Breaking into the plate: A 7.6 Mw fracture-zone earthquake adjacent

to the Central Indian Ridge

DelWayne R. Bohnenstiehl, Maya Tolstoy, and Emily ChappLamont-Doherty Earth Observatory of Columbia University, Palisades, New York, USA

Received 3 November 2003; revised 20 December 2003; accepted 30 December 2003; published 31 January 2004.

[1] On 15 July 2003 an extremely large (7.6 Mw) strike-slip earthquake initiated on or near the end of an activetransform along the northern Central Indian Ridge. Theevent propagated away from the plate boundary along thetypically inactive fracture zone, with a sense of slip thatopposes the active transform slip direction. Seismically andhydroacoustically determined aftershock locations delineatea 210 ± 25 km long mainshock rupture. The seismic momentand rupture dimensions imply a stress drop of 4.5–6.5 MPaand a mean slip of 3.0 ± 0.5 m. The largest aftershock(5.6 Mw) occurred on the active portion of a neighboringtransform at a distance of �160 km, where mainshock-induced static stress changes are predicted to promote failure.Near-axis fracture-zone earthquakesmay promote and inhibitridge-parallel diking along different spreading segments,perhaps contributing to inter-segment variability in the rateand asymmetry of spreading. INDEX TERMS: 3035 Marine

Geology and Geophysics: Midocean ridge processes; 3040 Marine

Geology and Geophysics: Plate tectonics (8150, 8155, 8157, 8158);

7230 Seismology: Seismicity and seismotectonics; 7215

Seismology: Earthquake parameters; 8150 Tectonophysics: Plate

boundary—general (3040). Citation: Bohnenstiehl, D. R.,

M. Tolstoy, and E. Chapp (2004), Breaking into the plate: A

7.6 Mw fracture-zone earthquake adjacent to the Central Indian

Ridge, Geophys. Res. Lett., 31, L02615, doi:10.1029/

2003GL018981.

1. Introduction

[2] Mid-ocean ridge plate boundaries are typically nar-row, with seismicity clustered spatially within 10–20 km ofthe spreading axis [Sykes, 1967]. More diffuse zones ofdeformation, however, have been recognized in some oce-anic regions. One such area occurs along the boundarybetween the Indian and Australian (or Capricorn) Plateswithin the near-equatorial regions of the Indian Ocean Basin[Wiens, 1985; Wiens et al., 1985, Royer and Gordon, 1997].It has been proposed that the relative motion of the Indianand Australian Plates can be described by a pole of rotationnear 5�S, 74�E, resulting in diffuse north-south compressionto the east of the pole and diffuse north-south extensionbetween the pole and the Central Indian Ridge (CIR) to thewest [Wiens et al., 1985; DeMets et al., 1994]. This letterdescribes a large magnitude fracture-zone earthquake withinthis western zone of extension (Figure 1) and explores itsimpacts on the CIR system.

2. Synopsis of Mainshock-Aftershock Activity

[3] On 15 July 2003, a 7.6 Mw earthquake occurredadjacent to the northern CIR, the boundary separating theSomalian Plate from the Indian and Australian Plates. Theevent’s epicenter lies near the eastern edge of a relativelyshort (�60 km long) right-stepping transform that offsetsnorthwest-trending spreading segments diverging at a full-rate of 36 mm/yr (Figure 2) [Drolia et al., 2000]. Since1976, only 12 shallow (<40 km depth) strike-slip events(null axis plunge >45�) of equal or larger magnitude havebeen recorded within the global Harvard Centroid MomentTensor (HCMT) catalog.[4] A total of 80 locatable aftershocks occurred in the

71 days following the mainshock. The rate of aftershockproduction approximates the modified Omori law [Utsu etal., 1995], with a decay constant ( p-value) of 0.95 ± 0.12(1s) estimated from the occurrence times of the largest28 aftershocks having Reviewed Event Bulletin (REB)body-wave magnitudes �4.1 [see Granville et al., 2002].This p-value is consistent with the global median of 1.1[Utsu et al., 1995] and values reported for aftershocksequences on the Discovery (�130 mm/yr), Siqueiros(�110 mm/yr) and Western Blanco (�60 mm/yr) Trans-forms in the Pacific Ocean [Bohnenstiehl et al., 2002].[5] Together, the seismically- and hydroacoustically-

derived epicenters (see Section 3.0) show a linear band ofaftershocks extending 210 ± 25 km off-axis along the fracturezone extension of a right-stepping transform (Figure 2). Theseaftershock events are interpreted as delineating the length (L)of the mainshock rupture. As the well-constrained epicenterof the mainshock lies to the southwest of its moment centroidposition and near the southwestern edge of this aftershockzone, the rupture appears to have propagated from the plateboundary—breaking into the Indian Plate along the fracturezone (Figure 2). The largest aftershock, which appears to be atriggered event, is two magnitude units smaller than themainshock (5.6 Mw). It occurs�12.3 days into the sequenceon the active portion of a neighboring transform (Figure 2).[6] The moment tensor solution for the mainshock indi-

cates dominantly right-lateral slip with a near vertical nodalplane striking parallel to the aftershock trend. The right-lateral sense of slip opposes the left-lateral slip along theadjacent active transform and is consistent with reactivationdue to the predicted north-south extension of �3 mm/yracross the diffuse Indian-Australian Plate boundary [Wiens,1985; Wiens et al., 1985]. The initiation of the mainshockrupture immediately adjacent to the CIR axis may reflectdifferential thermal stresses [e.g., Wiens and Stein, 1984],the weakness of this youngest portion of the fracture zone orthe concentration of stress due to the notch-like structure of

GEOPHYSICAL RESEARCH LETTERS, VOL. 31, L02615, doi:10.1029/2003GL018981, 2004

Copyright 2004 by the American Geophysical Union.0094-8276/04/2003GL018981$05.00

L02615 1 of 4

plate boundary’s ridge-transform geometry. The focal mech-anism and location of a 5.5 Ms earthquake that occurred in1973 suggest that this fracture zone, or perhaps its southernneighbor, may have been activated previously [Wiens andStein, 1984] (Figure 1). Although near-axis seismicity isconcentrated within the 5–8�S region, these events indicatethat a zone of significant moment release extends as farnorth as 1�S.

3. Data and Hydroacoustic Location Methods

[7] Teleseismically-derived location parameters weretaken from the REB and Preliminary Determination ofEpicenters (PDE) seismic catalogs. Regional seismic phases(Pn, Sn), as well as oceanic T-waves, also were recorded at atriad of hydroacoustic sensors located to the north of theDiego Garcia atoll, �500 km southeast of the mainshock

rupture (Figure 1). To verify the accuracy of the teleseismiclocations, single-station locations were computed using T-and Pn-arrival information. T-wave arrival time differences,derived from waveform cross-correlation, were used in aplane wave fitting inversion to determine the horizontalslowness components and estimate the back-azimuth to theepicenter [Del Pezzo and Giudicepietro, 2002]. Distanceswere calculated using Pn-T travel time differences, assum-ing upper mantle velocities of 8.1 km/s and T-wave prop-agation speeds of 1.485 km/s. Although the length of themainshock’s hydroacoustic coda obscured Pn arrivals dur-ing the early part of the sequence, 45 of the 80 events in theseismic catalog could be located using this method.

4. Rupture Dimensions and Scaling

[8] The downdip width (W) of the rupture can be con-strained by the near-axis thermal structure [e.g., Engeln etal., 1986]. Slip inversion studies focused on transform and

Figure 1. Moment tensor and historical focal mechanismsolutions. The CIR plate boundary (thick gray line) ismarked by normal-faulting earthquakes on planes strikingparallel to the ridge axis (cyan) and left-lateral strike-slipearthquakes associated with the active transforms (magenta).Off-axis activity dominated by right-lateral events onreactivated fracture zones (yellow) and normal-faultingearthquakes on east-west trending planes (blue), consistentwith north-south extension associated with a diffuse Indian-Australian Plate boundary. Earthquakes discussed in text arelabeled: 1- Mainshock, 15 July 2003, 7.6 Mw; 2-Largestaftershock, 28 July 2003, 5.6 Mw; 3- Historical event thatmay have occurred on same fracture zone as the mainshock,17 November 1973, 5.5 Ms; 4- Historical event on the activetransform extension of the mainshock rupture, 17 March1979, 5.6 Mw. Solid black dots show the aftershocklocations from Figure 2. Contour interval 1000-m. Triangleshows the Diego Garcia North (DGN) hydrophone station.

Figure 2. The well-located (±10 km, 95% confidencelevel) epicenter of the 7.6 Mw mainshock is shown as asolid white circle. Solid blue dots show the epicenters of80 aftershocks located by land-based seismic networks.Solid yellow dots show 45 independently derived T-waveepicenters, a subset of those located seismically. HCMTsolutions are shown with lines drawn to centroid locations.Together these datasets define a 210 km long band (thinwhite line) of seismicity that extends from the CIR plateboundary (gray line) along the fracture-zone (FZ) trace. Themedian error ellipse for seismic earthquake locations isshown in the lower right hand corner. Individual seismicerror ellipses and corresponding single station T-wavelocations (indicated by thin red lines) are shown foraftershocks that appear to be associated with ‘‘off-fault’’structures. The largest aftershock is well constrained by theseismic and T-wave locations to lie on transform T3 andexhibits a left-lateral sense of slip that is consistent with theactive transform slip direction. Epicenters to the south of themainshock may align with a neighboring fracture-zonetrace; however, the single station T-wave locations do notconfirm this.

L02615 BOHNENSTIEHL ET AL.: BREAKING INTO THE PLATE L02615

2 of 4

off-axis strike-slip activity have shown that most seismicslip is constrained above the depth of the 600�C isotherm,which corresponds to the approximate onset of plasticdeformation within the oceanic lithosphere [Engeln et al.,1986; Abercrombie and Ekstrom, 2001]. Figure 3 shows athermal cross-section (half-space model) along the activetransform and fracture-zone trace on the eastern flank of theridge axis. The predicted thermal structure allows for asignificant along-strike increase in the depth of rupture. Thisgeometry is extreme relative to predicted rupture geometrieson active transforms, where increasingly younger and olderlithosphere are juxtaposed [Engeln et al., 1986; Abercrombieand Ekstrom, 2001] or those within the interior of the plate,where a nearly horizontal thermal structure is predicted[Wiens and Stein, 1984].[9] Using the aftershock data to define L and the predicted

thermal structure to constrain W, the rupture area (A) isestimated to be between 2300 and 3200 km2. Given theobserved seismic moment (Mo) of 2.87 � 1020 N-m andthe relationship Mo = mdA, where d is the mean slip and m theshear modulus (35GPa), an average fault slip of 3.0 ± 0.5m isestimated. Across the length of its rupture, this slip eventaccommodates a north-south extension that is equivalent to�700 years of predicted Indian-Australian plate motion atthese longitudes.[10] The rupture dimensions imply a stress drop of 4.5–

6.5 MPa for this event, which is intermediate between typicalinterplate (�3Mpa) and intraplate (�10Mpa) stress drops, asproposed based on a global compilation [Kanamori andAnderson, 1975]. Recent oceanic intraplate events withinthe Wharton Basin (two 7.9 Mw earthquakes in June, 2000)and the Antarctic Plate (8.1 Mw earthquake in March, 1998)exhibit somewhat higher stress drops of �10–20 MPa[Abercrombie et al., 2003] and �16 MPa [Antolik et al.,2000], respectively. It has been suggested that lower stressdrops in the interplate setting might reflect less fault healingbetween more frequent ruptures [Kanamori and Anderson,1975]. Although the 15 July mainshock does not rupture theactive CIR plate boundary, there is an anomalous level of

near-axis earthquake activity associated with the youngfracture zones within the area (Figure 1) [Wiens et al., 1985;Radha Krishna et al., 1998]. Therefore, somewhat lowerstress drops might be expected for CIR fracture-zone earth-quakes, relative to the Wharton Basin events, which occuralong much older fracture zones (�70 Ma), or the AntarcticPlate event, which is not associated with any obvious struc-tural weakness.

5. Impacts on the Central Indian Ridge PlateBoundary

[11] Since the rotation between India and Australia beganaround 11Ma, �20 km of extension is predicted across thewestern portion of their diffuse boundary [Royer et al.,1997]. In the near-axis regions of the CIR, this divergence isaccommodated by fracture-zone reactivation and presum-ably a counter-clockwise rotation of the blocks between thefracture zones [Wiens, 1985]. Seismic observations andavailable morphologic data [Kamesh Raju, 1997; Drolia etal., 2000], however, do not show evidence for resulting CIRplate boundary reorganization [cf. Pockalny et al., 1997].This likely reflects the low long-term strain rates alongfracture zones, a consequence of the slow divergence(�3 mm/yr) and diffuse nature of the Indian-Australianboundary within the region.[12] To investigate the short-term (co-seismic) interaction

between right-lateral near-axis fracture earthquakes and theCIR (Figure 1), static stress changes resulting from the 15 Julymainshock are modeled using a finite-fault source within anelastic half-space [King et al., 1994]. A set of rectangular sub-

Figure 3. Thermal cross-section along transform T1 andits fracture-zone continuation (A-A0 on Figure 2). RTIindicates the ridge-transform intersection. The gray areaindicates the predicted rupture area, defined laterally by theextent of aftershock activity and at depth by the 600�Cisotherm. Dashed lines indicated the 600�C error limits,assuming end-member thermal diffusivities of 0.8 � 10�6

and 1.2 � 10�6 m2/s, and a mantle temperature of 1300�C.

Figure 4. (a) Map view of �CFC at 5 km depth for left-lateral slip on vertical receiver faults striking N 45�E (i.e.,active transforms). See text for model parameters. Slip onthe receiver faults is promoted in regions with positive�CFC and inhibited in regions with negative �CFC. (b)Map view of �sn at 5 km depth for vertical planes trendingparallel to the ridge axis (N 45�W). Axis parallel normalfaulting and diking will be promoted where �sn is positiveand inhibited where �sn is negative. Plate boundarymarked by dashed black line. Arrows indicate slip directionof the mainshock rupture and active transforms.

L02615 BOHNENSTIEHL ET AL.: BREAKING INTO THE PLATE L02615

3 of 4

faults with a constant slip of 3.0 m is used to approximate thesource geometry shown in Figure 3, with a rigidity of 35 GPaand a Poisson’s ratio of 0.25. Figure 4a shows the change incoulomb failure criteria (�CFC) [King et al., 1994] on a set ofvertical left-lateral receiver faults trending N 45�E (i.e., theactive transforms). �CFC = �sf + m0�sn, where �sf is thechange in shear stress ( positive in the direction of slip),�sn isthe change in normal stress (positive for unclamping), and m0

is the apparent friction. Assuming partially serpentinizedlithosphere, m0 is taken to be 0.4 [Estcartin et al., 2001].[13] Our model indicates that the right-lateral mainshock

should promote left-lateral slip (+�CFC) on the activeportions of neighboring transforms (T2, T3) (Figure 4a).Consistent with this prediction, the largest aftershock, a5.6Mw left-lateral earthquake, is located on the active portionof transform T3, �160 km northwest of the mainshockrupture. Our model also shows that left-lateral slip on theadjacent active transform T1, extending southwest from themainshock rupture, should be inhibited (��CFC) and that aright-lateral sense of slip is instead promoted. The distributionof aftershocks indicates that the rupture did not propagate intothe active portion of T1 (Figure 2); this likely reflects earlierleft-lateral loading in response to the differential motions ofthe plates across the transform. Since 1976, at least one left-lateral event of significant size has occurred on the activetransform, a 5.6 Mw event in March of 1979 (Figure 1).[14] With regard to the spreading centers, static stress

changes will tend to reduce ridge-normal compressivestresses (�sn) along the portion of the CIR axis to thenorth of the mainshock, with an increase in ridge-normalcompression to the south (Figure 4b). This will tend topromote and impede ridge parallel diking and normalfaulting within these respective regions. In contrast, due totheir geometry and sense of slip, active transform earth-quakes tend to promote ridge-normal extension on thesegments to either side of the transform.[15] Magnetic studies along the sections of the CIR that

adjoin the Indian-Australian diffuse plate boundary haveshown inter-segment variability in the rate and asymmetryof spreading, whichmay be accommodated by small jumps ofthe neovolcanic zone occurring during a short time interval[Kamesh Raju et al., 1997;Drolia et al., 2000]. Although theymay be infrequent, with recurrence times on the order of�102–103 yrs, it is possible that mega fracture-zone earth-quakes contribute to this process by influencing the positionof subsequent dike intrusion events (Figure 4b). On average, aone-meter wide dike must be emplaced every 30 years alongthe northern CIR—a frequency comparable to the decadalscales of post-seismic stress relaxation. This is unlikely to bethe sole mechanism, however, since small-offset ridge jumpsare not unique to the CIR [e.g., Kleinrock et al., 1997].

ReferencesAbercrombie, R. E., and G. Ekstrom (2001), Earthquake slip on oceanictransform faults, Nature, 410, 74–77.

Abercrombie, R. E., M. Antolik, and G. Ekstrom (2003), The June 2000Mw 7. 9 earthquakes south of Sumatra: Deformation in the Indian-Aus-tralian Plate, J. Geophys. Res., 108(B1), 2018, doi:10.1029/2001JB000674.

Antolik, M., A. Kaverina, and D. S. Dreger (2000), Compound rupture ofthe great 1998 Antarctic Plate earthquake, J. Geophys. Res., 105(B10),23,825–23,838.

Bohnenstiehl, D. R., M. Tolstoy, R. P. Dziak, C. G. Fox, and D. K. Smith(2002), Aftershock sequences in the mid-ocean ridge environment: ananalysis using hydroacoustic data, Tectonophysics, 354, 49–70.

DeMets, C., R. G. Gordon, and P. Vogt (1994), Location of the Africa-Australia-India triple junction and motion between the Australian andIndian plates: results from an aeromagnetic investigation of the CentralIndian and Carlsberg ridges, Geophys. J. Int., 118, 893–930.

Del Pezzo, E., and F. Giudicepietro (2002), Plane wave fitting method for aplane, small aperture, short period seismic array: a MATHCAD program,Comp. Geosc., 28, 59–64.

Drolia, R. K., I. Ghose, A. S. Subramanyam, M. M. Malleswara Rao,P. Kessarkar, and K. S. R. Murthy (2000), Magnetic and bathymetricinvestigations over the Vema Region of the central Indian Ridge: tectonicimplications, Mar. Geol., 167, 413–423.

Engeln, J. F., D. A. Wiens, and S. Stein (1986), Mechanisms and depth ofAtlantic transform earthquakes, J. Geophys. Res., 91, 548–577.

Estcartin, J., G. Hirth, and B. Evans (2001), Strength of slightly serpenti-nized peridotites: Implications for the tectonics of oceanic lithosphere,Geology, 29, 1023–1026.

Granville, J. P., W.-Y. Kim, and P. G. Richards (2002), An assessment ofseismic body-wave magnitudes published by the Prototype InternationalData Centre, Seismol. Res. Lett., 73, 893–906.

Kamesh Raju, K. A., T. Ramprasad, and C. Subrahmanyam (1997), Geo-physical investigations over a segment of the Central Indian Ridge,Indian Ocean, Geo-Marine Lett., 17, 195–201.

Kanamori, H., and D. L. Anderson (1975), Theoretical basis of someempirical relationships in seismology, Bull. Seismol. Soc. Am., 65,1073–1095.

Kleinrock, M. C., B. E. Tucholke, J. Lin, and M. A. Tivey (1997), Fast riftpropagation at a slow-spreading ridge: progressive tearing of an entirespreading segment, Geology, 25, 639–642.

King, G. C. P., R. S. Stein, and J. Lin (1994), Static stress changes and thetriggering of earthquakes, Bull. Seismol. Soc. Am., 84, 935–953.

Pockalny, R. A., P. J. Fox, D. J. Fornari, K. C. Macdonald, and M. R. Perfit(1997), Tectonic reconstruction of the Clipperton and Siqueiros fracturezones: Evidence and consequences of changes in plate motion over thelast 3 million years, J. Geophys. Res., 102(B2), 3167–3181.

Radha Krishna, M., R. K. Verma, and S. K. Arora (1998), Near-ridgeintraplate earthquakes in the Indian Ocean, Mar. Geol., 147, 109–122.

Royer, J.-Y., and R. G. Gordon (1997), The motion between the Capricornand Australian Plates, Science, 277, 1268–1274.

Royer, J.-Y., R. G. Gordon, C. DeMets, and P. R. Vogt (1997), New Limitson the motion between India and Australia since the chron 5 (11 Ma) andimplications for lithospheric deformation in the equatorial Indian Ocean,Geophys. J. Int., 129, 41–74.

Sykes, L. R. (1967), Mechanism of earthquakes and the nature of faultingon mid-ocean ridges, J. Geophys. Res., 72, 2131–2153.

Utsu, T., Y. Ogata, and R. S. Matsu’ura (1995), The centenary of the OmoriFormula for a decay law of aftershock activity, J. Phys. Earth, 43, 1–33.

Wiens, D. A. (1985), Historical seismicity near Chagos: a complex defor-mation zone in the equatorial Indian Ocean, Earth. Planet. Sci. Lett., 76,350–360.

Wiens, D. A., and S. Stein (1984), Intraplate seismicity and stresses inyoung oceanic lithosphere, J. Geophys. Res., 89, 11,142–11,464.

Wiens, D. A., C. DeMets, R. G. Gordon, S. Stein, D. Argus, J. F. Engeln,P. Lundgren, D. Quible, C. Stein, S. Weinstein, and D. F. Woods (1985),A diffuse plate boundary model for Indian Ocean tectonics, Geophys.Res. Lett., 12, 429–432.

�����������������������D. R. Bohnenstiehl, M. Tolstoy, and E. Chapp, Lamont-Doherty Earth

Observatory of Columbia University, 61 Route 9W, Palisades, NY 10964USA. (del@ldeo.columbia.edu)

L02615 BOHNENSTIEHL ET AL.: BREAKING INTO THE PLATE L02615

4 of 4