces,assuminga model crust of 6 km thicknessand water, crust, and mantle densi- tiesof 1.03,2.70,and3.30Mgtm3, respectively, and(2) theeffectsof mantledensitychanges dueto lithospheric cooling[Kuo and Forsyth,1988;Lin et al., 1990] basedon a three-dimensional modelof passivemantleupwelling[PhippsMorganandForsyth, 1988]. center segmentexhibits a much weaker RMBA low at its center (Figure8). The RMBA profileacross thenorthsegment is highly uniqueamongMAR segmentsin that it hasa deeplow (lower eventhanthe southernsegmentanomaly),but the low is centered at the end of the segmentrather than the middle. The causeof this asymmetryis uncertain. The mostconvincing evidencethatthe southsegment is or was in therecentpastdifferentthanbothcenterandnorthsegments is provided by off-axis RMBA [Lin et al., 1992; J. Lin et al., manuscriptin preparation,1995] (Figure9). Over the spansampledby thenear-axisswaths,theRMBA for thesouthsegment is on average~10-14mGal lowerthanfor thecenterandnorthsegments,implyingthat the southernsegmenthasaccretedcrustaveraging~1 km thickerfor the past~7 to 1 Ma. Again,thisimpliesthatthemantletemperature duringaccretion wasgreaterfor the intersegmentcomparisona negative correlationbetween mantle temperature/crustalthicknessand the rms height of abyssalhills, anda positivecorrelationbetweenmantletemperature/crustalthicknessandplan view aspectratio. Fault densityinterpretedfrom the HMR1 sidescansonardata is also markedlydifferentbetweenthe southsegmentand the center and north segments(G. Jar.slow et al., manuscriptin preparation,1995). Through the middle of each segmentand spanningcrustalages~ 1-7 Ma, the southsegmentexhibitsa fault density of 1.18 + 0.07 faults/km, whereasthe center and north Neat-Axis Gravity thesouthsegment thanfor thecenterandnorthsegments during • this spanof time [e.g., Suet al., 1994]. We thereforeinfer from - NorthSegment l, ß CenterSegment 0 SouthSegment il/' -3000 -3500 -4000 IC • 4500 25000 ' South Segment Center Segment North Segment I I I 25020 ' 25040 ' 26000 ' OC I 26020 ' IC oc I 26040 ' 27000 ' Latitude,Deg N Figure 9. RMBA values associatedwith the near-axisswaths (Figure 2a); each plotted value representsthe averageof the 25ø30'N 25',i5'N 26'60'N 26'iS'N 26'•0'N 26ø45'N RMBA field (seeFigure8 for details)calculatedoverthe length Latitude of eachnear-axisswath. Data from J. Lin et al. (manuscript in Figure 7. Bathymetryalong the axis of the neovolcaniczone of preparation, 1995). "IC" and "OC" indicate locationsof inner the Mid-AtlanticRidgeat the easternendof the surveyarea. comersandoutercomers,respectively. -5000 GOFFETAL.: ABYSSALHILLSIN THEATLANTICOCEAN segments have significantly smaller values of 0.76 + 0.03 faults/km and 0.70 + 0.02 faults/km,respectively. Comparing these values with center-of-segment parameter estimates in Figures4-6 implies a negativecorrelationbetweenfault density and both rms heightsand characteristic width, as well as a positive correlationbetweenfault densityandplan-viewaspectratio. SouthSegment4-5 m.y. Averages 4. Analysis of South Segment Figure 2b showspositionsof data swathsusedto analyzethe full run of the southsegment. Thesedata representa -29-m.y. record of a continuouslydevelopedspreadingsegment. There are, however, severalsecond-ordercomplexitiesin the evolution of the segment. At -22 Ma, the plate spreadingdirectionrotated counter-clockwise by -9 ø [Tucholkeet al., 1992; B. Tucholkeet al., manuscriptin preparation,1995). Prior to thistime, a distinct but small-offsetridge axis discontinuitydivided the southsegment in two (dashedlines,Figure 2). This boundaryhasa much weaker bathymetricexpressionthan the primary, larger-offset discontinuities boundingthe southsegment(solidlines,Figure2). As we will show later, this boundaryhad someeffect on abyssal hill morphology,but this effect was subsidiaryto the grossinside-outsidecornerasymmetryof the full southsegment. In the following text, we apply the term "secondarydiscontinuity"to identifythis structure. At youngercrustalagesof ~ 16 Ma and -11 Ma, structurallineamentscrosscutthe abyssalhill fabric and offset the magnetic anomalies(Figures 1 and 2). Thesestructuresappearto be the resultof rift-propagationevents(B. Tucholkeand M. Kleinrock, manuscriptin preparation,1995), andthey mostlikely have some effect on our estimatesof abyssalhill statisticalpropertiesas discussed below. In the following analysis,we averageabyssalhill parameters estimated from all swaths combined within each of a series of 4- to 5-m.y. crustalage ranges(age bins). This averagesout variability associated with inside-outside comerasymmetry,allowing us to investigatetemporalvariationsin abyssalhill morphology. Also, the two southernmost(inside corner) sampleswithin each age bin are excludedfrom the averaging. This maintainsour focus on "normal" abyssalhills, i.e., thosecreated on what is assumedto be full-thicknesscrustwith well-developed,ridge-parallel faultingIn addition, we examine rms height, swath by swath,within eachage bin, to determinealong-isochronvariation of abyssalhill characterwithin the segment. 22,515 ! 30 Ma -49O00' 25 Ma -48030 ' I 30 Ma -49o00 ' -48000 ' 20 Ma -47030 ' 15 Ma -47000' 10 Ma -46030 ' 5 Ma -46000 ' 0 Ma -45ø30 ' -45000 ' I 25 Ma -48030 ' -48000 ' 20 Ma -47030' 15 Ma -47000' 10 Ma -46030 ' 5 Ma -46ø00 ' -45030 ' 0 Ma -45000 ' Longitude,Deg West Figure 10. (a) RMS heightsaveragedwithin4- to 5-m.y. crustal age bins over the southsegmentswaths(Figure 2b) (excluding the southernmost two swaths, i.e., those associated with inside RMS Height corners) and plotted as a function of average longitude. (b) RMBA alongthe run of the southsegment,averaged,aswith rms heights above, to emphasizetemporal variations' each plotted value representsan averageof the RMBA field (seeFigure 8 for details)calculatedboth over the lengthof eachswathand over all swaths(exceptingthe southernmosttwo) in each age bin. Data are from J. Lin et al. (manuscriptin preparation,1995). Vertical bars represent1-rr error estimateson the averages. Horizontal bars are approximatelongitudinalspanof the samples. Crustal agesare indicated. RMS heightsestimatedalongthe run of the southsegment,averagedin the mannerdescribedabove,are plottedin Figure 10a. There is significant variation in observedrms height on time scalesrangingfrom 5 to -15 m.y. The largerms heightvaluesin the 12-7 Ma bin appearto be influencedby the propagatorpseudofault running through this region (Figure 2b). However, the pseudofaultin the 17-12 Ma bin hasno comparableeffect; rms heightshereare amongthe lowestvaluesobserved. RMS heightsfrom individualswathswithin the southsegment are plottedas a functionof swathlatitudein Figure 11 for the set of six agebins. The samplesin the 29-25 and 25-21 Ma agebins crossthe secondarydiscontinuitywithin the westernpart of the segment(Figures2b, 11a, and 1lb). Along-isochronvariationsin rms heightdisplaysimilar characterwithin eachof thesesix age bins. As with the near-axis data (Figure 4), there is a clear asymmetryassociatedwith the inside-outsidecomer positionin the segmentas a whole, with larger rms heightsin the southern, insidecomerpart;thisis particularlynotablein the 29-25 Ma, 2521 Ma, and 21-17 Ma age bins. (The large error bars on these plots are somewhatmisleading,becausewe are making observations regardingthe populationof estimationsratherthan a single estimation.For example,a weightedlinear inversionof the rms heightestimationsdisplayedin Figure 1la, 1lb, or 1lc resolvesa negativeslope(left to right) at -2rr or better.) In the 17-12 Ma agebin the trendis still present(mostlydue to the singlelow rms estimateof the northernmostsample),but high rms heightsalso appearnearthe outsidecomer,possiblybecauseof the propagator pseudofault.Similar propagatoreffectsmay accountfor the lack of any trend in the 12-7 Ma age bin. Thoughthe grossinside-outsidecomerpatternis the same,the southsegmentrms heightsplottedfor agebins 29-25 Ma, 25-21 Ma, 21-17 Ma (Figures 11a-11c) displaya structuredifferentin detail from that of the near-axis rms heights (Figure 4). Proceedingfrom outsideto insidecomer (northto south),the rms heightsincreaseand then decreasemarginallytoward the middle before increasingagain toward the inside corner. Though the error bars are generally large relative to this small "step"in rms 22,516 GOFFET AL.: ABYSSALHILLS IN THEATLANTICOCEAN a 29-25 Ma Secondary Segment Discontinuity IC 25o00 ' 25o20 ' 25o40 ' 26o00 ' 26o20 ' OC 26o40 ' 27o00 ' b oc 25-21 Ma !C 25o00 , 25o20 , 25o40 , Secondary Segment Discontinuity I i 26o00 ' 26o20 ' OC I 26o40 ' 27o00 ' 21-17 Ma IC • 25000 ' OC I I I I I 25020 ' 25040 ' 26000 ' 26020 ' 26040 ' 27o00 ' Latitude,Deg N Figure 11. Swath-by-swathestimatesof rms heightsfor six successive 4- to 5-m.y. crustalagebinsin the south segment,plottedas a functionof swathlatitude. Vertical barsrepresentestimated1-• errors. The spanof swaths whichsamplethe small-offsetsecondary discontinuity with thesouthsegment(seetext)aremarkedby the shaded region. "IC" and "OC" indicatelocationsof innercomersandoutercomers,respectively. GOFF ET AL.: ABYSSALHILLS IN THE ATLANTIC OCEAN 22,517 d 17-12 Ma oc IC 25000 ' I I I I I 25020 ' 25040 ' 26000 ' 26020 ' 26040 ' i I i i 25040 ' 26000 ' 26020 ' 26040 ' 26020 ' 26040 ' 27000 ' e 12-7Ma oc IC I 25000 ' 25020 ' 27000' f 7-2 Ma IC OC I 25000 ' 25020 ' 25040 ' 26000 ' 27000' Latitude,Deg N Figure 11. (continued) height, the consistencyof this feature at all three age bins suggeststhat it is real. In the 29-25 and 25-21 Ma agebins,the step is spatiallycoincidentwith the secondarydiscontinuityidentified by B. Tucholkeet al. (manuscriptin preparation,1995) (Figures 1la and 1lb). The continuation of thisstepstructureinto the 21- 17 Ma agebin (Figure 1l c) suggests that the secondarydiscominuity influencedtopographyin this youngercrest. However, thereis no gravity,magnetic,or bathymetricevidencefor continuation of the boundary(J. Lin et al., manuscriptit preparation, 1995;B. Tucholkeet al., manuscript it preparation,1995). RMS 22,518 GOFFETAL.:ABYSSAL HILLSIN THEATLANTICOCEAN heightswithin the 17-12 Ma age bin (Figure 1l d) displaya behavior similarto that of older agebins,i.e., a southwardincrease in rmsheight,followedby a decrease,andthenan increaseagain toward the inside corner. However, the initial increaseis large and abrupt,andit is followedby a lengthydecreasetowardthe insidecomer;thusit doesnot give the appearance of a "step"in an overall outside-to-insidecomer increasein rms height. This characteris differentfrom thatat the secondary discontinuity, and we infer thatit is relatedto the presenceof thepropagator pseudofault. Characteristic Width and Plan-View Aspect Ratio Characteristicwidths estimatedfrom the southsegmentdata set and averagedfor each 4- to 5-m.y. age bin are plottedin Figure 12a. Though somedetailsdiffer, the grossvariationin thisparameteris similarto thatof rmsheights(Figure10a):Both parameters startoff largein the 29-25 Ma agebin, generallydecreaseto the 25-21 Ma, 21-17 Ma, and 17-12 Ma age bins, increaseto maximum value in the 12-7 Ma agebin, anddecreaseto minimum value in the 7-2 Ma agebin. Individual estimations of characteristic width between 29 Ma and 12 Ma are, like rms heights,generallylargerat insidecomers thanat outsidecomers. However,thereis no obvioussteptoward the middle of the segmentas there appearsto be in the rms heights. The degreeof positivecorrelationbetweenrms height and characteristicwidth is mostobjectivelyquantifiedby the corSouthSegment,4-5 m.y.Averages i I, 30 Ma 30 Ma -49o00 ' , 25Ma I 20Mr 151Ma 110Ma I 5MaI, .48o30 ' -48o00 ' 25 Ma -48o30 ' -48o00 ' -47o30 ' 20 Ma -47030 ' -47o00 ' -46o30 ' 15 Ma -47000 ' -46o00 ' -45o30 ' 5 Ma -46030 ' between these two parametersis p = +0.48 with confidence >99%. Although this coefficient is not very high, it is well resolved,andit representsa minimumvaluebecausethe actualcorrelationis degradedby randomerrorsin bothparameters.Hence it is likely that characteristicwidth and rms heightare responding to similar physicalmechanismswhich control the formationof abyssalhills. Figure 12b plots plan view aspectratios estimatedfrom the southsegmentdata and averagedin the mannerdescribedearlier. The patternexpressedin this plot is similar and oppositeto the plots of averagedrms height and characteristicwidth in Figures 10a and 12a. Over both the near-axis and south segmentdata sets,rms heightand plan view aspectratio have a correlationcoefficientof p = -0.32 with confidence>99%. Overjust the south segmentdata set, the coefficientis p =-0.28 with confidence= 97%. As with the near-axissamples(Figure 6), individual estimationsof plan view aspectratiosover the entire southsegmentare almost uniformly smallestnear the inside corner. To quantify this observation,we calculated the average aspectratio for the southerntwo swaths in each segmentand over every age bin shown on Figure 2, as well as the average over all remaining swaths. The near-inside-cornerswathshave an average aspect ratio of 2.04 + 0.18, and the remaining sampleshave an average aspectratio of 3.70 + 0..21,almosta factor of 2 difference. ComparisonWith Other GeophysicalConstraints ,1, -49o00 ' relation coefficient [see Press et al., 1989]. For either the entire data setor just the southsegmentdata, the correlationcoefficient -46000 ' -45030 ' -45o00' 0 Ma -45000 ' Longitude,Deg West Figure 12. (a) Characteristicwidthsand (b) planview aspectratios averagedwithin each4-5 m.y. crustalagebin overthe south segmentswathsamples(Figure 2b), plottedas a functionof average longitude. Vertical barsrepresent1-rr error estimateson the averages.Horizontalbarsrepresentthe approximatelongitudinal spanof the samples.Crustalagesare alsoindicated. As in our analysisof near-axisabyssalhill morphology,we find a strong(thoughnot perfect) positive correspondence betweenalong-flow-linevariationsin RMBA andrmsheightsover the run of the southsegment(Figures10a and 10b). Figure 10b showsaveragedRMBA over the run of the southsegment(data from Linet al., [1992] and J. Lin et al. (manuscriptin preparation, 1995)). The averagingscheme,which is essentiallyidentical to that usedfor RMS heightsin Figure 10a, subduesRMBA variationsassociatedwith segmentation(i.e., focusedupwelling, inside/outsidecorner asymmetry)while emphasizingvariations associatedwith temporal variability in the accretionaryprocess. The principal difference betweenFigures 10a and 10b is that residualgravity in the 17-12 Ma age bin is large relativeto the correspondingaveragerms height. This may indicate that the propagatorpseudofaultrunningthroughthisregion(Figure2) has moreexpressionin gravity signaturethanit doesin bathymetry. Along-flow-line variations in averaged characteristic width (Figure 12a) and plan view aspectratio (Figure 12b) (with a negativecorrespondence) also displaysimilaritiesto the RMBA plottedin Figure 10b. The RMBA is interpretedprimarily to reflect the thicknessof the crust[e.g., Kuo and Forsyth,1988; Lin et al., 1990; Lin and Phipps Morgan, 1992; Tucholkeand Lin, 1994]: RMBA lows suggestthicker crust, and RMBA highs imply thinner crust. Variations in thicknessof the crust,to a first order, may in turn reflect variations in melt production and hence in mantle temperatureat the axis at the time of emplacement[Lin et al., 1990; Suet al., 1994]. Thus we infer that the quantitative characteristicsof abyssal hills are correlated with temporal variationsin both crustalthicknessand mantle temperatureat the ridge axis;wheremantletemperatureis relativelyhot andthicker crustis produced,abyssalhills generallyhave reducedheightand are narrower and more lineated, and where the mantle temperatureis cooler and thinnercrustproduced,abyssalhills are generallyhigher,wider, and lesslineated. The correlationbetweenRMBA and abyssalhill characteristics can alsobe demonstratedon unaveragedestimates.Over the en- GOFF ET AL.: ABYSSAL HILLS IN THE ATLANTIC OCEAN 22,519 mated sedimentthicknessover the southsegmentas derived by Webb and Jordan [1993]. Webb and Jordan [1993] created a highly innovative approachto modeling sedimentaccumulation correlationis also consistentwith inside-outsidecomer asymme- based on numerical sedimentationalgorithms. Comparing the try (largerrms heightsandthinnerci'ustat insidecorners). modelwith observedbathymetry,they foundthat sedimentloads Though less robust, well-resolvedcorrelationscan also be do not increasemonotonicallywith age within the surveyarea, demonstratedbetweenRMBA and both characteristicwidth (p = but ratherpeak at an estimated50 m in the range20-15 Ma before decreasingto -25 rn at -25 Ma. This observationis consis+0.31, confidence= 99%) andplan-viewaspectratio (p = -0.33, tent with actual accumulations observed in seismic reflection data confidence= 99%). The negativecorrespondence notedearlierbetweenabyssalhill collectedin the sameregionand may be explainedby a hypothestatisticalparametersand fault density(as interpretedfrom the sized deepeningof the calcite compensationdepth in the late HMR1 side scan data) cannot be demonstrated for the entire Miocene [Tucholkeand Vogt, 1979; Webb and Jordan, 1993]. southsegment.Figure 13a plotsaveragefault densityfor each The variationin sedimentload plottedin Figure 13b matches,in agebin calculatedalonga flow line swathwithin the centerof the reverse,the fault densitiesinterpretedfrom sidescansonardataat southsegment.The largeincreasein fault densityfrom the 12-7 crustalages>-7 Ma (Figure 13a). We note that the ~25-m increase in sediment load from ~25 Ma to the7-2 Ma agebinsis consistent with thelargeincreases in rmsheightandcharacteristic widthandthedecrease in apectratio Ma to ~17 Ma (Figure 13b) may partially explain the variationin over the sameage bins (Figure 10a), but theserelationshipsdo rms heightover corresponding age bins (Figure 10a). However, not continuefor older crustalages. H. Webb (personalcommunication, 1993)estimatesthat,evenfor The discrepancyat olderagesmay be relatedto the accumula- the moststronglypondedsedimentswithin the studyarea,an avtion of sediment cover, which can mask smaller faults and thus eragesedimentload of L will causea decreasein topographicrms decreaseobservedfault density. Figure 13b plots averageesti- height at mostby L and more likely by ~L/2. Thereforesedimentationis not likely a significantfactorin the observed~100-m decreasein averagerms heightfrom-29 Ma to ~17 Ma. tire data set, the correlation coefficient between RMBA and rms heightis p = +0.57 with confidence >99%. In additionto reflecting along-flow-linecorrelationsin gravity and rms height,this SouthSegment, 4-5 m.y.Averages 5. Results 30 Ma 25 Ma , 20 Ma I , 15 Ma I 10 Ma • 5 Ma 0 Ma I , ?49o00 ' -48o30 ' .48o00 ' .47o30 ' .47o00 ' -46030 ' -46000 ' -45030 ' -45000 ' and Discussion Principal results of statistical analysis of the Hydrosweep bathymetrywithinthe ONR AtlanticNaturalLab areasfollows: 1. Near-axisintrasegment variationin abyssalhill morphology is dominatedby asymmetrybetweeninside and outsidecorners; abyssalhills producedon inside corner crust have larger rms heightand characteristicwidth and smallerplan view aspectratiosthanabyssalhills producedat outsidecomers. 2. Insideto outsidecomerasymmetryin abyssalhill morphology generallypersistsin crustof the southsegmentover the full rangeof crustalagessampledin the survey. However,within the agebins 29-25 Ma and 25-21 Ma, the outsideto insidecomerincreasein rms heightsis slightlyoffsetby a negativestepnearthe segmentcenter,and this stepappearsto be associated with a secondary, small-offsetdiscontinuity. The stepin rms height persistsinto the 21-17 Ma age bin, althoughthe secondarydiscontinuity has no signaturein gravity, magnetics,or crustal geomorphology. 3. Along-flow-lineandintersegmentvariationsin abyssalhill morphologycorrelatewith variationsin RMBA, consistentwith earlier results from Shaw and Lin [1993] and Neumann and Forsyth [1995]. These variationsappearto correlateto variations in crustalthicknessand thusby inferenceto melt generation and temperatureof the mantle at the time of crustal accretion; lower-relief, narrower, and more lineated abyssalhills are produced when the crust being generatedis thicker (i.e., when the mantle beneaththe axis is hotter). 30 Ma .49o00 ' 25 Ma -48o30 ' .48o00 ' 20 Ma .47o30 ' 15 Ma .47o00 ' 10 Ma .46o30 ' 5 Ma -46o00 ' -45o30 ' 0 M• .45000 ' Longitude,Deg West Figure 13. (a) Fault densitiesaveragedwithin each4- to 5-m.y. crustalage bin in the southsegment(Figure 2b), plotted as a functionof averagelongitude. Fault densitieswere calculated from HMR1 sidescansonarinterpretations alonga flow line corridor approximatelyat the centerof the southsegment. Vertical bars representl-ix error estimateson the averages. Horizontal barsrepresentthe approximatelongitudinalspanof the samples. Crustalagesare alsoindicated. (b) Estimatedaveragesediment thicknessesover the south segment[from Webb and Jordan, 1993]. 4. Near-axisestimatesof averagefault density(as determined from analysisof HMR1 sidescanimagery)computedthroughthe centerof eachsegmentcorrelatewith colocatedestimatesof statistical parameters: negatively with both rms height and characteristic width andpositivelywith planview aspectratio. We discuss these results below. AbyssalHill Segmentation Goff [1991] documentednumeroustextural differencesbetweenMAR- and EPR-producedabyssalhills. To thesedifferenceswe add another:the along-isochronasymmetryof abyssal hill characterin responseto ridge segmentation.Along the EPR, abyssalhills appearto changein responseto the upwellingstruc- 22,520 GOFF ET AL.: ABYSSALHILLS IN THE ATLANTIC OCEAN ture and consequentvariationin mantletemperaturebeneatha ridge segment,with smallerrms height towardthe middle of a segmentandlargerrmsheighttowardthe ends[Goff, 1991;Goff et al., 1993]. AlongnorthernEPR segments, characteristic width and aspectratio tend to find their maximum and minimum, respectively,towardthe middleof the segment[Goff, 1991]. Along southernEPR segments,little or no along-isochronvariationin eithercharacteristic widthor aspectratiois observed[Goffet al., 1993]. In contrastto theseEPR observations,inside-outsidecor- rms heightvariationsseenin the 17-12 Ma and 12-7 Ma agebins (Figures 1ld and 1l e) are relatedto the propagatorpseudofaults traversingthoseregions(Figure 2b). AbyssalHills, Crustal Thickness,and Axial Mantle Temperature As with intrasegment variationsin abyssalhill morphology, intersegmentand along-flow-linevariationscan be correlatedto variations in relative crustal thickness inferred from RMBA ner asymmetriesappearto dominateabyssalhill development within segmentsof the slow spreadingMAR [Tucholkeet al., 1992; Tucholke and Lin, 1994]. This is clearly documentedin along-axis variation in abyssalhill rms height, characteristic width, and aspectratio (Figures4-6). The differentcharacterof MAR andEPR abyssalhills in relation to ridge segmentationis very similar to differencesin other geologicand geophysicalparameters.Only MAR segmentsdisplay significant inside-outsidecorner asymmetryin elevation, gravity signature,apparentcrustalthickness,and bulk composition [Tucholkeand Lin, 1994]. This asymmetryis thoughtto be causedby low-angle, detachmentfaulting in the slow spreading (Figures10 and 12). Both correlationsindicatethatlarger,wider abyssalhills are generatedin thinnercrust,suggesting that lithosphericstrength,as modulatedby crustalthickness,playsa primary role in controllingextensionalfaulting and the resulting abyssalhill size and shape. However, while it is hypothesized that inside-outside comerasymmetryin crustalthicknessis a resultof tectonicthinningby low-angledetachment faulting[Dick et al., 1981; Brown and Karson, 1988; Karson, 1990; Tucholke and Lin, 1994], intersegmentand flow line variationsin crustal thicknesslikely resultfrom spatialvariations[Lin et al., 1990] and temporalvariations[Tucholkeand Lin, 1994], respectively, in axial mantletemperatureandmagmatism.Thusaxial thermal MAR crust [Dick et al., 1981; Karson and Dick, 1983, 1984; structure,either directly or indirectly,may alsocontrolthe charKarson, 1990; Tucholkeand Lin, 1994]. Two implicationsof the acterof abyssalhills. The effectsof crustalthicknessand axial detachment model which will affect inside-outside comer surface mantletemperatureare expectedto be similar:Eitherlower axial morphologycanbe consideredin the contextof our observations. mantle temperature[e.g., McNutt, 1984] or a thinnercrustassoFirst, crustal thickness will have an effect on the mechanical ciatedwith reducedmelt generationin the presenceof lowertemstrengthof the lithosphere,andin particular,thickercrustmay reperatureswould imply greaterlithosphericstrength. It is theresult in weaker lithosphere[e.g., Shaw and Lin, 1994]. From fore difficult to separatethe two effects. simplified theoreticalwork relating faulting and morphologyto An apparentcorrelationof abyssalhill structurewith axial lithosphericstrength[e.g., Malinverno and Cowie, 1993; Shaw thermal structureis also observedwithin segmentsat the EPR, and Lin, 1993], we would predict faults with larger throw and but there is little variationbetweensegments.RMS heightson wider spacing(and consequentlylarger and wider abyssalhills) EPR flanks tend to be smallesttoward the middle of segments on thinned, inside comer crust, which is consistent with our oband increasetoward the ends;this is a patternwhich correlates servations. Second,volcanicedificesare likely to be more comwith the inferred focusingof mantleupwelling beneaththe ridge mon on outsidecorner crust (hangingwall of detachmentfault) axis [Goff, 1991; Cochranet al., 1993; Goff et al., 1993]. MAR than on inside corner crust [Tucholke and Lin, 1994]. Because abyssalhills alsomay be influencedby focusedmantleupwelling rms heightsincreasefrom outsidecomers(more volcanic)to inbeneathridge segments,but if so, this signalis mostlyobscured side comers(less volcanic), it appearsthat roughnesscreatedby by the inside-outsidecornerasymmetry. The small minimum in faulting mustdominateover roughnesscreatedby volcanoes. rmsheightandmoresignificantminimumin characteristic width in the near-axis swathsof the south segment(Figures4 and 5) Division of the South Segment may be one example of abyssalhill formationrespondingto inB. Tucholke et al. (manuscript in preparation, 1995) have side-outsidecorner asymmetrysuperimposedon focusedmantle identifieda small-offset,secondarydiscontinuityspanningcrustal upwelling. We anticipatethat as the averagetemperatureof the ages29-21 Ma within the southsegment. This raisesa question mantle at the axis increases,abyssalhill formationwill respond aboutthe suitabilityof treatingdatafrom bothnorthandsouthof more to mantle upwelling and less to the factorscontrollinginside-outsidecomer asymmetry. this boundaryas observationsfrom a singlesegment. It is interesting,however,that over the full run of the southsegment,there is a positivetrend in rms height going from north to south(from Abyssal Hills and Faulting outside to inside corner; Figures 1 l a-1 ld, 1l e). Thus the south segment, as defined by the larger-offset discontinuitiesat its northernand southernmargins,hasbehavedoverall as a complete inside-outside corner system throughout its history. Superpimosed on the positivetrendof rmsheightin the 29-25 Ma and 25-21 Ma age bins we do seea smallbut consistentnegative stepwhich seemsto be associatedwith the crossingfrom inside to outsidecomer acrossthe secondarydiscontinuity(Figures1l a and 1lb). Thus it appearsthat even suchminor offsetsmay influence lithosphericpropertiesthat control faulting and abyssal hill formation,althoughthe influenceis reducedas offsetlength is reduced. The smallnegativestepin rmsheightwithin the interiorof the southsegmentpersiststhroughcrustalages21-17 Ma, despitethe lack of any discordancein crustal geomorphology,gravity, or magnetics. Whether this persistenceis somehowrelatedto the prior secondarydiscontinuityor to someotherunrelatedfactoris presentlyunknown. We believethat the complexstructureof the Two kinds of circumstantialevidence have been presentedin the resultsand above discussionfavoring the assertionthat normal faulting, as opposedto volcanic edifice formation,is the principalabyssalhill formingprocessat slowspreadingrates:(1) Assumingthe validityof the detachment model(in particular,that volcanic edifices are more common on outside comer crust), in- side-outsidecomerasymmetryin abyssalhill characteristics (rms heightsare larger on inside corner crust) can only be accounted for if faulted morphologyis much larger than volcanicedifice morphology;and (2) fault density(as interpretedfrom sidescan sonardata)computedwithin the centerof eachsegmentcorrelates with colocatedestimatesof statisticalparameters(negativelywith both rms height and characteristicwidth and positively with aspectratio) in the near-axis (low sediment)regions. We note that the latter evidencecould have other explanations. For example,in the eventthat volcanismwere dominant,faultingcould be controlledby the locationsof volcanicedifices,andthe corre- GOFF ET AL.: ABYSSAL HILLS IN THE ATLANTIC OCEAN lationbetweenstatisticalparametersandfault densitycouldbe an artifice of the separatecorrelationof each with volcanicedifice formation. However, we believe the first evidenceis a strong constraintand sufficientfor our interpretation;the secondevidencedoesnot contradictthisinterpretation. Whether MAR abyssalhills are generatedpredominantlyby faultingor volcanicprocesses hasbeendebatedvigorouslyin the literature(seealsoGoff[1991] andShawand Lin [1993] for more extensive discussionand references). Rona et al. [1974] concludedfrom a studyof abyssalhills eastof the MAR near 23øN that both volcanicand tectonicprocesses play an importantrole in their formation. Kong et al. [1988] and Pockalnyet al. [1988] observedthat a large volcanicedifice which formed in the rift valley near 23øN was similar in size and shapeto abyssalhill structuresoff axis and thusproposedthat abyssalhills in this region are extinct volcanicedificeswhich have been rafted away. However, beforea volcanicedifice canbe transferredto the ridge flank, it mustpassoverthe rift mountains,andin the process,it is subjectedto extensive normal faulting with individual fault throws of the order of hundreds of meters [Macdonald and Luyendyk,1977]. Smith and Cann [1990] suggestthat such faulting so drasticallyalterssurfacemorphologythat volcanic edificesconstructedwithin the medianvalley can no longerbe identified as distinct featuresfollowing the faulting process. While many volcanicfeaturesmay exit the rift valley relatively intact [Jaroslow et al., 1994], theseearlier studies,like our data, are consistent with the assertionthatnormalfaultingis the dominantabyssalhill formingmechanism. 22,521 sideandoutsidecorners,abyssalhill morphologyalsoresponds to temporal and intersegment variations in crustal thickness and/or thermal structure at the axis. 4. Assumingthe validity of the detachmentmodel, insideoutsidecomerasymmetryin abyssalhill characteristics canonly be accounted for if faultedmorphologyis muchlargerthanvolcanicedifice morphology.This interpretationis consistentwith theobservation thatin near-axisdata,faultdensity(asdetermined from analysisof HMR1 sidescanimagery)correlates negatively with variationsin both rms height and characteristicwidth and positivelywith variationsin aspectratio. However,thesecorrelationsare not generallyobservedoff axis alongthe mn of the southsegment.This maybe attributableto sedimentcover. Acknowledgments. This researchwas supportedby the Office of Naval Researchunder grants N00014-92-J-1214, N00014-94-I-0197, N0014-90-J-1621, andN0014-94-1-0466. G.E.J.was supported by ONR AASERT grant N00014-93-I-1153, and additional supportto J.L. was providedby NSF grant OCE93-00708. We are indebtedto M. Orr, J. Kravitz, and M. Badiey for their supportin developmentof the research plan for the ONR Atlantic Natural Laboratory. We also thank T. Reed and P. Lemmond for their efforts in processingand display of the Hydrosweepdata on which this study was basedand to the crew and scientific party on Ewing Cruise 9208 for their assistancein data acquisitionand processing. Thoughtful reviews from M. Cannat, J. Karson,and an anonymousassociateeditor on an earlier draft helped improvethismanuscriptconsiderably.UTIG contribution1154. WHOI contribution 9004. References 6. Conclusions Bell, T. H., Statisticalfeaturesof sea-floortopography,Deep Sea Res., Our principalresultsof the statisticalanalysisand relatedinterpretationsare as follows: 1. As suggested by Tucholkeet al. [1992] and Tucholkeand Lin [ 1994],abyssalhill morphologyis stronglyinfluencedby the inside-outsidecorner geometryof slow spreadingmid-ocean ridge segments.This is evidentin our near-axisdata sets,where abyssalhills originatingat insidecomershavelargerrmsheight andcharacteristic width,andsmallerplanview aspectratiosthan abyssalhillsoriginatingat outsidecomers(Figures4-6). Insideoutsidecomerasymmetryin abyssalhill morphologymay be a directconsequence of tectonicthinningof insidecomercrestby low-angledetachment faulting[Dick et al., 1981;Karson,1990; Tucholkeand Lin, 1994]. 2. Thereis anoverallpositivetrendin rmsheightgoingfrom outsideto insidecornercrestover the full 29-m.y. mn of the southsegment.Within thispositivetrendis a slightnegativestep in rmsheightnearthe middleof the segmentin the 29-25 Ma, 2521 Ma, and 21-17 Ma agebins (Figures11a-11c). In the 29-25 Ma and25-21 Ma agebins,thisstepcorrelateswith theoutsideto insidecomer transitionassociatedwith a small-offsetsecondary discontinuity(Figures2, 1la, and 1lb). Thusjust asthereis a hierarchyof offsetmagnitudeandbathymetricexpression of ridge axis discontinuities,there may be a correlativehierarchyof effectson development of abyssalhills adjacentto the discontinuities. 3. ResidualmantleBougeranomaliescorrespond positively with both intersegmentand flow line variationsin abyssalhill rmsheightandcharacteristic width andnegativelywith planview 22, 883-892, 1975. Bell, T. H., Mesoscaleseafloor roughness,Deep Sea Res., 26A, 65-76, 1978. Blackman,D. K., and D. W. Forsyth,Isostaticcompensation of tectonic featuresof the Mid-Atlantic Ridge:25ø-27ø30'S,J. Geophys.Res.,96, 11,741-11,758, 1991. Brown, J. R., and J. A. Karson,Variationsin axial processes on the MidAtlanticRidge:The medianvalley of the MARK area,Mar. Geophys. Res., 10, 109-138, 1988. Chayes,D. N., D. W. Caress,W. B. F. Ryan, A. Malinverno,W. Menke, and C. Keely, Status of Hydrosweepon the R/V Ewing: System operationand data processing(abstract),Eos Trans. AGU, 72 (44), Fall Meet. suppl.,488, 1991. Cochran,J. R., J. A. Goff, A. Malinverno,D. J. Fornari,C. Keeley, and X. Wang, Morphology of a "superfast"mid-oceanridge crest and flanks:The East PacificRise, 7ø-9øS,Mar. Geophys.Res., 15, 65-75, 1993. Detrick, R. S., H. D. Needham,and V. Renard,Gravity anomaliesand crustalthicknessvariationsalong the Mid-Atlantic Ridge between 33øNand40øN, J. Geophys.Res.,100, 3767-3787, 1995. Dick, H. J. B., G. Thompson,andW. B. Bryan,Low-anglefaultingand steady state emplacement of plutonic rocks at ridge-transform intersections(abstract),Eos, Trans. AGU, 62, 406, 1981. Escartin, J., and J. Lin, Ridge offsets, normal faulting, and gravity anomaliesof slow spreadingridges,J. Geophys.Res.,100, 6163-6177, 1995. Fox, C. G., and D. E. Hayes, Quantitativemethodsfor analyzingthe roughness of the seafloor,Rev. Geophys.,23, 1-48, 1985. Gilbert, L. E., and A. Malinverno, A characterizationof the spectral densityof residualoceanfloor topography,Geophys.Res.Lett., 15, 1401-1404, 1988. Goff, J. A., Stochasticmodelingof seafloormorphology,Ph.D. thesis, 266 pp., Mass. Inst. of Technol./Woods Hole Oceanogr. Inst., Cambridge,1990. aspectratio (Figures 10 and 12). Lower-relief, narrower, and Goff, J. A., A global and regional stochasticanalysisof near-ridge abyssalhill morphology,J. Geophys.Res.,96, 21,713-21,737,1991. more lineatedabyssalhills are producedwhen the crestbeing Goff, J. A., Quantitativecharacterizationof abyssalhill morphology generatedis thicker (i.e., when the mantle beneaththe axis is hotalongflow lines in the Atlantic Ocean,J. Geophys.Res., 97, 9183ter), andhigher,wider, lesslineatedabyssalhills are produced 9202, 1992. when the crestbeing generatedis thinner(i.e., when the mantle beneaththe axis is colder). This suggeststhat in additionto respondingto differencesin mechanisms of crustalaccretionat in- Goff, J. A., and T. H. Jordan, Stochasticmodeling of seafloor morphology:Inversionof Sea Beam data for second-orderstatistics,J. Geophys.Res., 93, 13,589-13,608, 1988. 22,522 GOFFETAL.: ABYSSALHILLSIN THEATLANTICOCEAN Goff, J. A., and T. H. Jordan, Stochastic modeling of seafloor morphology: Resolutionof topographic parameters by SeaBeamdata, IEEE J. OceanicEng., 14, 326-337, 1989. Goff, J. A., A. Malinvemo, D. J. Fomari, andJ. R. Cochran,Abyssalhill segmentation: quantitative analysisof theEastPacificRiseflanks7øS9øS,J. Geophys.Res.,98, 13,851-13,862,1993. Hayes,D. E., andK. A. Kane,The dependence of seafloorroughness on spreading rate,Geophys. Res.Lett.,18, 1425-1428,1991. Jaroslow,G. E., D. K. Smith, and B. E. Tucholke,The off-axisrecordof volcanism in the western North Atlantic Ocean (abstract), EosTrans. PhippsMorgan, J., E. M. Parmentier,and J. Lin, Mechanismsfor the origin of mid-oceanridge axial topography:Implicationsfor the thermaland mechanicalstructureof accretingplate boundaries,J. Geophys.Res.,92, 12,823-12,836, 1987. Pockalny,R. A., R. S. Detrick,andP. J. Fox, Morphologyandtectonics of the Kane transformfrom Sea Beam bathymetrydata,J. Geophys. Res., 93, 3179-3193, 1988. Press,W. H., B. P. Flannery,S. A. Teukolsky,and W. T. Vetteding, Numerical Recipes,701 pp., CambridgeUniv. Press,Cambridge, Mass., 1989. Rona, P. A., R. N. Harbison,and S. A. Bush,Abyssalhills of the eastern AGU, 75 (44), Fall Meet. suppl.,660-661, 1994. central North Ariantic, Mar. Geol., 16, 275-292, 1974. Karson,J. A., Seafloorspreading on theMid-AtlanticRidge:Implications for thestructure of ophiolites andoceaniclithosphere produced at slow Scheirer,D. S., and K. C. Macdonald, Variation in cross-sectionalarea of the axial ridgealongtheEastPacificRise:Evidencefor the magmatic spreading environments, in Proceedings of theSymposium TROODOS budgetof a fast spreadingcenter,J. Geophys.Res.,98, 7871-7885, 1987, edited by J. Malpas et al., pp. 547-555, Geol. Surv. Dep., 1993. Nicosia,Cyprus,1990. Karson, J. A., and H. J. B. Dick, Tectonics of ridge-transform Schouten,H., and C. Denham, Deconvolutionof faultednormal seafloor in the North Atlantic FAMOUS area (abstract),Eos Trans. AGU, 64, intersections at the KaneFractureZone,Mar. Geophys. Res.,6, 51-98, 1983. 314, 1983. J.P., andK. C. Macdonald,High insidecomersat ridgeKarson,J. A., and H. J. B. Dick, Deformedandmetamorphosed oceanic Severinghaus, transformintersections, Mar. Geophys.Res.,9, 353-367, 1988. cruston the Mid-AtlanticRidge,Offolin',9, 279-302, 1984. Shaw,P. R., Ridge segmentation, faulting,and crustalthicknessin the Kong,L. S. L., R. S. Detrick,P. J. Fox, L. A. Mayer,andW. B. F. Ryan, Atlantic Ocean,Nature, 358, 490-493, 1992. The morphology andtectonics of theMARK areafromSeaBeamand of variationsin faulting SeaMARC I observations (Mid-AtlanticRidge23ø N), Mar. Geophys. Shaw,P. R., andJ. Lin, Causesandconsequences styleat the Mid-AtlanticRidge,J. Geophys. Res.,98, 21,839-21,851, Res., 10, 59-90, 1988. 1993. Krause,D.C., and H. W. Menard, Depth distributionand bathymetric Shaw, W. J., and J. Lin, Three-dimensionalmodelling of thermoclassification of someseafloorprofries,Mar. Geol.,3, 169-193, 1965. Kuo,B.-Y., andD. W. Forsyth,Gravityanomalies of theridge-transform mechanicalstructureof oceanicspreadingcenters:Effectsof alongaxisthicknessvariationsassociated with segmentation andhot spots systemin the South Atlantic between 31 and 34.5øS: Upwelling (abstract),EosTrans.AGU, 75 (16), SpringMeet. suppl.,331, 1994. centersand variationsin crustalthickness,Mar. Geophys.Res., 10, Smith, D. K., and J. Cann, Hundredsof small volcanoeson the median 205-232, 1988. valley floor of the Mid-AtlanticRidgeat 24ø-30øN,Nature,348, 152Lin, J., andJ. PhippsMorgan,The spreadingrate dependence of three155, 1990. dimensional mid-oceanridgegravitystructure,Geophys. Res.Lett.,19, 13-16, 1992. Sparks, D. W., E. M. Parmentier, and J. Phipps Morgan, Three dimensional mantleconvection beneatha segmented spreading center: Lin, J. G., M. Purdy, H. Schouten,J.-C. Sempere,and C. Zervas, Implications for along-axisvariations in crustalthickness andgravity, Evidencefrom gravitydatafor focusedmagmaticaccretionalongthe J. Geophys.Res.,98, 21,977-21,995,1993. Mid-AtlanticRidge,Nature, 344, 627-632, 1990. Lin, J., B. E. Tucholke, M. C. Kleinrock, and J. A. Goff, Variations in Su, W., C. Z. Mutter, J. C. Mutter, and W. R. Buck, Some theoretical predictionson the relationshipsamong spreadingrate, mantle crustalfaultingandmagmaticaccretionalongtheMid-AtlanticRidge temperature,and crustalthickness,J. Geophys.Res.,99, 3215-3227, andoff-axis:Resultsfrom the westernflank of the MAR at 25-27 deg 1994. N (abstract),Eos Trans.AGU, 73 (43), Fall Meet. suppl.,538, 1992. Macdonald,K. C., andB. P. Luyendyk,Deep-towstudiesof thestructure Tucholke,B. E., andJ. Lin, A geologicmodelfor the structureof ridge segments in slow-spreading oceancrust,J. Geophys. Res.,99, 11,937of the Mid-Atlantic Ridge near lat 37ø N., Geol. $oc. Am. Bull., 88, 621-636, 1977. 11,958, 1994. Malinvemo,A., Inversesquare-root dependence of mid-ocean-ridge flank Tucholke, B. E., and H. Schouten,Kane fracturezone, Mar. Geophys. Res., 10, 1-39, 1988. roughness on spreading rate,Nature,352, 58-60, 1991. Malinvemo, A., and P. A. Cowie, Normal faultingandthe topographic Tucholke,B. E., andP. R. Vogt, WesternNorthAtlantic:Sedimentary evolutionandaspectsof tectonichistory,Initial Rep.Deep SeaDrill roughnessof mid-oceanridge flanks, J. Geophys.Res.,98, 17,92117,935, 1993. Malinvemo, A., and L. E. Gilbert, A stochasticmodel for the creation of abyssalhill topographyat a slow spreadingcenter,J. Geophys.Res., 94, 1665-1675, 1989. Malinvemo, A., and R. A. Pockalny, Abyssal hill topographyas an indicatorof episodicityin crustalaccretion,Earth Planet. $ci. Lett., 99, 154-169, 1990. McDonald,M. F., andE. J. Katz, Quantitative methodfor describing the regionaltopographyof the oceanfloor, J. Geophys.Res.,74, 25972607, 1969. McNutt, M. K., Lithosphericflexureandthermalanomalies, J. Geophys. Proj., 43, 791-825, 1979. Tucholke, B. E., J. Lin, and M. C. Kleinrock, Crustal structure of spreading segments on thewesternflankof theMid-AtlanticRidgeat 25ø25'Nto 27ø10'N(abstract),Eos Trans.AGU, 73 (43), Fall Meeting Suppl.,537-538, 1992. Wang, X., and J. R. Cochran,Gravity anomalies,isostasy,and mantle flow at the East Pacific Rise crest, J. Geophys.Res., 98, 19,50519,532, 1993. Webb,H. F., andT. H. Jordan,Quantifyingthedistributionandtransport of pelagicsedimentson youngabyssalhills, Geophys.Res.Lett., 20, 2203-2206, 1993. Res., 89, 11,180-11,194, 1984. Menard,H. W., Sea floor spreading,topographyand the secondlayer, J. A. Goff, Universityof TexasInstitutefor Geophysics,8701 North Science,157, 923-924, 1967. 78759-8397. (e-mail: Menard, H. W., and J. Mammerickx,Abyssalhills, magneticanomalies MoPac Expressway, Austin, TX [email protected]) and the East Pacific Rise, Earth Plan. $ci. Lett., 2, 465-472, 1967. G. E. Jaroslow, J. Lin, and B. E. Tucholke, Woods Hole Neumann,G. A., andD. W. Forsyth,High resolutionstatistical estimation Institution,WoodsHole, MA 02153. of seafloormorphology:Obliqueand orthogonalfabricon the flanks Oceanographic M. C. Kleinrock, Box 6214 Station B, Vanderbilt University, of the Mid-AtlanticRidge,34ø-35.5øS, Mar. Geophys. Res.,in press, 1995. PhippsMorgan, J., and D. W. Forsyth, Three-dimensionalflow and temperatureperturbationsdue to a transformoffset:Effectson oceanic anduppermantlestructure, J. Geophys. Res.,93, 2955-2966,1988. Nashville, TN 37235. (ReceivedOctober3, 1994;revisedAugust3, 1995; accepted August10,1995)
© Copyright 2026 Paperzz