Author's personal copy · Author's personal copy 50 E. Contreras-Reyes, D. Carrizo / Physics of the...

This article appeared in a journal published by Elsevier. The attachedcopy is furnished to the author for internal non-commercial researchand education use, including for instruction at the authors institution

and sharing with colleagues.

Other uses, including reproduction and distribution, or selling orlicensing copies, or posting to personal, institutional or third party

websites are prohibited.

In most cases authors are permitted to post their version of thearticle (e.g. in Word or Tex form) to their personal website orinstitutional repository. Authors requiring further information

regarding Elsevier’s archiving and manuscript policies areencouraged to visit:

http://www.elsevier.com/copyright

http://www.elsevier.com/copyright

Author's personal copy

Physics of the Earth and Planetary Interiors 186 (2011) 49–58

Contents lists available at ScienceDirect

Physics of the Earth and Planetary Interiors

journa l homepage: www.e lsev ier .com/ locate /pepi

Control of high oceanic features and subduction channel on earthquake rupturesalong the Chile–Peru subduction zone

Eduardo Contreras-Reyesa,∗, Daniel Carrizoa,b

a Departamento de Geofísica, Facultad de Ciencias Físicas y Matemáticas, Universidad de Chile, Santiago, Chileb Advanced Mining Technology Center AMTC, Facultad de Ciencias Físicas y Matemáticas, Universidad de Chile, Santiago, Chile

a r t i c l e i n f o

Article history:Received 8 September 2010Received in revised form 8 March 2011Accepted 8 March 2011

Edited by: George Helffrich.

Keywords:BathymetryAsperityBarrierSubduction channelEarthquake rupture1960 and 2010 megathrust earthquakes

a b s t r a c t

We discuss the earthquake rupture behavior along the Chile–Peru subduction zone in terms of the buoy-ancy of the subducting high oceanic features (HOF’s), and the effect of the interplay between HOF andsubduction channel thickness on the degree of interplate coupling. We show a strong relation betweensubduction of HOF’s and earthquake rupture segments along the Chile–Peru margin, elucidating howthese subducting features play a key role in seismic segmentation. Within this context, the extra increaseof normal stress at the subduction interface is strongly controlled by the buoyancy of HOF’s which islikely caused by crustal thickening and mantle serpentinization beneath hotspot ridges and fracturezones, respectively. Buoyancy of HOF’s provide an increase in normal stress estimated to be as high as10–50 MPa. This significant increase of normal stress will enhance seismic coupling across the subduc-tion interface and hence will affect the seismicity. In particular, several large earthquakes (Mw ≥ 7.5) haveoccurred in regions characterized by subduction of HOF’s including fracture zones (e.g., Nazca, Challengerand Mocha), hotspot ridges (e.g., Nazca, Iquique, and Juan Fernández) and the active Nazca-Antarcticspreading center. For instance, the giant 1960 earthquake (Mw = 9.5) is coincident with the linear pro-jections of the Mocha Fracture Zone and the buoyant Chile Rise, while the active seismic gap of northChile spatially correlates with the subduction of the Iquique Ridge. Further comparison of rupture char-acteristics of large underthrusting earthquakes and the locations of subducting features provide evidencethat HOF’s control earthquake rupture acting as both asperities and barriers. This dual behavior can bepartially controlled by the subduction channel thickness. A thick subduction channel smooths the degreeof coupling caused by the subducted HOF which allows lateral earthquake rupture propagation. This mayexplain why the 1960 rupture propagates through six major fracture zones, and ceased near the MochaFracture Zone in the north and at the Chile Rise in the south (regions characterized by a thin subductionchannel). In addition, the thin subduction channel (north of the Juan Fernández Ridge) reflects a hetero-geneous frictional behavior of the subduction interface which appears to be mainly controlled by thesubduction of HOF’s.

© 2011 Elsevier B.V. All rights reserved.

1. Introduction

Along the Chile–Peru Trench the oceanic Nazca plate subductsbeneath the South American plate at a current rate of ∼6.5 cm/year(Khazaradze and Klotz, 2003). The Nazca plate hosts a large pop-ulation of high oceanic features (HOF’s) including hotspot tracksand oceanic fracture zones (FZ’s). Perhaps the most striking fea-ture of the oceanic plate is the Chile Rise, an active spreadingcenter that marks the boundary between the oceanic Nazca andAntarctic plates and is currently subducting at ∼46◦S beneaththe South American plate. Bathymetric data show four mainoceanic hotspot tracks: Nazca, Iquique, Copiapó and Juan Fer-

∗ Corresponding author. Tel.: +56 2 9784296; fax: +56 2 6968686.E-mail address: [email protected] (E. Contreras-Reyes).

nández Ridges (Fig. 1). The Nazca Ridge (NR) was formed atthe Easter Island hotspot on the Pacific-Farallon/Nazca spread-ing center (Pilger, 1984), while the Juan Fernández Ridge (JFR) isan off-ridge hotspot formed at the JFR hotspot onto 27 Myr oldoceanic crust (Yánez et al., 2001). The oceanic Nazca plate is furthersegmented by several oceanic fracture zones formed at the Pacific-Nazca and Antarctic-Nazca spreading centers (Tebbens et al.,1997).

Most of the seafloor features hosted by the oceanic Nazca plateare in the throes of being subducted (Herron et al., 1981; Yánezet al., 2001; Rosenbaum et al., 2005). Subduction of a HOF maymodify the geodynamics and tectonic setting of the outer forearcregion, and disrupt and erode material from the overriding plate.In addition, subduction of a HOF is expected to influence dramat-ically the degree of coupling across the subduction interface andmay affect seismicity, in particular the size and frequency of large

0031-9201/$ – see front matter © 2011 Elsevier B.V. All rights reserved.doi:10.1016/j.pepi.2011.03.002


50 E. Contreras-Reyes, D. Carrizo / Physics of the Earth and Planetary Interiors 186 (2011) 49–58

Fig. 1. (A) Main oceanic bathymetric features and most significant earthquakes (red elipses) from the 19th to 21st centuries are shown. Bathymetry is taken from Sandwelland Smith (1997) and main oceanic fracture zones are taken from Tebbens et al. (1997). (B) Earthquake ruptures (Mw > 7.0) were compiled from Beck et al. (1998), Comte andPardo (1991), Cisternas et al. (2005) and Bilek (2010). (For interpretation of the references to color in this sentence, the reader is referred to the web version of the article.)

earthquakes (Kelleher and McCann, 1976; Cloos, 1992; Scholz andSmall, 1997).

The slip distribution at the subduction interface is highly het-erogeneous in terms of the strength or frictional characteristics ofthe material in the fault zone. These stress heterogeneities havebeen recognized as asperities (Kanamori, 1994) and barriers (Dasand Aki, 1977) to earthquake rupture which control the seismicmoment release and rupture area. A seismic asperity is an area withlocally increased friction and exhibits little aseismic slip during theinterseismic period relative to the surrounding regions. Once theshear yield stress �o (critical shear stress required for failure) alongthese heterogeneities is reached by the accumulated interseismicshear stress �1, the asperity concentrates the coseismic momentrelease and slip during the earthquake. The regions with little or noslip during rupture propagation are generally called barriers andcontrol the size of the earthquake rupture area (ERA) (Aki, 1979;Kanamori and McNally, 1982).

Once the earthquake initiates (�1 ≥ �o), the rupture front prop-agates through regions where the dynamic stress associated withthe rupture process is larger than �o. If the rupture front encoun-ters an obstacle where the dynamic stress is lower than �o, the faultmotion slows down and eventually stops (barrier zone). In otherwords, the interplay between the amount of dynamic stress and �o

will define whether a region behaves as an asperity with large slipor behaves as a barrier with no slip. Therefore, it is fundamentalinvestigate the shear yield stress distribution along the subductioninterface. A heterogeneous distribution of �o might be useful to pre-dict the location of local barriers and asperities before occurrence ofan earthquake. �o varies over time and depends on frictional prop-erties, fluid pore-pressure, and state stress anomaly. For example,subduction of a rigid seamount will result in an increase of normalstress �n and hence �o preventing eventual rupture (strong cou-pling). Likewise, buoyancy of HOF’s will enhance seismic couplingand anomalous increase of �o. In contrast, an increase of fluid porepressure will reduce �o facilitating rupture propagation.

Since HOF’s dramatically modify the geodynamics of the outerforearc and interplate boundary conditions, they are probably themost obvious candidates for asperities and/or barriers. The fric-tional behavior of HOF’s is clearly influenced by the increase ofnormal stress at the subduction interface caused by the excessbuoyancy of these features (Scholz and Small, 1997). Furthermore,Scholz and Small (1997) claim that the “geometric” normal stresseffect due to the need of the overriding plate to accommodate theshape of the HOF is a significant additional source of increasing nor-mal stress at the subduction interface. Thus, the anomalous strongcoupling associated with HOF’s (buoyancy and geometry) might


E. Contreras-Reyes, D. Carrizo / Physics of the Earth and Planetary Interiors 186 (2011) 49–58 51

play an important role on seismicity. This appears to be the casefor South America, where many large earthquakes occur in regionswhere these large Nazca plate features enter the trench.

We study the spatial relationship between large underthrust-ing earthquakes (Mw ≥ 7.5) and bathymetric heterogeneities on thesubducting Nazca oceanic plate along the Chile–Peru subductionzone. Such heterogeneities are manifested by subduction of bathy-metric anomalies (HOF’s), which are likely to affect the seismiccoupling across the subduction interface due to their buoyancy.For instance, hotspot seamounts usually host thick oceanic crustand magmatic material that may have erupted, ponded beneathor intruded the base of the oceanic crust (magmatic underplating)(Koppers and Watts, 2010). On the other hand, oceanic fracturezones are characterized by a high degree of hydration in both thecrust and mantle, and they commonly comprise bodies of serpen-tinites deep into the oceanic upper mantle (Contreras-Reyes et al.,2008). Thus, buoyancy is linked directly to structural or chemicalalteration of the crust or mantle beneath the oceanic feature thatstores rocks less dense than its surroundings. We use publishedseismic velocity models across the Nazca Ridge (NR), JFR and MochaFZ in order to estimate the anomalous normal stress associated withthe buoyancy of these HOF’s. In addition, several HOF’s are identi-fied from bathymetric data and compared with published coseismicslip models of large underthrusting earthquakes in order to exploreif subducting HOF correlate with zones of large slip (asperities)and/or little slip (barriers).

In addition, the amount of subducted sediment may also playan important role on megathrust seismicity since the subduc-tion channel might smooth the subduction interface resulting ina homogenous coupled region. It can be expected that a largehomogenous coupled region facilitates long lateral rupture propa-gation (Ruff, 1989; Scherwath et al., 2009; Contreras-Reyes et al.,2010) and hence the earthquake size. Therefore, a thin subductionchannel above a subducting HOF may reflect the presence of a seis-mic barrier while a thick subduction channel would tend to reducethe effect of the anomalous coupled region caused by the subduct-ing HOF. In order to test this hypothesis for some earthquakesrecorded along the Chilean subduction we use a compilation ofgeodetic data and seismic data constraining the subduction channelthickness and we discuss the interplay between subduction chan-nel thickness and the dual asperity-barrier behavior. Identificationof both subduction channel thickness variation and magnitude ofbuoyancy forces associated to HOF’s would provide a very usefulcontext for future seismological works.

2. Relation between high oceanic features and earthquakerupture areas along the Chile–Peru subduction zone

The Chile–Peru subduction zone is an extremely active con-vergent plate margin producing large earthquakes (Mw > 8) aboutevery 10 years and capable of generating at least one tsunamigenicmega-thrust earthquake per century. This is documented in his-torical accounts and instrumentally well-recorded events that dateback to the 14th century (Darwin, 1851; Beck et al., 1998; Bilek,2010). Fig. 1A shows the location of the rupture areas of the largestevents documented over the 19th to 21st centuries. These ERA’sare spatially linked to the locations of the main oceanic incomingfeatures of the Nazca plate. One of the most important observationsis that at least one rupture limit for each event coincides approx-imately with a subducting HOF, and about half of the ERA’s arebounded by HOF’s. In the following we discuss some examples:

(a) The southern Peru–northern Chile region (15.5–23◦S): thisregion is characterized by two active seismic gaps that mark thezones of the (Mw = 8.7) 1868 southern Peru and the (Mw = 8.9)

1877 northern Chile mega-thrust earthquakes (Comte andPardo, 1991). The southern Peru 1868 rupture segment brokepreviously in 1604 and 1784 in a similar, approximately 450 kmlong rupture. The northwest part of that segment broke againin an Mw = 8.4 earthquake in 2001, filling at least partially thesouth Peru gap (Ruegg et al., 2001). This event propagated for∼70 km before encountering a 6000 km2 fault area that acted asa barrier. Approximately 30 s after the start of rupture, this bar-rier began to break when the main rupture front was ∼200 kmfrom the epicenter (Robinson et al., 2006). In this case, the sub-ducted feature acted as both a barrier and an asperity duringa single event and was associated with the oceanic Nazca FZ(Robinson et al., 2006). Further to the south, the northern Chileseismic gap (18.5–23◦S) is roughly coincident with the subduc-tion of the Iquique Ridge (IR). Furthermore, the 2007 Tocopillaearthquake (Mw = 7.7) spatially correlates with the southernpart of the IR, while the 1995 Antofagasta event (Mw = 8.0) isbounded by the IR and Tal–Tal ridge in the north and south,respectively (Figs. 1A and 2).

(b) Central Chile (31–37.5◦S): the most recent large Chileanearthquake occurred on February 27th, 2010 (Mw = 8.8) andpropagated northward and southward achieving a final rupturelength of about 450 km (34–38◦S; Lay et al., 2010). In the sameregion, 176 years ago in 1835 a large earthquake with an esti-mated magnitude of Mw ∼ 8.5 was reported by Darwin (1851).North of the rupture area of the 2010 earthquake, two megath-rust earthquakes occurred: the 1906 (Mw ∼ 8.4; Beck et al.,1998) and 1985 (Ms = 7.8; Barrientos, 1995) events. The south-ern limit of the 2010 earthquake is coincident with the MochaFZ, while the northern limits of the 1906 and 1985 earthquakesare coincident with the subduction of JFR (Fig. 1). Similarly, thesouthern limit of the 1943 (Ms = 7.9) earthquake correlates withthe JFR.

(c) The 1960 mega-thrust event (37.5–46◦S): the largest instru-mentally recorded earthquake, which occurred in 1960 witha magnitude of Mw = 9.5, ruptured over a length of almost1000 km from ∼37.5◦S to 46◦S near the Chile Triple Junction(CTJ) of the Nazca, Antarctic and South American plates (Morenoet al., 2009). The Chile Rise is currently subducting at the CTJ,which is a region characterized by a shallow trench associatedwith the buoyancy of this active spreading ridge (Herron et al.,1981). The 1960 earthquake ruptured across six major fractureszones and it stopped near the Mocha FZ to the north and atthe CTJ to the south. In this case, the fractures zones involveddid not act as barriers, except for the Mocha FZ and Chile Risewhich are the northern and southern earthquake rupture limits,respectively.

Other examples showing spatial correlations between ERA loca-tions and incoming HOF’s can be inferred from Fig. 1A. In addition,such spatial correlations also appear to apply for the historical seis-mic segmentation of the last 500 years (see Fig. 1B). Consideringthe historical ERA estimates, not every limit of an ERA coincideswith the subduction of an HOF, but at least one limit of each ERAis coincident with a prominent HOF as is shown in Fig. 1A, andhalf of the events are bounded at both limits by HOF’s. Thus, thelarge earthquakes of the Chile–Peru margin appear to be highlysegmented by bathymetric features. Obviously, the uncertainty ofearthquake rupture areas are large for historical reports and thecorrelation between HOF and ERA position is far from perfect forold events (Fig. 1). Nevertheless, events instrumentally recordedduring the 20th to 21st centuries present a reasonably good correla-tion between HOF and ERA position (Fig. 1). Additionally, geodeticaldata provide further constraints for the size of ERA’s and fault slipdistribution as we discuss in the following section.



Fig. 2. (A) Southern Peru-northern Chile seismic gap. Ellipses represent the approximate extent of rupture zones of the 1868 (Mw = 8.7) and 1877 (Mw = 8.9) events afterComte and Pardo (1991). The rupture area of the 2001 Arequipa earthquake (Mw = 8.4) is almost coincident with the middle part of the 1868 rupture area. In northern Chile,no great event had been recorded since 1877. The Mw = 8.1 Antofagasta earthquake of 1995 July 30 ruptured an 180 km long segment located south of the 1877 rupture zone.The 2007 Tocopilla earthquake Mw = 7.7 is almost coincident with the southern part of the 1877 rupture zone. For those earthquakes with known distributed slip we use theoutermost contour to represent the rupture area (for the 1995 earthquake from Chlieh et al. (2004), for the 2001 Arequipa earthquake from Pritchard et al. (2007) and for the2007 earthquake Béjar-Pizarro et al. (2010). Starts denote epicenter locations. (B) Fault slip model after Chlieh et al. (2004) and Béjar-Pizarro et al. (2010) for the 1995 and2007 events, respectively. Slip distribution with 3 m contour interval of the 2001 Arequipa earthquake based on the coseismic fault slip model of Pritchard et al. (2007). Grayand salmon regions represent the swell associated to the Iquique and Tal–Tal Ridges, respectively. Red dots denote the largest seamounts identified in A. (For interpretationof the references to color in this sentence, the reader is referred to the web version of the article.)

3. Geodetical observations

Geodetical data provide further and crucial evidence for seis-mic coupling at the subduction interface, allowing identificationof regions with high coseismic slip (asperities) or no slip (barri-ers). In this section, we show a compilation of fault slip models for(1) the active seismic gap of the southern Peru-north Chile region(15.5–23◦S), (2) Central Chile (31–38◦S), and (3) the region of thelargest ever recorded 1960 earthquake (∼38–46◦S).

3.1. The active seismic gap of south Peru-north Chile

The region between the Ilo Peninsula (15.5◦S, South Peru) andthe Mejillones Peninsula (23.5◦S, North Chile) represents a majorseismic gap that has not experienced a significant megathrustearthquake since the 1868 earthquake (Mw = 8.7) and the 1877earthquake (Mw = 8.9) (Comte and Pardo, 1991, Fig. 2A). The 1995Antofagasta (Mw = 8.1) and the 2001 Arequipa (Mw = 8.4) earth-quakes appear to have provided an extra load at both extremitiesof the remaining ∼500 km length unruptured segment. After the1877 event and before the 2007 Tocopilla earthquake, a few events(Mw < 7.5) have been reported in the region (Comte and Pardo,1991; Tichelaar and Ruff, 1991) but they were not large enoughto release a significant part of the 9 m of slip deficit accumulated inthe gap in the last 130 years (Béjar-Pizarro et al., 2010).

Fig. 2B shows the slip distribution with 3 m contour interval ofthe 2001 Arequipa earthquake (Mw = 8.4) based on the coseismicslip model of Pritchard et al. (2007). This event only partially filledthe rupture zone of the 1868 earthquake (Fig. 2A). The rupture frontpropagates southward and encounters the Nazca FZ at 70 km fromthe hypocenter. Although, this HOF apparently acted as barrier, the

rupture continued around this barrier, which remained unbrokenfor ∼30 s and then began to break when the main rupture front was∼200 km from the epicenter (Robinson et al., 2006).

The 1995 Antofagasta earthquake (Mw = 8.1) ruptured a 180 km-long segment just south of the main 1877 gap. Geodetic slipinversions give a consistent picture of the coseismic slip distribu-tion during the earthquake with a single asperity slipping 5–10 m(Chlieh et al., 2004; Fig. 2B). The earthquake rupture area is lim-ited by the southern part of the Iquique Ridge in the north and theTal–Tal ridge in the south (Fig. 2B). Thus, these features appear toact as barrier for rupture propagation.

The 2007 Tocopilla event (Mw = 7.7) ruptured the deeper part ofthe seismogenic interface (30–50 km) and did not reach the surface,indicating that the shallow part of the seismogenic interface (fromthe trench to 30 km depth) remains unbroken with the exception ofthe southern edge of the rupture (at ∼23◦S), where coseismic slipseems to have propagated up to ∼25 km depth (Béjar-Pizarro et al.,2010). The fault slip model shows slip concentrated in two mainasperities with maximum values lower than 2 m (Béjar-Pizarroet al., 2010). The rupture area of this event is roughly coincidentwith southern region of the IR which may have arrested the updiprupture propagation (Fig. 2B).

3.2. Central Chile (31–37.5◦S)

The rupture area of the most recent megathrust earthquake; the2010 event (Mw = 8.8), is characterized by two regions of high seis-mic slip with maximum slip of 10–15 m (Lay et al., 2010; Morenoet al., 2010). Between these asperities, the rupture bridged a zonethat was creeping interseismically with consistently low coseismicslip. The bilateral rupture propagated towards the south, where one



Fig. 3. (A) Earthquake segmentation along the south central Chile margin is indicated by ellipses that enclose the approximate rupture areas of historic earthquakes (Becket al., 1998; Ruegg et al., 2009). Starts denote epicenter location. (B) Fault slip distribution of the 2010 Maule earthquake obtained by inversion of teleseismic P waves, SHwaves, and resulting R1 source–time functions after Lay et al. (2010). Slip distribution at 0.5 m contour interval of the 1985 earthquake based on the dislocation model ofMendoza et al. (1994).

asperity failed, and bridged a previously creeping zone to the north,where a second asperity failed (Moreno et al., 2010).

The southern limit of the rupture area is fairly coincident withthe subduction of the Mocha FZ (Fig. 3A), and hence this HOF mighthave acted as seismic barrier. In the northen limit of the 2010earthquake, the rupture apparently stops some 100 km south ofthe subduction JFR according to the teleseismic fault slip model ofLay et al. (2010) (Fig. 3B). Moreno et al. (2010) proposed that pre-stress was lowered in the northern periphery of the 2010 ERA byprevious earthquakes such as 1906 and 1985 events resulting in alow-stress barrier. Nevertheless, the JFR appears to be the northernbarrier for both the 1906 and 1985 events, and further north thesouthern barrier of the 1943 earthquake (Ms = 7.9) (Fig. 3). Thus,the JFR plays a crucial role on seismic segmentation in south cen-tral Chile. In fact, at ∼72.5◦W/32.5◦S, magnetic anomalies revealthe subduction of a large seamount belonging to the Juan Fernán-dez hotspot track (Yánez et al., 2001). The size of this seamount isabout 40 km in diameter and 4 km high, and therefore equivalent tothe size of the O’Higgins guyot which host a wide crustal root and amagmatic underplating body beneath the Moho (Kopp et al., 2004,see Section 4). Thus, the subduction of the JFR might considerablyaffect seismicity in this segment of the Chilean margin.

3.3. The 1960 megathrust earthquake (Mw = 9.5)

Fig. 4 shows the fault slip model inverted by Moreno et al. (2009)using the geodetic data compiled by Plafker and Savage (1970). Thenorthern limit of the 1960 ERA is coincident with both the Peninsulade Arauco and the subducting Mocha FZ, while in the south faultslip ceased a few kilometres north of the CTJ. The most constrainedregion of the model is north of Isla de Chiloé, where most geodeticaldata were acquired (Moreno, personal communication). Here, thetwo main slip patches are bounded by fractures zones, and theseHOF’s present a reduction of fault slip or increase of coupling.

4. Seafloor morphology and buoyant zones of the oceanicNazca plate

Individual features on the subducting Nazca oceanic plate arelarge, reaching elevations of up to 5 km above the surrounding

regional seafloor depth and widths up to 70 km (Fig. 1). The buoy-ancy of these features is a key factor controlling interplate seismiccoupling and depends on the local mode of isostatic compensa-tion. Fig. 5 shows the seismic structure of the NR (Hampel et al.,2004), easternmost portion of the JFR (Kopp et al., 2004), andthe Mocha FZ (Contreras-Reyes et al., 2008) derived from wide-angle seismic transects. The NR comprises of an anomalously thickcrust (∼14 km), twice the thickness of the typical oceanic Nazcacrust, and characterized by underplating (Hampel et al., 2004)(Fig. 5A). Both the thick crust and the underplated subcrustal mate-rial represent anomalous bodies with density similar to gabbro, butlower than the surrounding mantle peridotite. The easternmostportion of the JFR corresponds to the O’Higgins seamount group(3–4 km high) which shows a moderately thick crust according tothe seismic model of Kopp et al. (2004) (Fig. 5B). However, themajor buoyancy source is owed to reduced mantle velocities andhence densities. That is, low upper mantle velocities in the rangeof 7.5 ≤ Vp ≤ 8.0 km/s are commonly attributed to underplating ofmelt bodies beneath the crust which are usually centered at the vol-canic edifice, and an asymmetric distribution is rather uncommon.Fig. 4B shows that the reduced mantle velocities extend furthertrenchward suggesting mantle serpentinization in the outer riseregion (Kopp et al., 2004). Thus, the serpentinized mantle also rep-resents an extra buoyant force since the bulk density of peridotitedecreases with the degree of hydration. Regardless if low mantleseismic velocities and densities are caused by mantle serpentiniza-tion or magmatic underplating, both processes are an importantsource for buoyancy of the oceanic lithosphere.

Contreras-Reyes et al. (2008), reported both low crustal andupper mantle velocities beneath the Mocha FZ which were inter-preted in terms of an intensely fractured and hydrated upperoceanic lithosphere. The maximum depth of mantle serpentiniza-tion is up to 6–8 km within the uppermost mantle, providing animportant buoyancy source (Fig. 5C). Thus, buoyancy due to crustaland upper mantle hydration may contribute an important mecha-nism for increased coupling at the subduction interface.

Table 1 and Fig. 5 summarize the resulting anomalous normalstress, ��n, associated with the buoyancy of the NR, JFR and MochaFZ. ��n ranges between 10 and 50 MPa providing a significant



Fig. 4. (A) Rupture area of the mega-thrust 1960 earthquake. Start denotes epicenter location. (B) Fault slip model after Moreno et al. (2009) and main oceanic bathymetricfeatures identified. Note the strong spatial correlation between HOF’s and low slip values, which suggests that HOF’s enhance strong coupling and hence reduced fault slip.

increase of coupling at the plate interface compared to the appar-ent stress drop of about 30 MPa reported as an average for Chileaninterplate earthquakes (Leyton et al., 2009). Our ��n estimates arebased only on the buoyancy of HOF’s and represent a minimum ofthe actual anomalous coupling. Actually, ��n is likely to be higherdue to the additional normal stress associated with the accommo-dation of a subducting high relief topographic feature (topographiceffect). Scholz and Small (1997) roughly estimated that a typicalseamount 4 km high and 60 km wide results in an extra normalstress of 100 MPa. Thus, the resulting ��n is expected to be higherthan estimates shown in Table 1 and Fig. 5 for large seamounts suchas the NR and JFR.

5. Discussion

5.1. Role of subducting HOF’s on seismic coupling

The spatial correlation between ERA locations and incomingHOF’s indicates that seamounts and fracture zones play a key rolein the dynamics of the earthquake rupture along the Chile–Peru

subduction zone. For instance, part of the southern Peru seismicgap and the entire northern Chile seismic gap is coincident withthe subduction of the IR, which likely controls at least in part thedegree of coupling in the seismogenic zone. The IR is composed bya number of seamounts (1.0–2.0 km high) and it is surrounded bya smooth and broad region of shallow seafloor (swell). This swellis 200–300 km wide and more than 500 m in elevation (Fig. 2). Thegray region representing the IR’s swell shown in Fig. 2B is basedon the shallow topography of this HOF inferred from the bathy-metric data. The extension and distribution of the swell is not cleardue to presence of the prominent outer rise topography. McNuttand Bonneville (2000) proposed that the swell is mainly a con-sequence of the buoyancy of magmatic underplating beneath theMoho or crustal thickening, as imaged by seismic refraction exper-iment. Unfortunately, the deep structure of the IR has not yet beenimaged by seismic refraction data and a high resolution image of theIR deep structure is unknown. Nevertheless, the 3D density modelof Tassara et al. (2006) shows that the IR hosts an anomalously thickcrust ∼15 km thick providing an important source of buoyancy. Asmooth and uplifted topography (like the IR swell) might provide a

Table 1The anomalous normal stress ��n is calculated as ��n = ��gH + �g�H, where the ��n (buoyancy force) depends on the thickness of the corresponding anomalouscrustal thickness (�Hc), thickness of the underplated magmatic material beneath the crust and/or thickness of the serpentinized mantle (�Hm). ��n also depends on themantle–crust density contrast (��mc = 530 kg/m3), “normal” mantle–serpentinized mantle density contrast (��m = 230 kg/m3) and/or “normal” crust-altered crust densitycontrast (��c = 50 kg/m3). g = 9.81 m/s2 is the acceleration due to gravity. �Hm and �Hc are based on the information shown in Fig. 5.

Oceanic feature �Hc (km) ��mc (kg/m3) �Hm (km) ��m (kg/m3) ��c (kg/m3) ��n (MPa)

NR 7.4 530 3.0 230 0 50JFR 2.0 530 3.0 230 0 10Mocha FZ 0.0 – 8.0 230 50 20



Fig.

5.(u

pp

er)

Stru

ctu

rean

din

terp

reta

tion

ofth

e(A

)N

azca

Rid

ge,(

B)

east

ern

mos

tp

orti

onof

the

Juan

Fern

ánd

ezR

idge

,an

d(C

)M

och

aFZ

base

don

the

2Dse

ism

icve

loci

tym

odel

ofH

amp

elet

al.(

2004

),K

opp

etal

.(20

04),

and

Con

trer

as-R

eyes

etal

.(20

08),

resp

ecti

vely

.Map

loca

tion

sof

seis

mic

pro

file

sar

esh

own

inFi

g.1A

.(be

low

)D

irec

tco

mp

aris

onof

thes

eH

OF’

sst

ruct

ure

wit

hty

pic

alN

azca

ocea

nic

cru

st(6

.5km

thic

k).T

he

anom

alou

sn

orm

alst

ress

��

n(b

uoy

ancy

forc

e)d

epen

ds

onth

eth

ickn

ess

ofth

eco

rres

pon

din

gan

omal

ous

cru

stal

thic

knes

s(�

Hc)

and

onth

ickn

ess

ofth

eu

nd

erp

late

dm

agm

atic

mat

eria

lben

eath

the

cru

stan

d/o

rth

ickn

ess

ofth

ese

rpen

tin

ized

man

tle

(�H

m).

��

nal

sod

epen

ds

onth

em

antl

e–cr

ust

den

sity

con

tras

t(�

�m

c=

530

kg/m

3),

“nor

mal

”m

antl

e–se

rpen

tin

ized

man

tle

den

sity

con

tras

t(�

�m

=23

0kg

/m3)a

nd

“nor

mal

”cr

ust

-alt

ered

cru

std

ensi

tyco

ntr

ast(

��

c=

50kg

/m3).

See

furt

her

det

ails

inTa

ble

1.D

ensi

tyva

lues

are

take

nfr

omth

ed

ensi

tym

odel

ofTa

ssar

aet

al.(

2006

).



Fig. 6. Cartoons showing the subduction interface with a thin (a) and thick (b) subduction channel cover. A thin subduction channel is typical for the northern Chile andPeruvian margins (see Contreras-Reyes et al. (2010) for a review), whereas a thick subduction channel thicker than 1.0 km correlates with the rupture area of the 1960earthquake (Scherwath et al., 2009; Contreras-Reyes et al., 2010). A thick subduction channel is expected to smooth the plate interface contact and favor earthquake rupturepropagation. FAP denotes frontal accretionary prism.

high degree of coupling homogeneously distributed. Furthermore,the IR subduction is highly oblique beneath South America, effec-tively spreading the anomalously coupled zone over a wider region.Similarly, the obliqueness of other subducted HOF’s can extend theanomalously coupled region along strike (e.g., Nazca FZ, ChallengerFZ, Iquique FZ and Mocha FZ). For large HOF’s such as the IR, obliquesubduction can explain the large size and recurrence interval ofgreat earthquakes.

The impact of subducting seamounts on the degree of couplinghas been broadly documented (e.g., Scholz and Small, 1997; Cloosand Shreve, 1996) in contrast to the effect of fracture zones on seis-mic coupling which has been given less attention. Even though theimportance of oceanic fracture zones on seismic segmentation hasbeen proposed by some authors (e.g., Ruff and Miller, 1994), themechanism responsable for such segmentation has not been diss-cused much. We proposed here that the deep serpentinized bodieswithin the uppermost mantle hosted by oceanic FZs affect dra-matically the buoyancy and hence the degree of coupling (Fig. 5).Consequently, their impact on seismicity is of similar importanceto hotspot seamounts. For instance, the subduction of the Nazca FZ,Iquique FZ, Challenger FZ and Mocha FZ appears to strongly corre-late with the northern and southern limits of several ERA’s (Fig. 1A).In most cases these FZ’s coincide with the ERA’s borders and henceact as seismic barriers.

5.2. The interplay between HOF and subduction channelthickness and its impact on rupture propagation

An additional and fundamental element affecting the couplingdegree along the subduction interface is the subduction channelthickness (e.g., Ruff, 1989). For example, a rough subducting HOFcovered by a thin subduction channel translates into highly hetero-geneous strength distribution, and between regions of high reliefwould behave aseismic due to its weak coupling, which will resultin small earthquake generation (Ruff, 1989). This process is illus-trated in Fig. 6A, which shows the typical structure of the northern

erosional Chile margin characterized by a thin subduction chan-nel. A thin subduction channel would result in a potential asperityand large earthquake nucleation if the dynamic stress is larger thanthe yield shear stress, otherwise it will behave as a seismic barrier.On the other hand, a thick subduction channel (Fig. 6B) will allowa smoother strength distribution along the plate interface result-ing into highly homogeneous strength distribution (Ruff, 1989).It can be expected that a thick subduction channel would reducecoupling above topographic heterogeneities and it would increaseplate coupling between high topographies (Fig. 6B). Thus, these twoprocesses together will favor a homogeneous plate contact and con-sequently earthquake rupture propagation and generation of giantearthquakes.

Recently, seismic studies have reported large variations ofthe subduction channel thickness along the Chile–Peru margin(Scherwath et al., 2009; Contreras-Reyes et al., 2010). North of theJFR, the erosional northern Chile and Peruvian margin are charac-terized by a thin subduction channel, typically thinner than 300 m(Reichert et al., 2002). On the other hand, the subduction channelsouth of the JFR and north of the Mocha FZ is relatively thin (≤1.0 kmthick) compared to the subduction channel south of the Mocha FZand north of the CTJ (>1.5 km) (Contreras-Reyes et al., 2010). Thus,we would expect that south of the Mocha FZ, the thick subductionchannel has a strong influence on earthquake rupture propagation.In contrast, north of the Mocha FZ where the subduction channelbecomes thinner we would expect that the HOF’s mainly controlthe size of the earthquake. Although the limits of several earth-quakes do not perfectly correlate with HOF subduction, this may beexplained by local anomalous regions of thick subduction channelsthat would tend to homogenize the subduction interface facilitatingrupture propagation.

The Mocha FZ is a remarkable oceanic feature controlling thesize of both the 1960 and 2010 earthquakes. In fact, the highly cou-pled region where the Mocha FZ subducts appears to have arrestedthe southward propagation of the recent Maule 2010 earthquake(Fig. 3). On the other hand, the Mocha FZ stopped in the north the



rupture propagation of the megathrust 1960 event (Fig. 4). Thisearthquake then propagated all the way south to just north of theCTJ. The thick subduction channel along the rupture zone may haveprovided enough smoothness for a long rupture propagation acrossthe fracture zones involved. The thin subduction channels north ofthe Mocha FZ and near the CTJ were too thin to reduce plate cou-pling, and rupturing ceased before crossing the Mocha FZ and theChile Rise in the north and south, respectively (Scherwath et al.,2009). In addition, the fault-slip model shows that the slip patches(Moreno et al., 2009) are approximately bounded by the subduc-tion of fracture zones, which suggests that these HOF’s tend to actas barriers, but both the large amount of energy carried by the rup-ture front and the thick subduction channel have allowed furtherrupture propagation (Fig. 4).

A giant earthquake can also be generated in a region with a thinsubduction channel, if the topography of the subducting feature israther smooth and wide enough as for instance an oceanic plateau.This case could be the active seismic gap of north Chile, where theIR subducts and no major earthquake (Mw > 8.5) has occurred sincemore than 120 years. The smooth topography and deep crustal rootof the IR (Tassara et al., 2006) favors efficiently the plate contact,seismic coupling and elastic energy accumulation during the inter-seismic period. When the accumulated shear stress becomes largerthan critical stress required for failure, a large earthquake may gen-erate with a rupture area of similar dimensions to the IR.

6. Summary

We propose that two important factors affect the earthquakerupture behavior: (1) the buoyancy of the subducting HOF, and (2)the effect of the interplay between HOF and subduction channelthickness on the degree of coupling. We show a strong relationbetween subduction of oceanic features and earthquake rupturesegments along the Chile–Peru margin, elucidating how these sub-ducting features play a key role in seismic segmentation. Withinthis context, the extra increase of normal stress at the subductioninterface is strongly controlled by the buoyancy of HOF’s. Thesefeatures generate anomalously coupled regions acting as barriersand/or asperities. This dual behavior depends mainly on the inter-play between the magnitude of the yield shear stress in the vicinityof the HOF and the energy carried by the rupture front. A thick sub-duction channel smooths the degree of coupling and hence reducethe yield shear stress at the subduction interface allowing lateralearthquake rupture propagation. Future work should focus on moreseismic imaging HOF’s and subduction channels in order to betterunderstand their internal structures to provide better constraintson rupture earthquake dynamics modelling and seismic hazardassessment.

Acknowledgments

We thank Javier Ruiz and Sergio Ruiz for fruitful discussions.Eduardo Contreras-Reyes acknowledges the support of the ChileanNational Science Foundation (FONDECYT) project # 11090009. Wethank Marta Béjar-Pizarro and Marcos Moreno for providing thedigital coseismic slip models. We also thank M.A. Gutscher, ananonymous reviewer, and the Editor George Helffrich for com-ments on the manuscript. We appreciate comments on the Englishprovided by Dave Rahn.

References

Aki, K., 1979. Characterization of barriers on an earthquake fault. J. Geophys. Res. 84,6140–6148.

Beck, S., Barrientos, S., Kausel, E., Reyes, M., 1998. Source characteristics of historicearthquakes along the central Chile subduction zone. J. South Am. Earth Sci. 11,115–129.

Barrientos, S., 1995. Dual seismogenic behavior: the 1985 Central Chile earthquake.Geophys. Res. Lett. 22 (24), 3541–3544.

Béjar-Pizarro, M., et al., 2010. Asperities and barriers on the seismogenic zonein North Chile: state-of-the-art after the 2007 Mw 7.7 Tocopilla earthquakeinferred by GPS and InSAR data. Geophys. J. Int. 183, 390–406.

Bilek, S.L., 2010. Seismicity along the South American subduction zone: review oflarge earthquakes, tsunamis, and subduction zone complexity. Tectonophysics495, 1–2, doi:10.1016/j.tecto.2009.02.037.

Chlieh, M., de Chabalier, J.B., Ruegg, J.C., Armijo, R., Dmowska, R., Campos, J., Feigl,K.L., 2004. Crustal deformation and fault slip during the seismic cycle in theNorth Chile subduction zone, from GPS and InSAR observations. Geophys. J. Int.158 (2), 695–711.

Cloos, M., 1992. Thrust-type subduction-zone earthquakes and seamount asperities:a physical model for seismic rupture. Geology 20, 601–604.

Cloos, M., Shreve, R., 1996. Shear-zone thickness and the seismicity of Chilean andMarianas-type subduction zones. Geology 24, 107–110.

Cisternas, M., et al., 2005. Predecessors of the giant 1960 Chile earthquake. Nature437 (15), 404–407, doi:10.1038/nature03943.

Comte, D., Pardo, M., 1991. Reapprisal of great historical eathquakes in NorthernChile and Southern Peru seismic gaps. Nat. Hazards 4, 23–44.

Contreras-Reyes, E., Grevemeyer, I., Flueh, E.R., Reichert, C., 2008. Upper lithosphericstructure of the subduction zone offshore southern Arauco Peninsula, Chile at∼38◦S. J. Geophys. Res. 113, B07303, doi:10.1029/2007JB005569.

Contreras-Reyes, E.R., Flueh, E., Grevemeyer, I., 2010. Tectonic control on sedimentaccretion and subduction off south-central Chile: Implications for coseismicrupture processes of the 1960 and 2010 megathrust earthquakes. Tectonics 29,TC6018, doi:10.1029/2010TC002734.

Darwin, C., 1851. Geological Observations on Coral Reefs, Volcanic Islands and onSouth America. Londres, Smith, Elder and Co, p. 768.

Das, S., Aki, K., 1977. Fault plane with barriers: a versatile earthquake model. J.Geophys. Res. 82, 5658–5670.

Hampel, A., Kukowski, N., Bialas, J., Huebscher, C., Heinbockel, R., 2004. Ridge sub-duction at an erosive margin: the collision zone of the Nazca Ridge in southernPeru. J. Geophys. Res. 109, B02101, doi:10.1029/2003JB002593.

Herron, E.M., Cande, S.C., Hall, B.R., 1981. An active spreading center collides with asubduction zone, a geophysical survey of the Chile margin triple junction. Mem.Geol. Soc. Am. 154, 683–701.

Kanamori, H., 1994. Mechanics of earthquakes. Annu. Rev. Earth Planet. Sci. 22,207–237.

Kanamori, H., McNally, K.C., 1982. Variable rupture mode of the subductionzone along the Ecuador–Colombia coast. Bull. Seism. Soc. Am. 72–74, 1241–1253.

Khazaradze, G., Klotz, J., 2003. Short- and long-term effects of GPS measured crustaldeformation rates along the south central Andes. J. Geophys. Res. 108, 2289,doi:10.1029/2002JB001879.

Kelleher, J., McCann, W., 1976. Buoyant zones, great earthquakes, and unstableboundaries of subduction. J. Geophys. Res. 81 (26), 4885–4896.

Kopp, H., Flueh, E.R., Papenberg, C., Klaeschen, D., 2004. Seismic investi-gations of the O’Higgins Seamount Group and Juan Fernandez Ridge:aseismic ridge emplacement and lithosphere hydration. Tectonics 23, TC2009,doi:10.1029/2003TC001590.

Koppers, A., Watts, A.B., 2010. Intraplate seamounts as a window into deep earthprocesses. Oceanography 23 (1), 42–57.

Lay, T., Ammon, C.J., Kanamori, H., Koper, K.D., Sufri, O., Hutko, A.R., 2010. Tele-seismic inversion for rupture process of the 27 February 2010 Chile (Mw = 8.8)earthquake. Geophys. Res. Lett. 37, L13301, doi:10.1029/2010GL043379.

Leyton, F., Ruiz, J., Campos, J., Kausel, E., 2009. Intraplate and interplate earthquakesin Chilean subduction zone: a theoretical and observational comparison. Phys.Earth Planet Int. 175, 1–2.

McNutt, M., Bonneville, A., 2000. A shallow, chemical origin for the Marquesas Swell.Geochem. Geophys. Geosyst. 1, 1014, doi:10.1029/1999GC000028.

Mendoza, C., Hartzell, S., Monfret, T., 1994. Wide-band analysis of the 3 March 1985Central Chile earthquake: overall source process and rupture history. Bull. Seism.Soc. Am. 84 (2), 269–283.

Moreno, M., Rosenau, M., Oncken, O., 2010. Maule earthquake slip correlateswith pre-seismic locking of Andean subduction zone. Nature 467, 198–202,doi:10.1038/nature09349.

Moreno, M., Bolte, J., Klotz, J., Melnick, D., 2009. Impact of megathrust geometryon inversion of coseismic slip from geodetic data: Application to the1960 Chileearthquake. Geophys. Res. Lett. 36, L16310, doi:10.1029/2009GL039276.

Pilger, R.H., 1984. Cenozoic plate kinematics, subduction and magmatism, SouthAmerican Andes. J. Geol. Soc. Lond. 141, 793–802.

Plafker, G., Savage, J., 1970. Mechanism of the Chilean earthquake of May 21 and 22,1960. Geol. Soc. Am. Bull. 81, 1001–1030.

Pritchard, M., Norabuena, E., Ji, C., Boroschek, R., Comte, D., Simons, M., Dixon, T.,Rosen, P., 2007. Geodetic, teleseismic, and strong motion constraints on slipfrom recent southern Peru subduction zone earthquakes. J. Geophys. Res. 112,B03307, doi:10.1029/2006JB004294.

Reichert, C., Schreckenberger, B., SPOC Team, 2002. Fahrtbericht SONNE-Fahrt SO-161 Leg 2 and 3 SPOC, Subduktionsprozesse vor Chile-BMBF-Forschungsvorhaben 03G0161A- Valparaiso 16.10.2001–Valparaiso 29.11.2001,Bundesanst. fuer Geowissenschaften und Rohstoffe, Hannover, Germany.

Rosenbaum, G., Giles, D., Saxon, M., Betts, P., Weinberg, R., Duboz, C., 2005.Subduction of the Nazca Ridge and the Inca Plateau: insights into theformation of ore deposits in Peru. Earth Planet Sci. Lett. 239, 18–32,doi:10.1016/j.epsl.2005.08.003.



Robinson, D.P., Das, S., Watts, A.B., 2006. Earthquake rupture stalled by a subductingfracture zone. Science 312, 1203–1205.

Ruegg, J.C., Rudloff, A., Vigny, C., Madariaga, R., de Chabalier, J.B., Campos, J., Kausel,E., Barrientos, S., Dimitrov, D., 2009. Interseismic strain accumulation measuredby GPS in the seismic gap between Constitución and Concepción in Chile. Phys.Earth Planet Int. 175 (1–2), 78–85.

Ruegg, J.-C., Oclay, M., Lazo, D., 2001. Co, post and pre(?)-seismic displacementsassociated with the Mw 8.4 Southern Peru earthquake of 23 June 2001 fromcontinuous GPS measurements. Seism. Res. Lett. 72, 673–678.

Ruff, L., 1989. Do trench sediments affect great earthquake occurrencein subduction zones? Pure Appl. Geophys. 129, 263–282, doi:10.1007/BF00874629.

Ruff, L., Miller, A., 1994. Rupture process of large earthquakes in the northernMexico subduction zone. Pure Appl. Geophys. 142, 101–172, doi:10.1007/BF00875970.

Sandwell, D., Smith, W., 1997. Marine gravity anomaly from Geosat and ERS 1 satel-lite altimetry. J. Geophys. Res. 102 (B5), 10039–10054.

Scherwath, M., Contreras-Reyes, E., Flueh, E., Grevemeyer, I., Krabbenhoeft, A.,Papenberg, C., Petersen, C.J., Weinrebe, R.W., 2009. Deep lithospheric structuresalong the southern central Chile margin from wide-angle P-wave modelling.Geophys. J. Int. 179 (1), 579–600.

Scholz, C.H., Small, C., 1997. The effect of seamount subduction on seismic coupling.Geology 25, 487–490.

Tassara, A., Goetze, H.-J., Schmidt, S., Hackney, R., 2006. Three-dimensional densitymodel of the Nazca plate and the Andean continental margin. J. Geophys. Res.111, B09404, doi:10.1029/2005JB003976.

Tebbens, S.F., Cande, S.C., Kovacs, L., Parra, J.C., LaBrecque, J.L., Vergara, H., 1997. TheChile ridge: a tectonic framework. J. Geophys. Res. 102 (B6), 12035–12060, 726,doi:10.1029/96JB02581.

Tichelaar, B., Ruff, L., 1991. Seismic coupling along the chilean subduction zone. J.Geophys. Res. 96 (B7), 11997–12022.

Yánez, G., Ranero, C.R., Díaz, J., 2001. Magnetic anomaly interpretation across thesouthern central Andes (32–34◦S): The role of the Juan Fernández Ridge in thelate tertiary evolution of the margin. J. Geophys. Res. 106 (B4), 6325–6345.

Author's personal copy · Author's personal copy 50 E. Contreras-Reyes, D. Carrizo / Physics of the...

Documents

Transcript of Author's personal copy · Author's personal copy 50 E. Contreras-Reyes, D. Carrizo / Physics of the...