Earth and Planetary Science Letters 405 (2014) 142–155

Contents lists available at ScienceDirect

Earth and Planetary Science Letters

www.elsevier.com/locate/epsl

Anatomy of a megathrust: The 2010 M8.8 Maule, Chile earthquake

rupture zone imaged using seismic tomography
Stephen P. Hicks a,∗ , Andreas Rietbrock a , Isabelle M.A. Ryder a , Chao-Shing Lee b ,
Matthew Miller c
a
School of Environmental Sciences, University of Liverpool, Liverpool, Merseyside, L69 3GP, UK
b
Institute of Applied Geosciences, National Taiwan Ocean University, Keelung, Taiwan
c
Departamento de Geofísica, Universidad de Concepción, Concepción, Chile

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

Article history: Knowledge of seismic velocities in the seismogenic part of subduction zones can reveal how material
Received 25 May 2014 properties may influence large ruptures. Observations of aftershocks that followed the 2010 M w 8.8
Received in revised form 19 August 2014 Maule, Chile earthquake provide an exceptional dataset to examine the physical properties of a
Accepted 26 August 2014
megathrust rupture zone. We manually analysed aftershocks from onshore seismic stations and ocean
Available online 16 September 2014
bottom seismometers to derive a 3-D velocity model of the rupture zone using local earthquake
Editor: P. Shearer
tomography. From the trench to the magmatic arc, our velocity model illuminates the main features
Keywords: within the subduction zone. We interpret an east-dipping high P-wave velocity anomaly (>6.9 km/s)
Maule earthquake as the subducting oceanic crust and a low P-wave velocity (<6.25 km/s) in the marine forearc as the
Chile accretionary complex. We find two large P-wave velocity anomalies (∼7.8 km/s) beneath the coastline.
subduction These velocities indicate an ultramafic composition, possibly related to extension and a mantle upwelling
seismic zone during the Triassic.
seismic tomography
We assess the role played by physical heterogeneity in governing megathrust behaviour. Greatest slip
OBS
during the Maule earthquake occurred in areas of moderate P-wave velocity (6.5–7.5 km/s), where the
interface is structurally more uniform. At shallow depths, high fluid pressure likely influenced the up-dip
limit of seismic activity. The high velocity bodies lie above portions of the plate interface where there
was reduced coseismic slip and minimal postseismic activity. The northern velocity anomaly may have
acted as a structural discontinuity within the forearc, influencing the pronounced crustal seismicity in
the Pichilemu region. Our work provides evidence for how the ancient geological structure of the forearc
may influence the seismic behaviour of subduction megathrusts.
© 2014 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY-NC-ND
license (http://creativecommons.org/licenses/by-nc-nd/3.0/).

1. Introduction teristics of megathrust earthquakes, Lay et al. (2012) subdivide the

subduction megathrust into five distinct depth domains. Knowl-
Understanding the physical processes that control the behaviour edge of fault properties could reveal what governs this megathrust
of subduction zone megathrust faults remains one of earthquake segmentation and the physical origins of asperities and barriers.
seismology’s main goals. The conceptual subduction zone asper- Seismic imaging methods can shed light on fault zone material
ity model (Lay and Kanamori, 1981) was developed to explain properties. However, many subduction megathrusts lie offshore,
the inhomogeneous moment release of large earthquakes. Asperi- where seismic instrumentation is deficient. This uneven station
ties are regions of the plate interface that produce the strongest coverage reduces imaging capability and the mislocation of off-
slip, whereas barriers inhibit rupture. Roughness on the down- shore earthquakes. With its coastline lying <100 km to the trench
going plate, such as seamounts can act as either asperities (e.g. in places, the Central Chile margin (Fig. 1), is an ideal natural lab-
Abercrombie et al., 2001) or barriers (e.g. Kodaira et al., 2004). oratory to study the subduction interface.
Alternatively, features within the overriding plate, such as crustal In 2010, an M w 8.8 earthquake struck the Maule region of
batholiths (Sobiesiak et al., 2007), forearc basins (e.g. Song and Si- Central Chile. Following the earthquake, a dense seismometer net-
mons, 2003; Fuller et al., 2006) and faults (e.g. Audin et al., 2008) work was deployed on the forearc to record aftershocks (e.g.
can also influence rupture behaviour. Based on the seismic charac- Rietbrock et al., 2012). Fortunately, this onshore network was sup-
plemented by ocean-bottom seismometer (OBS) deployments, dra-
matically improving station coverage (Fig. 2). Studies of preseis-
* Corresponding author. mic locking (e.g. Moreno et al., 2010), the coseismic rupture (e.g.

http://dx.doi.org/10.1016/j.epsl.2014.08.028
0012-821X/© 2014 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/3.0/).
S.P. Hicks et al. / Earth and Planetary Science Letters 405 (2014) 142–155 143

Fig. 2. Map of the Maule segment along the South Central Chile margin coloured by
bathymetry/topography. The main rupture characteristics of the Maule earthquake
Fig. 1. Geotectonic characteristics and simplified geological map of the South Central are shown. The purple star shows the rupture’s epicentre (Hayes et al., 2013) and
Chilean margin. Morphotectonic units from Glodny et al. (2007) and geological map the blue contours represent coseismic slip distribution (>6 m) in 2 m intervals
redrawn after SERNAGEOMIN (2003), Melnick and Echtler (2006) and Vásquez et al. (Moreno et al., 2012). Triangles correspond to seismic stations used in the velocity
(2011). The red box in the inset map gives the location of the study area. Labels inversions. Black crosses indicate the horizontal grid nodes used in the 3-D inver-
show the names of the locations referred to in this paper. (For interpretation of the sion and red circles show preliminary locations of the 669 earthquakes used in the
references to colour in this figure legend, the reader is referred to the web version velocity inversions. Thick black lines show the location of cross-sections that are
of this article.) shown in this paper. (For interpretation of the references to colour in this figure
legend, the reader is referred to the web version of this article.)

Moreno et al., 2012) and postseismic deformation (e.g. Lin et al., 2. Characteristics of the Central Chile subduction zone
2013) describe the behaviour of the Maule megathrust at dif-
ferent stages of the seismic cycle. This behaviour suggests that Along the Central Chilean margin, earthquakes are driven by
spatial variations in physical properties could exist in the fault subduction of the Nazca plate beneath the South American plate
zone. at a rate of 74 mm/yr (e.g. DeMets et al., 2010). Large earthquake
The purpose of our study is to detect physical heterogene- ruptures along the margin tend to occur within highly coupled
ity along the Maule megathrust using seismic tomography. In segments (Métois et al., 2012). We focus on the Maule segment,
a preliminary tomographic study of the rupture zone based on which is demarcated by the Mocha Block to the south and the
automatically-determined P- and S-wave arrival times from early Juan Fernández Ridge to the north (Contreras-Reyes et al., 2013).
The Maule segment last fully ruptured during the great 1835 Con-
aftershocks, Hicks et al. (2012) identified a large seismic veloc-
cepción earthquake and was recognised as a mature seismic gap
ity anomaly lying along the plate interface. This previous study,
(e.g. Ruegg et al., 2009).
however, used only observations from onshore stations, resulting
Active- and passive-source seismic studies have established the
in reduced offshore imaging capability. To build on this work, we
deep subsurface structure in parts of the Maule segment. The
have manually analysed aftershock data and incorporated OBS sta- 7 km thick oceanic crust subducts at an angle of ∼6◦ beneath the
tions to derive a detailed three-dimensional (3-D) velocity model trench (Contreras Reyes et al., 2008; Moscoso et al., 2011), steep-
of the rupture zone. We examine the quality of our velocity model ening to ∼15◦ beneath the coastline (e.g. Haberland et al., 2009;
by analysing the resolution matrix and by testing characteristic Hayes et al., 2012). The marine forearc comprises two domains:
models. We compare fault velocity structure with behaviour dur- the frontal accretionary prism and the paleo-accretionary com-
ing the seismic cycle, providing valuable insight into the physical plex (outer wedge) (e.g. Contreras Reyes et al., 2008). The conti-
origins of asperities and barriers. nental Moho intersects the subducting plate beneath the eastern
144 S.P. Hicks et al. / Earth and Planetary Science Letters 405 (2014) 142–155

coastal ranges; however, its exact position is debated. Based on between April and June 2010, with many stations active until
local earthquake tomography, Hicks et al. (2012) place the slab- September 2010. Onshore data alone, however, is insufficient for
mantle intersection at ∼50 km depth, similar to that estimated accurately imaging the offshore region, where most coseismic and
south of the Maule segment (e.g. Haberland et al., 2009). Con- aftershock activity occurred.
versely, Dannowski et al. (2013) postulate that based on receiver Fortunately, following the Maule earthquake, UK and Taiwanese
functions, the intersection lies at ∼38 km depth in the Maule seg- institutions deployed two separate OBS networks in the rupture
ment. Further east, the continental crust thins to ∼30 km, defin- area (Fig. 2). The Taiwanese deployment comprised 17 OBS that
ing the underlying arch-shaped, high velocity continental mantle were initially active from 15th July to 8th August 2010. The sta-
(e.g. Haberland et al., 2009). Hicks et al. (2012) describe a high tions were then moved northward in a second stage from 14th
P-wave velocity anomaly on the plate interface beneath the coastal August to 6th September. The UK deployment had 10 OBS instru-
cordillera at 36◦ S that was interpreted as a subducted seamount. ments offshore of the Arauco Peninsula from August 2010 to March
The geology of the coastal cordillera (Fig. 1) encompasses a late- 2011.
Paleozoic paired intrusive-metamorphic belt with two series of
metasediments (e.g. Martin et al., 1999). The Western Series con- 4.2. Catalogue selection, travel-time picks and initial event locations
stitutes low-grade metapsammopelitic rocks with intercalations
of metabasite derived from an ancient accretionary prism. Along For a well-resolved tomography model, we require a uniform
the eastern coastal cordillera, late Paleozoic granite batholiths in- source–receiver distribution across the rupture area to ensure that
trude the Eastern Series. The composition of these granites in- resolving capability is even across the model. Due to the uneven
dicates a shallow, crustal-derived source (Lucassen et al., 2004). aftershock distribution, we used a set of criteria to select the event
Although these Paleozoic rocks dominate the surface geology, out- dataset. From an automatic catalogue (Rietbrock et al., 2012), we
crops of Triassic plutons lie along the coastline. Some of these first chose events with large magnitudes (M l > 4) so that many
intrusions contain fayalite, indicating a mantle-derived magmatic clear onsets are recorded throughout the network. Second, we sub-
source (Vásquez and Franz, 2008). The intrusions were emplaced divided the area into smaller 2500 km2 blocks and selected an
when the margin was undergoing post-orogenic collapse and rift- equal number of events in each, ensuring an even event distri-
ing, marking the transitional period between Gondwanan amalga- bution. Third, we selected events that were located within each
mation and contemporary Andean-style subduction (Vásquez et al., OBS network during their operational periods. Fortunately, both
2011). the northern and southern OBS networks were simultaneously ac-
tive for 15 days (22/08/10–06/09/10); therefore, as a final step, we
3. The 2010 Maule earthquake chose events located between these two networks during this time
window.
On February 27, 2010, an M w 8.8 earthquake ruptured a ∼500 The above criteria resulted in an initial catalogue of 710 events.
km long portion of the Maule segment. The earthquake nucleated We manually determined onset times of P- and S-waves for these
offshore, 25 km from the coastline (Fig. 2) (Hayes et al., 2013). events using the SDX software (http://doree.esc.liv.ac.uk:8080/sdx).
Coseismic slip models for the rupture show that most slip oc- Based on onset time uncertainties, we assigned each observation
curred between the trench and the coastline. The models further a weight as follows: Weight 0 (<0.04 s); Weight 1 (0.04–0.1 s);
reveal that two asperities were ruptured during the earthquake: Weight 2 (0.1–0.2 s); Weight 3 (0.2–1 s); Weight 4 (>1 s). Us-
one to the north, the other to the south of the mainshock epicen- ing these onset times, we located the events inside the one-
tre (Fig. 2). In this paper, we refer to the coseismic slip models dimensional (1-D) velocity model of Haberland et al. (2006) using
of Moreno et al. (2012) and Lin et al. (2013) since these use the HYPO71PC (Lee and Valdes, 1985). We rejected events with an az-
most complete datasets. Teleseismic back-projection reveals that imuthal GAP >270◦ and with fewer than 12 P-wave and 4 S-wave
high frequency radiation came from a deeper part of the fault, of observations. We also rejected all observations with weights of 3
which the overall pattern suggests triggering on distinct portions and 4. Applying these criteria reduced the initial dataset by 6%,
of megathrust (Kiser and Ishii, 2011). leaving a high-quality catalogue of 669 events (Fig. 2) with 38,000
Following the rupture, most aftershocks occurred along the P-wave and 13,000 S-wave onset times.
plate interface at 10–35 km depth, with a second band at
40–45 km depth (Lange et al., 2012; Rietbrock et al., 2012). Hicks 5. Velocity inversion strategy
et al. (2012) showed that the resulting gap in seismicity coin-
cides with the location of a high P-wave velocity anomaly. Another We used a staggered velocity inversion scheme (e.g. Haberland
feature of the aftershock sequence was intense shallow, normal et al., 2009; Collings et al., 2012) in which we inverted for 1-D
faulting seismicity in the north, near Pichilemu (e.g. Ryder et al., model, followed by a coarse 2-D model, a fine 2-D model, and fi-
2012). Bedford et al. (2013) and Lin et al. (2013) indicate that most nally a 3-D model. Such a strategy ensures that a smooth regional
postseismic deformation was aseismically released, assumed to be velocity model in the trench-perpendicular direction can be estab-
afterslip. Bedford et al. (2013) show that most afterslip occurred lished without leaving velocity artefacts (from the 1-D model) in
seaward of the coastline; Lin et al. (2013) suggest most afterslip poorly resolved regions, which may affect event locations. Estab-
occurred landward of the coastline. lishing a 2-D model with robust earthquake locations is therefore
an important step before resolving any 3-D (trench-parallel) ve-
4. Seismic data locity variations. At each step, we chose damping values from
trade-off curves of data variance versus model complexity (e.g.
4.1. Temporary seismic networks Eberhart Phillips, 1986). The inversion parameters at each stage
are summarised in Table S1.
Following the Maule earthquake, teams from Chile, the US and
Europe installed seismometers in the rupture area to record after- 5.1. 1-D inversion
shock seismicity. The International Maule Aftershock Deployment
(IMAD) comprised ∼160 three-component broadband instruments We selected events located inside the network (GAP < 180◦ ),
on the continental forearc (Fig. 2). Most stations were deployed leaving 627 events for the 1-D inversion. We inverted for P-wave
within one month after the earthquake. Station coverage peaked velocity (v p ) and S-wave velocity (v s ) using VELEST (Kissling et al.,
S.P. Hicks et al. / Earth and Planetary Science Letters 405 (2014) 142–155 145

Fig. 3. Results of the 1-D velocity model inversion. (a) 1-D velocity models showing the range of starting models, range of inverted models (grey shading), and the final
velocity model. (b) Map of 1-D station delay terms. The reference station is indicated by the black triangle.

1994). VELEST requires that all stations are located within the up- sion, we tested 2000 initial models that were generated by ran-
permost layer. However, the greatest station elevation is 2.2 km domly perturbing the velocity of each layer in our starting model
and deepest OBS station lies 5.4 km below sea level; a model (Haberland et al., 2006). In each inversion, we used a v p / v s ratio of
with an 8 km thick uppermost layer is impractical. We negated 1.79, as determined from Wadati diagram analysis. From the 2000
this problem by following the strategy of Husen et al. (1999), set- inversions, we selected the model with the lowest RMS residual as
ting station elevations to zero and allowing station terms to absorb our best v p model and then inverted for a 1-D v s model.
systematic travel-time errors. We kept station damping low in the The best-fitting 1-D model (Fig. 3a) has low v p of 5.1 km/s at
1-D inversion to ensure that station terms accounted for station shallow depths, increasing to 6.1 km/s at 5 km depth. Velocities of
elevations and regional 2-D velocity variations. For the 1-D inver- greater than 7 km/s are reached at 20 km depth. The uppermost
146 S.P. Hicks et al. / Earth and Planetary Science Letters 405 (2014) 142–155

layer of the minimum 1-D model is poorly constrained due to the nique has been used in local earthquake tomography studies (e.g.
lack of shallow events and expected velocity structure differences Haberland et al., 2009; Collings et al., 2012).
between the offshore and onshore areas. Velocities in the upper-
most layer are ∼0.8 km/s slower than the model of Haberland et 6. Inversion resolution
al. (2006), a result of the greater quantity of offshore observations
in our study. 6.1. Resolution tests
Station delays are important for generating accurate 1-D event
locations if the subsurface has significant 2-D velocity variations. Our first resolution test focussed on the full model resolution
We find large negative delays for both P-waves (>−3.0 s) and matrix (MRM). Nodes with good ray coverage have large diagonal
S-waves (>−2.7 s) at OBS stations located offshore of the trench elements of the MRM; nodes with poor coverage have small diag-
axis (Fig. 3b). The P-wave station terms increase toward the mag- onal elements of the MRM due to their dependency on adjacent
matic arc, with stations here having the largest P-wave delays grid points. This effect is known as smearing. We estimated the
(<1.7 s). There is also a trench-parallel variation in onshore P-wave magnitude of smearing by calculating the spread function (Toomey
station terms, with values more positive in the south. In contrast, and Foulger, 1989), which assesses the ratio between off-diagonal
S-wave delays are large and positive at stations on the continental and diagonal terms. To accurately visualise the direction and size
shelf (up to 2.9 s). The systematic trench-perpendicular variation of smearing of nodes, we contoured each row of the MRM at the
in station delays mainly results from dipping structures in the sub- 70% value of its corresponding diagonal element.
surface, and to a lesser extent from station elevation differences. As another resolution test, we analysed the sensitivity of our
velocity models to the event catalogue; we achieved this by per-
5.2. 2-D and 3-D tomographic inversion forming a bootstrap resampling. Compared to formal MRM anal-
ysis, bootstrap resampling is useful because noise is intrinsically
For the 2-D and 3-D tomographic inversions, we used contained within the dataset. Calvert et al. (2000) suggest that
SIMUL2000 (Thurber, 1983; Eberhart Phillips and Michael, 1998). such event-based resampling should produce similar results to re-
This algorithm simultaneously inverts for seismic velocity and sampling individual picks. We randomly resampled the event cat-
hypocentral parameters using an iterative damped least squares alogue, forming a catalogue of 530 events (80% of our original
method; it uses a direct inversion for v p / v s ratio to account for the catalogue). Once the 530 events were chosen, we randomly chose
reduced number of high-quality S-wave observations compared to duplicate events, ensuring the catalogue was of the same size as
P-wave observations. Velocities are inverted on a rectangular grid for the actual inversion (589 events). Using the same inversion
of nodes with linear B-spline interpolation. In each inversion, we workflow, the velocity models were then stored and the process
did not invert for station corrections and kept hypocentres fixed repeated 100 times.
for the first three iterations. As a final assessment of imaging capability, we carried out
We first inverted for a coarse 2-D model, in which a 3-D grid restoring resolution tests. By designing synthetic velocity models,
was used, but velocity nodes in the along-strike direction were we assessed the capacity of our dataset to resolve the geome-
kept fixed, effectively forming a 2-D inversion. The 2-D inversion try and amplitude of velocity anomalies. We calculated synthetic
grid had a uniform horizontal grid spacing of 25 km and 10 km travel times using the true source–receiver geometry. To reflect the
spacing in depth. Beneath the outer rise and trench, we linked true quality of our observations, Gaussian noise was added to the
vertically adjacent nodes at depths of 15 km or greater due to travel times with a standard deviation depending on onset time
the diminished ray coverage. Linking nodes (Thurber and Eber- uncertainties (0.04–0.2 s). We then inverted the travel-times us-
hart Phillips, 1999) allows for coarser node spacing in parts of ing the same workflow as per our real inversion. We designed
the model. Without linking nodes, we found that the inversion in- the synthetic velocity model with two objectives in mind. First,
troduced low velocity artefacts in this part of the model, biasing we examined whether our inversion could constrain the geome-
earthquake depths beneath the outer rise. For the fine 2-D and 3-D try and amplitude of high velocity bodies lying along the plate
inversions, we used a minimum horizontal grid spacing of 16 km interface that were identified by Hicks et al. (2012). Second, we
and a minimum vertical spacing of 8 km. For the 3-D inversion, tested whether we could resolve lateral variations of v p / v s ratio in
we introduced a set of nodes in the trench-parallel direction, each the subducting oceanic crust. Our input model also comprised the
55 km apart, providing 13 nodes to image velocity variations in main domains of the Central Chile subduction zone identified in
the trench-parallel direction (Fig. 2). This nodal spacing allows for previous studies (e.g. Haberland et al., 2009). The final 3-D model
a finer resolution image than that achieved by Hicks et al. (2012). used as input for the restoring resolution test is shown in Fig. 4
For the coarse 2-D inversion, we used events inside the network ( v p ) and Fig. S3 ( v p / v s ).
(GAP < 180◦ ) to establish the main velocity structure. Nonethe-
less, a test showed that when the closest stations were located 6.2. 2-D model resolution
more than 40 km from an earthquake, its depth was poorly con-
strained. This finding was important for events located in between MRM analysis of the 2-D v p model (Fig. S1) shows that most
the two OBS networks (Chen et al., 1982; Frohlich et al., 1982). nodes have large diagonal elements and symmetric resolution ker-
Therefore, we only chose events that had the closest two stations nels. This finding indicates that in most parts, the model is well
at an epicentral distance of less than 40 km. These criteria resulted resolved up to the trench and to depths of 80–90 km beneath
in a catalogue of 589 events for the coarse 2-D inversion. Once the the Central Depression. Compared with the tomographic model
main velocity structure was established, we subsequently relaxed of Hicks et al. (2012) for the Maule region, the inclusion of OBS
the GAP criterion to 210◦ for the fine 2-D inversion to improve ray data dramatically improves imaging resolution of the marine fore-
coverage at the edges. arc. Beneath the magmatic arc, at depths of 20–90 km, diagonal
Finally, to improve the lateral imaging capability of our inver- elements are small with large and elongate resolution contours, in-
sions, we applied an extra smoothing technique. We performed dicating poor resolution. Resolution is also poor west of the trench,
two additional inversions, each shifting the horizontal node loca- beneath the outer rise. From the results of the bootstrapped event
tions by a third of the nodal spacing, and calculating the average resampling (Fig. S2), v p is well constrained throughout the model
velocity of these. For the 3-D inversion, we also performed this (σ < 0.01 km/s), indicating that the model is insensitive to the
averaging in the trench-parallel direction. Such an averaging tech- event selection criteria. Overall, the continental forearc (10–40 km
S.P. Hicks et al. / Earth and Planetary Science Letters 405 (2014) 142–155 147

Fig. 4. (left) Synthetic 3-D v p input model and (right) corresponding inversion results for the restoring resolution test. Locations of the five cross-sections are shown in Fig. 2.
Black crosses indicate the location of inversion nodes and white dots indicate the locations of the 669 earthquakes used. Regions of the model with good resolution are
bounded by the thick grey line. Regions with reduced resolution are faded; regions with no resolution are left blank. The location of the coastline is denoted by the white
triangle.

depth) is the best-resolved part of the 2-D v p model. The resolu- Based on the combined interpretation of the results from our
tion tests indicate that the size of resolved features varies with resolution tests, spread values of less than 2.1 and 4.2 represent
depth due to the irregular source–receiver distribution. For ex- well-resolved areas in the 2-D v p and v p / v s ratio models, re-
ample, we are unable to sufficiently resolve a thin low-velocity spectively. Laterally, we are able to resolve the subducting oceanic
structure at 50–80 km depth (Fig. S4), such as that imaged at the lithosphere from the trench to ∼80 km depth beneath the mag-
base of the continental mantle by Haberland et al. (2009). This matic arc. Both the marine and continental forearcs are well re-
poor resolution at greater depths is due to the lack of intermediate solved.
depth aftershocks beneath the Central Depression and magmatic
6.3. 3-D model resolution
arc (Fig. 2).
For the 2-D v p / v s ratio model, spread values are higher and
From the restoring resolution tests (Fig. 4), the 3-D inversion is
resolution kernels are larger than in the v p model. This is likely able to accurately recover the shape and amplitude of the input
due to our dataset comprising 66% fewer S-wave than P-wave ob- anomalies. In the continental mantle, v p is overestimated by ∼13%
servations, and fewer S observations at larger epicentral distances. and velocities in the shallow marine forearc are ∼30% too fast.
Some vertical smearing is present at shallow and deep nodes be- Amplitudes of high velocity anomalies in the subducting crust are
neath the marine forearc, but overall, the well-resolved areas are well recovered. Likewise, for v p / v s ratio, the inversion is able to
comparable to that of the v p model. The maximum σ for v p / v s is recover most features of the input model, but is not able to fully
0.04. recover patches of high v p / v s ratio in the subducting lithosphere
148 S.P. Hicks et al. / Earth and Planetary Science Letters 405 (2014) 142–155

Fig. 5. Resolution estimate of our 3-D velocity model based on analysis of the model resolution matrix. At each node, the different colours indicate spread function values,
black circles represent diagonal elements of the resolution matrix and green lines correspond to the 70% contour of the resolution kernel. Locations of the five cross-sections
are shown in Fig. 2. Black crosses indicate the location of inversion nodes. Regions of the model with good resolution are bounded by the thick grey line. The location of the
coastline is denoted by the white triangle. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

(Fig. S3). In the Central Depression, shallow marine forearc and 7. Results and discussion
continental mantle, v p / v s values are exaggerated by around 2%.
MRM analysis of the 3-D model (Fig. 5) shows that the well- 7.1. Description and interpretation of velocity models
resolved area in each cross-section is broadly similar to that of
the 2-D v p model. Nodes in the southernmost cross-section have The 2-D and 3-D models show the regional first-order velocity
strong smearing beneath the magmatic arc. The offshore region is structure previously observed for the Central Chilean margin (e.g.
poorly resolved at ∼36◦ S due to the lack of OBS coverage here. Haberland et al., 2009). The 2-D model is shown on a cross-section
In the north, however, nodes are well resolved up to the trench in Fig. 6. The 3-D model is displayed on vertical and horizontal
because of the denser OBS coverage. Along the onshore forearc, sections in Fig. 7 and Fig. 8, respectively. Features described in this
resolution is best in the north due to abundant crustal seismic- section correspond to labels in the figures.
ity. Although its resolution is lower, the v p / v s ratio anomalies
are real (as shown by the restoring resolution test) but anomalies 7.1.1. Subducting oceanic lithosphere
could be averaged over a distance greater than the node spacing The most prominent feature of the velocity models is an east-
(Eberhart Phillips et al., 2005). Overall, spread values of less than dipping structure with high v p (6.9–8 km/s), (labelled ‘oc’). This
2.1 indicate good resolution in the 3-D v p model. The resolving ca- feature also has a strong v p gradient and elevated v p / v s ratio of
pability of the 3-D v p / v s ratio model is broadly similar to that of 1.80–1.85 (Poisson’s ratio of 0.28–0.29) along the main band of
the v p model, with spread values of less than 4.1 indicating good seismicity. These velocities are in agreement with previous stud-
resolution. ies in the area (e.g. Contreras Reyes et al., 2008; Haberland et al.,
S.P. Hicks et al. / Earth and Planetary Science Letters 405 (2014) 142–155 149

composition (Christensen, 1996). This interpretation is supported

by the widespread outcrops of granite across the Coastal Cordillera
(Fig. 1) that likely extend through the upper crust (Groß and
Micksch, 2008). Such velocities could also correspond to metab-
asite and metagreywacke compositions of the Western and Eastern
Series, respectively (Christensen, 1996; Krawczyk et al., 2006). In
the shallow crust (<10 km depth), low v p (<6 km/s) corresponds
to sediments in the Central Depression (labelled ‘cd’).
In the lower forearc, prominent high v p anomalies (7.6–
8.0 km/s) lie beneath the coast. One is located at 36◦ S (hereafter
referred to as the Cobquecura anomaly; labelled ‘CA’); the other to
the north at 34◦ S (hereafter referred to as the Pichilemu anomaly;
labelled ‘PA’), (Figs. 7 and 8). Based on the 7.5 km/s v p contour,
the Cobquecura anomaly is up to 40 km wide and 20 km thick, in-
tersecting the plate interface at its base. This anomaly represents a
significant velocity increase of 8% relative to the input 2-D model
(Fig. S5). The smaller Pichilemu anomaly lies further above the
plate interface. Crucially, both anomalies have a slightly elevated
v p / v s ratio of ∼1.81 (Poisson’s ratio of 0.28) and a strong posi-
tive signature in the forearc Bouguer gravity field, which helps to
Fig. 6. 2-D velocity model plotted on a cross-section oriented perpendicular to the
trench. Crosses indicate grid nodes and white circles give the 2-D locations of events constrain their composition. More specifically, there is a moderate
from the tomographic inversion. Based on analysis of the MRM and the characteris- positive correlation between gravity and v p in the lower forearc
tic model tests, limits of well-resolved areas are given by the thick grey line. Regions (Fig. 8b).
with diminished resolution are faded; unresolved regions are left blank. Labels refer Interpretation of the seismic velocities and gravity signal asso-
to the following features that are discussed in the text: mf = marine forearc; oc =
ciated with the Cobquecura and Pichilemu anomalies clearly in-
oceanic crust; cf = upper forearc crust; cd = Central Depression; cm = continental
mantle. The thick black line indicates the location of our calculated plate interface. dicates dense, ultramafic material. Weakly-serpentinised peridotite
The purple star denotes the hypocentral location of the Maule earthquake (Hayes et (<20%) at these depths can explain the velocities (Christensen,
al., 2013). The location of the coastline is denoted by the white triangle. (For inter- 2004). Serpentinised peridotite at the continental forearc’s base
pretation of the references to colour in this figure legend, the reader is referred to could represent a subducted oceanic topographic high, such as a
the web version of this article.)
seamount. However, our improved velocity model indicates that
the Cobquecura anomaly is larger and seismically faster than pre-
2009) and indicative of either hydrated oceanic crustal material
viously thought (Hicks et al., 2012). Given these findings, we find a
(e.g. Hacker and Abers, 2004) or high pore fluid pressure. Based
seamount interpretation less plausible. Dense material in the lower
on velocity contours alone and the high v p / v s ratio anomaly, we
forearc could be a deeper manifestation of the Paleozoic granite
are unable to trace the oceanic lithosphere to depths of more than
batholith; however, the elevated v p / v s ratio rules out the possi-
50 km, indicating that the crust has a lower pore pressure or is
bility of residual intrusive material lying at its base (Husen et al.,
less hydrated at these depths. From our tomographic images, we
2000; Reyners et al., 2006).
are unable to determine the exact location of the oceanic Moho.
The surface projection of the ultramafic bodies corresponds rea-
Assuming a 7 km thick crust (e.g. Contreras Reyes et al., 2008), the
sonably well with the location of Triassic intrusions (Fig. S7). South
mid-lower oceanic mantle has v p of 7.8–8.5 km/s and a low v p / v s
of 37.5◦ S, a notable absence of high velocity bodies (Haberland et
ratio (1.70–1.76) (Poisson’s ratio of 0.24–0.26), suggesting it is not
al., 2009) is consistent with a lack of Triassic intrusions in this
hydrated.
region (Fig. 1). In contrast to the Paleozoic granites, these intru-
sions contain a stronger mantle source signature (Vásquez et al.,
7.1.2. Marine forearc
2011), possibly relating to the underlying ultramafic bodies. The
The offshore forearc region (labelled ‘mf ’) comprises low v p
Triassic intrusions were emplaced during extension (Vásquez et al.,
(4.75–6.25 km/s) and a high v p / v s ratio of (1.9–2.2). The location
2011) that was facilitated by either slab detachment (Parada et al.,
of such velocities is consistent with sediments and meta-sediments
1999), asthenospheric upwelling (Franzese and Spalletti, 2001) or
in the frontal prism and outer wedge, respectively. These v p / v s
slab steepening (Vásquez et al., 2011). It is plausible that any of
ratios correspond to a Poisson’s ratio of 0.31–0.37. Tsuji et al.
these mechanisms could have emplaced mantle material beneath
(2008) relate such values to overpressured sediments in a pore
the ancient volcanic arc; remnants of this material could now lie
pressure model for the Nankai accretionary prism. Beneath the
at the base of the present-day forearc.
coastline, a strong horizontal v p gradient represents the bound-
Interpreting which mechanism is responsible for the emplace-
ary between the outer wedge and the upper continental fore-
ment of these anomalous blocks of mantle material depends on
arc. Studies find similar velocities in the marine forearc south of
their regional extent. Although large outcrops of Triassic intrusives
the Maule segment (Contreras Reyes et al., 2008; Haberland et
have been mapped north of the Maule segment, in the Valparaiso
al., 2009) and in other subduction zones (Reyners et al., 2006;
area (Fig. 1), they do not extend south of the Maule segment.
Collings et al., 2012). Overpressured fluids could result from either Given the discrete nature of these blocks, it is possible that the
dehydration of the oceanic crust (e.g. Kodaira et al., 2004) or the tectonic process responsible for their emplacement was more lo-
smectite–illite transition at 100–150 ◦ C (Moore and Saffer, 2001). calised and limited to the central Chile region. Understanding the
Based on a thermal model for South Central Chile (Völker et al., structure and petrology of such localised blocks of mantle material
2011), the temperature of this phase transformation coincides with in the lower forearc crust could provide constraints on the genesis
the eastward limit of the frontal prism. of arc magmas and could decipher models of localised slab pro-
cesses, such as slab melting.
7.1.3. Continental forearc
The upper continental forearc (labelled ‘cf ’) beneath the Coastal 7.1.4. Continental mantle
Cordillera has a moderate v p of 5.5–7.0 km/s and reduced v p / v s Beneath the easternmost Coastal Cordillera, at depths of 25–
ratio of ∼1.71 (Poisson’s ratio of 0.24), consistent with a granitic 35 km, v p exceeds 7.25 km/s, defining a dome-shaped feature that
150 S.P. Hicks et al. / Earth and Planetary Science Letters 405 (2014) 142–155

Fig. 7. 3-D velocity model plotted along five cross-sections, of which the locations are labelled on the far right and shown in Fig. 2. Labels refer to the following features
that are discussed in the text: PA = Pichilemu anomaly; CA = Cobquecura anomaly. Resolution limits are defined in the same way as Fig. 6. The location of the coastline is
denoted by the white triangle.

is present along the margin (labelled ‘cm’). Its location, regional plate interface estimates (e.g. Haberland et al., 2009; Hayes et
geometry and velocity suggest that it represents the continental al., 2012), fitting a second-order polynomial through these events.
mantle wedge. Low-moderate v p / v s ratios of ∼1.76 (Poisson’s ratio Since seismicity did not reach the trench, we fixed the shal-
of 0.26) in its western part are indicative of unserpentinised man- lowest part of the interface to the trench at 7 km depth (e.g.
tle (Carlson and Miller, 2003). Beneath the magmatic arc, we ob- Contreras Reyes et al., 2008). In 3-D, we followed a similar work-
serve areas of elevated v p / v s , suggesting hydrated asthenospheric flow proposed by Hayes and Wald (2009). We find that our re-
mantle, although its extent is unresolved. vised interface geometry (Figs. 6, 7 and Fig. S7) corresponds to the
By inspection of the velocity contours that intersect the plate mean depth of thrust mechanism aftershocks (Agurto et al., 2012;
interface, we find that the continental Moho, represented by a v p Hayes et al., 2013), (Fig. S8). By projecting this interface through
of ∼7.75 km/s, intersects the subducting plate interface at a depth our 3-D velocity model, we can assess the velocity structure of the
of 45–50 km. This depth estimate is similar to that of Bohm (2004) plate interface. To account for both hypocentre and velocity un-
and Haberland et al. (2009), but deeper than that of Dannowski et certainties, we estimate the error in plate interface velocity (see
al. (2013). From our perspective, high v p anomalies in the lower Note S1).
forearc could contaminate the lower resolution and laterally aver- The velocity structure of the plate interface region shows a clear
aged receiver function image, resulting in a misplaced continental zonation of seismic properties with depth (Fig. 9 and Fig. S9).
Moho. Given the spacing of nodes with depth, (see Section 5.2), our
model represents the average plate interface velocity over a total
7.1.5. Plate interface zone thickness of 6–8 km. Therefore, any thin (<4 km thick) low veloc-
The curved region of seismicity, dipping away from the trench, ity layers along the plate interface (e.g. Haberland et al., 2009) are
defines the plate interface. To define geometry of the fault, we unlikely to be imaged. Along the shallowest part of the plate inter-
selected all events that lie within ±15 km depth of previous face, beneath the frontal prism, v p is relatively low (∼6.6 km/s)
S.P. Hicks et al. / Earth and Planetary Science Letters 405 (2014) 142–155 151

Fig. 8. (a) 3-D velocity model plotted as depth sections. Labels refer to the following features that are discussed in the text: mf = marine forearc; oc = oceanic crust; cf
= upper forearc crust; cd = Central Depression; PA = Pichilemu anomaly; CA = Cobquecura anomaly. Resolution limits are defined in the same way as Fig. 6. (b) Forearc
Bouguer gravity anomaly (Hicks et al., 2012) overlain by contours of v p model at 25 km depth. Contours are given for v p in the range 7.25–8 km/s. Inset: scatter plot showing
the correlation between forearc Bouguer gravity anomaly and forearc v p at 25 km, sampled at 0.2◦ intervals of latitude and longitude.

Fig. 9. Seismic velocities along the plate interface from the 2-D model plotted as a function of depth for (a) v p and (b) v p / v s ratio. Grey shading shows our estimated
uncertainty in these seismic velocities. (c) Histogram of plate interface aftershock earthquake depths and normalised coseismic slip distribution as a function of depth. Trench-
perpendicular slip profiles traverse the northern slip asperity. Aftershock hypocentres are relocated from the catalogue of Rietbrock et al. (2012). (d) Depth-segmentation of
the plate interface region based on interpretation of the seismic velocities and seismic character of the Maule megathrust. Letters in the left-hand column correspond to the
megathrust domains of Lay et al. (2012).

and v p / v s ratio is high (1.87–2.05). Beneath the outer wedge, face has moderate v p (6.25–7 km/s) and elevated v p / v s ratio of
v p sharply increases with depth (∼0.03 km/s per km) and v p / v s 1.88 (Fig. 10a). The interface was also strongly locked (>90%) be-
ratio decreases to ∼1.84. Beneath the lower forearc crust, v p lev- neath the coast, at the base of the Cobquecura ultramafic body.
els out at ∼7.5 km/s, but reaches up to 8 km/s where the plate For the Arauco peninsula region, Moreno et al. (2014) identify a
interface intersects the high velocity bodies (Fig. 9a). At depths of correlation between weakly locked regions of the interface and ele-
∼44 km, v p decreases to ∼7.4 km/s. Beneath the Central Depres- vated v p / v s ratio due to high fluid pressures. We also find a strong
sion, v p further increases to ∼8.5 km/s and v p / v s ratio decreases correlation (r = −0.71) for the region south of 37◦ S (Fig. S12).
to ∼1.78. Given this segmented velocity structure of the plate in- However, we do not find such a clear correlation for the whole
terface, the next logical step is to consider whether this influences rupture zone (Fig. S11a). This finding indicates that factors con-
megathrust behaviour. trolling preseismic locking may change across the Maule segment.

7.2. Correlating megathrust behaviour with plate interface velocity

structure 7.2.2. Nucleation and rupture of the Maule earthquake
The Maule earthquake nucleated in a region of high v p
7.2.1. Preseismic coupling of the megathrust (∼7.2 km/s) and strong dip-parallel v p gradient, at the periph-
Based on the preseismic locking model of Moreno et al. (2010), ery of the Cobquecura anomaly (Fig. 10b). This inference is in line
the Maule megathrust before the 2010 earthquake was strongly with Tassara (2010), who shows that ruptures along the Andean
coupled (>90%) beneath the outer wedge, where the plate inter- margin generally nucleate at the edge of geological heterogeneities
152 S.P. Hicks et al. / Earth and Planetary Science Letters 405 (2014) 142–155

Fig. 10. Distribution of (top) v p and (bottom) v p / v s ratio along the plate interface region compared with behaviour of the Maule megathrust over the seismic cycle. Labels
refer to the features discussed in the text and are defined in the caption of Fig. 7. (a) Contours of preseismic locking of the megathrust (Moreno et al., 2010) (blue lines)
are for locking of greater than 0.7 and given in intervals of 0.1. (b) Mainshock epicentral location from Hayes et al. (2013). Coseismic slip distribution (Moreno et al., 2012)
plotted as for Fig. 2. White circles give the location of high frequency energy release during the Maule earthquake (Kiser and Ishii, 2011). (c) Contours of afterslip >1 m (Lin
et al., 2013) (blue lines) plotted in 0.2 m intervals. Grey circles show the distribution of relocated plate interface aftershock seismicity. (For interpretation of the references
to colour in this figure legend, the reader is referred to the web version of this article.)

in the forearc. A local increase in stress at the edge the Cobquecura than 7.25 km/s. The northern and southern slip asperities corre-
anomaly could have led to onset of the rupture. spond to weaker v p gradient with depth. This more homogeneous
Following nucleation, most coseismic slip occurred along the velocity structure is partly due to the lack of high v p anomalies be-
plate interface beneath the outer wedge, where we find mod- neath the coast in the Constitución and Arauco regions. Therefore,
erate v p (∼6.5 km/s) and elevated v p / v s ratio (∼1.86), (Figs. 9 it is possible that slip localisation during the Maule earthquake
and 10b). Coseismic slip was minimal beneath the frontal prism was affected by the presence of long-lived mafic bodies in the
(v p < 6.25 km/s; v p / v s ratio >1.85) and beneath the continental lower forearc. Overall, there is strong negative correlation between
mantle wedge (v p > 7.5 km/s; v p / v s ratio <1.8). Coseismic slip v p and coseismic slip for the down-dip portion of the rupture
was reduced beneath the crustal forearc where v p becomes more (Fig. S13).
S.P. Hicks et al. / Earth and Planetary Science Letters 405 (2014) 142–155 153

At shallow depths along the fault (<17 km), coseismic slip was seismic characteristics deviate from Domain C. Therefore, we in-
reduced by 50% (Lin et al., 2013). At these depths, v p sharply stead propose a new domain, ‘g’, for the Maule megathrust. At
decreases and v p / v s ratio abruptly increases to >2 (Fig. 9). Com- greater depths (44–50 km), the predominance of high frequency
parison of slip with seismic velocity at shallow depths suggests radiation and localised aftershock clusters agrees well with Do-
that overpressure conditions (Spinelli et al., 2006) and material main C, where v p indicates a return to normal interface structure.
with low rigidity at the base of the frontal prism inhibited slip. It has been proposed that the maximum depth of the seismogenic
The smectite–illite transition may contribute to this overpressure zone is influenced by the intersection of the 350–450 ◦ C isotherm
(Moore and Saffer, 2001) and influence the transition between ve- with the plate interface (e.g. Hyndman et al., 1997). Based on
locity strengthening and velocity weakening regimes (e.g. Saffer Völker et al. (2011) however, these temperatures are not reached
and Marone, 2003). until 65–90 km depth, so temperature is unlikely to play a role in
For the high frequency part of the rupture (Kiser and Ishii, the deeper segmentation of the Maule megathrust.
2011), there is a striking interaction with velocity structure along
the plate interface (Fig. 10b). The locations of high frequency en- 8. Conclusions
ergy release appear to step down around the high velocity anoma-
lies, with most energy coming from deeper regions of the plate In this paper, we have presented a detailed seismic velocity
interface that have intermediate v p and elevated v p / v s ratio. High model (v p and v p / v s ratio) of the 2010 M w 8.8 Maule, Chile
frequency energy release during an earthquake can be caused by earthquake rupture zone. A dense aftershock dataset from onshore
sudden changes in rupture speed or slip (Madariaga, 1977). Based and offshore seismic networks allowed an in-depth study of phys-
on the coseismic slip distribution (Moreno et al., 2012), we pro- ical properties along the megathrust interface and in the overlying
pose that the rupture was slowed by the high velocity anoma- forearc.
lies beneath the coast, generating stopping phases at the rupture’s Our velocity model shows that two high velocity anomalies
down-dip termination. with v p > 7.5 km/s lie beneath the coast, at the base of the con-
tinental forearc in the central and northern parts of the rupture
7.2.3. Postseismic deformation following the Maule earthquake zone. We interpret these high velocity anomalies as large (up to
To investigate the relationship between velocity structure and 15 km thick), dense bodies of ultramafic peridotite. A comparison
aftershock seismicity, we relocated the full 2010 aftershock cata- with the surface geology suggests that these bodies could be relic
logue of Rietbrock et al. (2012), (see Supplementary Note 2 for a blocks of mantle.
discussion of how combining OBS data with high-quality, manu- By comparing the location of these anomalous bodies with the
ally determined onset times affects earthquake locations). We find behaviour of the earthquake, we show that they may have played
that the Cobquecura and Pichilemu anomalies clearly lie within the a role in controlling the down-dip and along-strike distribution of
distinct gap of plate interface aftershock seismicity (Fig. 10c and slip during the rupture. It also appears that these anomalies influ-
Fig. S15). Intense crustal seismicity also occurred along the west- enced the location of high frequency seismic energy. Hence, long-
ern edge of the Pichilemu anomaly. It is plausible that structural lived structural and compositional heterogeneities in the forearc
heterogeneity in the lower forearc influenced the distribution of can act as a rupture barrier during large earthquakes and can in-
plate interface and shallow crustal seismicity following the Maule fluence the along-strike segmentation of ruptures along the Central
earthquake. Tomographic images from Japan also show a similar Chile margin. These ultramafic blocks are also associated with a
relationship between crustal high velocities and aftershock seis- distinct gap in plate interface aftershock seismicity. The Pichilemu
micity (Kato et al., 2010, 2013). Significant afterslip occurred in velocity anomaly acted as a structural discontinuity, influencing
regions of intermediate plate interface v p (7.2–7.5 km/s) and ele- the focus of normal faulting aftershock seismicity in the overriding
vated v p / v s ratio (1.82–1.84) (Fig. 10c and Fig. S10c). This finding crust. At the shallow end of the seismogenic zone, overpressured,
suggests that afterslip following the Maule earthquake may have low rigidity sediments at the base of the frontal prism likely inhib-
been compositionally driven. ited shallow rupture during the Maule earthquake. The megathrust
fault beneath the Cobquecura anomaly was strongly locked be-
7.3. Depth-varying rupture properties and fault material properties fore the Maule earthquake, yet experienced little coseismic slip.
This part of the fault was therefore recognised by Moreno et al.
Based on the depth variation of subduction zone rupture char- (2012) as experiencing slip deficit. Since this study, there has been
acteristics, Lay et al. (2012) classify the megathrust into four dis- minimal postseismic slip along this portion of the fault, raising
tinct domains (A–D). From the physical properties and seismo- questions over the current state of stress and overall rheological
genic characteristics of the Maule megathrust, we can draw par- nature of the Cobquecura anomaly.
allels with this classification (Fig. 9c). The plate interface beneath Overall, P-wave velocities of greater than 7.5 km/s inhibited
the frontal prism (<17 km depth) was largely aseismic during seismic and aseismic slip both during and after the Maule earth-
the Maule rupture, corresponding to Domain A. The high v p / v s quake. Therefore, an understanding of physical properties along the
and Poisson’s ratios provide direct evidence for low rigidity ma- plate interface could help in determining the seismic hazard of
terial existing at shallow depths that was proposed by Bilek and a subduction zone. Seismic velocity could be used as a proxy in
Lay (1999) to influence aseismic behaviour and occasional slow, other subduction zones to estimate rupture size potential and the
tsunamigenic earthquakes. The depth of the plate interface beneath regions conducive to high frequency seismic energy release and lo-
the outer wedge (17–30 km) corresponds to Domain B, the most calisation of shallow crustal aftershock seismicity.
seismogenic part of the megathrust. For the Maule earthquake,
the greatest coseismic slip and aftershock activity occurred in this Acknowledgements
domain, and is where the plate interface is structurally more ho-
mogeneous. The IMAD dataset is available through IRIS and GFZ (http://
Domain C should theoretically lie at 35–55 km depth beneath www.iris.edu/dms/dmc; http://www.webdc.eu). We are grateful to
the continental forearc. However, at 30–44 km depth, where the all field crews that participated in the onshore and offshore de-
plate interface intersects ultramafic bodies in the lower forearc, co- ployments. We thank IRIS/PASSCAL, CNRS-INSU, GFZ and GEF/SEIS-
seismic slip was reduced and postseismic slip was minimal; such UK for providing the onshore seismic instruments. The Universidad
154 S.P. Hicks et al. / Earth and Planetary Science Letters 405 (2014) 142–155

de Chile and Universidad de Concepción provided valuable sup- Eberhart Phillips, D., Michael, A.J., 1998. Seismotectonics of the Loma Prieta, Cal-
port in the field, and we are also grateful to the CAP/OSU/IGM ifornia, region determined from three-dimensional V p , V p / V s , and seismicity.
J. Geophys. Res. 103, 21099. http://dx.doi.org/10.1029/98JB01984.
GPS teams for advice and logistical assistance. The National Tai-
Eberhart Phillips, D., Reyners, M., Chadwick, M., Chiu, J.M., 2005. Crustal hetero-
wan Ocean University and UK Ocean Bottom Instrument Consor- geneity and subduction processes: 3-D V p , V p / V s and Q in the southern
tium (OBIC) provided OBS instruments. Dr. Victor Ariel Gallardo North Island, New Zealand. Geophys. J. Int. 162, 270–288. http://dx.doi.org/
(Oceanography Department, Universidad de Concepción), Captain 10.1111/j.1365-246X.2005.02530.x.
Juan Vilches and Freddy Echeverria provided vital logistics sup- Franzese, J.R., Spalletti, L.A., 2001. Late Triassic–early Jurassic continental ex-
tension in southwestern Gondwana: tectonic segmentation and pre-break-
port for the offshore deployment. We are grateful to the Armada
up rifting. J. South Am. Earth Sci. 14, 257–270. http://dx.doi.org/10.1016/
de Chile for providing vessels. NERC funded the UK deployments S0895-9811(01)00029-3.
(grant NE/I005420/1). Amaya Fuenzalida provided insightful com- Frohlich, C., Billington, S., Engdahl, E.R., Malahoff, A., 1982. Detection and lo-
ments on the manuscript. S.P.H. is funded by a NERC studentship, cation of earthquakes in the Central Aleutian Subduction Zone using is-
land and ocean bottom seismograph stations. J. Geophys. Res. 87, 6853–6864.
NE/J50015X/1. We thank the editor and two anonymous reviewers
http://dx.doi.org/10.1029/JB087iB08p06853.
for their insightful comments. Fuller, C.W., Willett, S.D., Brandon, M.T., 2006. Formation of forearc basins
and their influence on subduction zone earthquakes. Geology 34, 65–68.
Appendix A. Supplementary material http://dx.doi.org/10.1130/g21828.1.
Glodny, J., Gräfe, K., Echtler, H., Rosenau, M., 2007. Mesozoic to Quaternary con-
tinental margin dynamics in South-Central Chile (36–42◦ S): the apatite and
Supplementary material related to this article can be found on-
zircon fission track perspective. Int. J. Earth Sci. 97, 1271–1291. http://dx.doi.org/
line at http://dx.doi.org/10.1016/j.epsl.2014.08.028. 10.1007/s00531-007-0203-1.
Groß, K., Micksch, U., 2008. The reflection seismic survey of project TIPTEQ—
References the inventory of the Chilean subduction zone at 38.2◦ S. Geophys. J. Int.
http://dx.doi.org/10.1111/j.1365-246X.2007.03680.x.
Abercrombie, R.E., Antolik, M., Felzer, K., Ekström, G., 2001. The 1994 Java tsunami Haberland, C., Rietbrock, A., Lange, D., Bataille, K., Dahm, T., 2009. Structure of the
earthquake: slip over a subducting seamount. J. Geophys. Res. 106, 6595–6607. seismogenic zone of the southcentral Chilean margin revealed by local earth-
http://dx.doi.org/10.1029/2000JB900403. quake traveltime tomography. J. Geophys. Res. 114, B01317. http://dx.doi.org/
Agurto, H., Rietbrock, A., Ryder, I., Miller, M., 2012. Seismic-afterslip characterization 10.1029/2008jb005802.
of the 2010 Mw 8.8 Maule, Chile, earthquake based on moment tensor inver- Haberland, C., Rietbrock, A., Lange, D., Bataille, K., Hofmann, S., 2006. Interac-
sion. Geophys. Res. Lett. 39, L20303. http://dx.doi.org/10.1029/2012GL053434. tion between forearc and oceanic plate at the south-central Chilean margin
Audin, L., Lacan, P., Tavera, H., Bondoux, F., 2008. Upper plate deformation and seis- as seen in local seismic data. Geophys. Res. Lett. 33. http://dx.doi.org/10.1029/
mic barrier in front of Nazca subduction zone: the Chololo Fault System and 2006GL028189.
active tectonics along the Coastal Cordillera, southern Peru. Tectonophysics 459, Hacker, B.R., Abers, G.A., 2004. Subduction Factory 3: an Excel worksheet and macro
174–185. http://dx.doi.org/10.1016/j.tecto.2007.11.070. for calculating the densities, seismic wave speeds, and H2 O contents of minerals
Bedford, J., Moreno, M., Baez, J.C., Lange, D., Tilmann, F., Rosenau, M., Heidbach, O., and rocks at pressure and temperature. Geochem. Geophys. Geosyst. 5, Q01005.
Oncken, O., Bartsch, M., Rietbrock, A., Tassara, A., Bevis, M., Vigny, C., 2013. http://dx.doi.org/10.1029/2003GC000614.
A high-resolution, time-variable afterslip model for the 2010 Maule Mw = Hayes, G.P., Bergman, E., Johnson, K.L., Benz, H.M., Brown, L., Meltzer, A.S., 2013.
8.8, Chile megathrust earthquake. Earth Planet. Sci. Lett. 383, 26–36. http:// Seismotectonic framework of the 2010 February 27 M w 8.8 Maule, Chile earth-
dx.doi.org/10.1016/j.epsl.2013.09.020. quake sequence. Geophys. J. Int. 195, 1034–1051. http://dx.doi.org/10.1093/gji/
Bilek, S.L., Lay, T., 1999. Rigidity variations with depth along interplate megath- ggt238.
rust faults in subduction zones. Nature 400, 443–446. http://dx.doi.org/10.1038/ Hayes, G.P., Wald, D.J., 2009. Developing framework to constrain the geometry of
22739. the seismic rupture plane on subduction interfaces a priori—a probabilistic
Bohm, M., 2004. 3-D Lokalbebentomographie der sudlichen Anden zwischen 36◦ approach. Geophys. J. Int. 176, 951–964. http://dx.doi.org/10.1111/j.1365-246X.
und 40◦ S. Freie Universitat, Berlin, Germany. 2008.04035.x.
Calvert, A., Sandvol, E., Seber, D., Barazangi, M., Roecker, S., Mourabit, T., Vi- Hayes, G.P., Wald, D.J., Johnson, R.L., 2012. Slab1.0: a three-dimensional model
dal, F., Alguacil, G., Jabour, N., 2000. Geodynamic evolution of the lithosphere of global subduction zone geometries. J. Geophys. Res. 117. http://dx.doi.org/
and upper mantle beneath the Alboran region of the western Mediterranean: 10.1029/2011JB008524.
constraints from travel time tomography. J. Geophys. Res. 105, 10871–10898. Hicks, S.P., Rietbrock, A., Haberland, C.A., Ryder, I.M.A., Simons, M., Tassara, A.,
http://dx.doi.org/10.1029/2000JB900024. 2012. The 2010 M w 8.8 Maule, Chile earthquake: nucleation and rupture prop-
Carlson, R.L., Miller, D.J., 2003. Mantle wedge water contents estimated from seismic agation controlled by a subducted topographic high. Geophys. Res. Lett. 39.
velocities in partially serpentinized peridotites. Geophys. Res. Lett. 30, 51–54. http://dx.doi.org/10.1029/2012GL053184.
http://dx.doi.org/10.1029/2002GL016600. Husen, S., Kissling, E., Flueh, E., Asch, G., 1999. Accurate hypocentre determination
Chen, A.T., Frohlich, C., Latham, G.V., 1982. Seismicity of the forearc marginal in the seismogenic zone of the subducting Nazca Plate in northern Chile using a
wedge (accretionary prism). J. Geophys. Res. 87, 3679–3690. http://dx.doi.org/ combined on-/offshore network. Geophys. J. Int. 138, 687–701. http://dx.doi.org/
10.1029/JB087iB05p03679. 10.1046/j.1365-246x.1999.00893.x.
Christensen, N.I., 1996. Poisson’s ratio and crustal seismology. J. Geophys. Res. 101, Husen, S., Kissling, E., Flueh, E.R., 2000. Local earthquake tomography of shallow
3139–3156. http://dx.doi.org/10.1029/95JB03446. subduction in north Chile: a combined onshore and offshore study. J. Geophys.
Christensen, N.I., 2004. Serpentinites, peridotites, and seismology. Int. Geol. Rev. 46, Res. 105, 28183–28198. http://dx.doi.org/10.1029/2000JB900229.
795–816. http://dx.doi.org/10.2747/0020-6814.46.9.795. Hyndman, R., Yamano, M., Oleskevich, D., 1997. The seismogenic zone of sub-
Collings, R., Lange, D., Rietbrock, A., Tilmann, F., Natawidjaja, D., Suwargadi, B., duction thrust faults. Isl. Arc 6, 244–260. http://dx.doi.org/10.1111/j.1440-1738.
Miller, M., Saul, J., 2012. Structure and seismogenic properties of the Mentawai 1997.tb00175.x.
segment of the Sumatra subduction zone revealed by local earthquake trav- Kato, A., Igarashi, T., Obara, K., Sakai, S., Takeda, T., Saiga, A., Iidaka, T., Iwasaki, T., Hi-
eltime tomography. J. Geophys. Res. 117, B01312. http://dx.doi.org/10.1029/ rata, N., Goto, K., Miyamachi, H., Matsushima, T., Kubo, A., Katao, H., Yamanaka,
2011JB008469. Y., Terakawa, T., Nakamichi, H., Okuda, T., Horikawa, S., Tsumura, N., Umino,
Contreras Reyes, E., Grevemeyer, I., Flueh, E.R., Reichert, C., 2008. Upper lithospheric N., Okada, T., Kosuga, M., Takahashi, H., Yamada, T., 2013. Imaging the source
structure of the subduction zone offshore of southern Arauco peninsula, Chile, regions of normal faulting sequences induced by the 2011 M9.0 Tohoku-Oki
at ∼38◦ S. J. Geophys. Res. 113, B07303. http://dx.doi.org/10.1029/2007JB005569. earthquake. Geophys. Res. Lett. 40, 273–278. http://dx.doi.org/10.1002/grl.50104.
Contreras-Reyes, E., Jara, J., Maksymowicz, A., Weinrebe, W., 2013. Sediment load- Kato, A., Miyatake, T., Hirata, N., 2010. Asperity and barriers of the 2004 Mid-Niigata
ing at the southern Chilean trench and its tectonic implications. J. Geodyn. 66, Prefecture earthquake revealed by highly dense seismic observations. Bull. Seis-
134–145. http://dx.doi.org/10.1016/j.jog.2013.02.009. mol. Soc. Am. 100, 298–306. http://dx.doi.org/10.1785/0120090218.
Dannowski, A., Grevemeyer, I., Kraft, H., Arroyo, I., Thorwart, M., 2013. Crustal Kiser, E., Ishii, M., 2011. The 2010 M w 8.8 Chile earthquake: triggering on multi-
thickness and mantle wedge structure from receiver functions in the ple segments and frequency-dependent rupture behavior. Geophys. Res. Lett. 38,
Chilean Maule region at 35◦ S. Tectonophysics 592, 159–164. http://dx.doi.org/ L07301. http://dx.doi.org/10.1029/2011GL047140.
10.1016/j.tecto.2013.02.015. Kissling, E., Ellsworth, W.L., Eberhart-Phillips, D., Kradolfer, U., 1994. Initial refer-
DeMets, C., Gordon, R.G., Argus, D.F., 2010. Geologically current plate motions. Geo- ence models in local earthquake tomography. J. Geophys. Res. 99, 19635–19646.
phys. J. Int. 181, 1–80. http://dx.doi.org/10.1111/j.1365-246X.2009.04491.x. http://dx.doi.org/10.1029/93JB03138.
Eberhart Phillips, D., 1986. Three-dimensional velocity structure in northern Califor- Kodaira, S., Iidaka, T., Kato, A., Park, J.O., Iwasaki, T., Kaneda, Y., 2004. High pore fluid
nia Coast Ranges from inversion of local earthquake arrival times. Bull. Seismol. pressure may cause silent slip in the Nankai Trough. Science 304, 1295–1298.
Soc. Am. 76, 1025–1052. http://dx.doi.org/10.1126/science.1096535.
S.P. Hicks et al. / Earth and Planetary Science Letters 405 (2014) 142–155 155

Krawczyk, C.M., Mechie, J., Lüth, S., Tašárová, Z., Wigger, P., Stiller, M., Brasse, H., Parada, M.A., Nyström, J.O., Levi, B., 1999. Multiple sources for the Coastal
Echtler, H.P., Araneda, M., Bataille, K., 2006. Geophysical signatures and active Batholith of central Chile (31–34◦ S): geochemical and Sr–Nd isotopic evi-
tectonics at the South-Central Chilean margin. In: The Andes Active Subduction dence and tectonic implications. Lithos 46, 505–521. http://dx.doi.org/10.1016/
Orogeny. Springer, New York, pp. 171–192. S0024-4937(98)00080-2.
Lange, D., Tilmann, F., Barrientos, S.E., Contreras-Reyes, E., Methe, P., Moreno, M., Reyners, M., Eberhart Phillips, D., Stuart, G., Nishimura, Y., 2006. Imaging subduc-
Heit, B., Agurto, H., Bernard, P., Vilotte, J.-P., 2012. Aftershock seismicity of the tion from the trench to 300 km depth beneath the central North Island, New
27 February 2010 Mw 8.8 Maule earthquake rupture zone. Earth Planet. Sci. Zealand, with V p and V p / V s . Geophys. J. Int. 165, 565–583. http://dx.doi.org/
Lett. 317–318, 413–425. http://dx.doi.org/10.1016/j.epsl.2011.11.034. 10.1111/j.1365-246X.2006.02897.x.
Lay, T., Kanamori, H., 1981. An asperity model of large earthquake sequences. In: Rietbrock, A., Ryder, I., Hayes, G., Haberland, C., Comte, D., Roecker, S., Lyon-Caen, H.,
Maurice Ewing Series: Earthquake Prediction: An International Review. Wash- 2012. Aftershock seismicity of the 2010 Maule Mw = 8.8, Chile, earthquake: cor-
ington, DC, pp. 579–592. relation between co-seismic slip models and aftershock distribution? Geophys.
Lay, T., Kanamori, H., Ammon, C.J., Koper, K.D., Hutko, A.R., Ye, L., Yue, H., Rushing, Res. Lett. 39, L08310. http://dx.doi.org/10.1029/2012GL051308.
T.M., 2012. Depth-varying rupture properties of subduction zone megathrust Ruegg, J.C., Rudloff, A., Vigny, C., Madariaga, R., De Chabalier, J.B., Campos, J., Kausel,
faults. J. Geophys. Res. 117, B04311. http://dx.doi.org/10.1029/2011JB009133. E., Barrientos, S., Dimitrov, D., 2009. Interseismic strain accumulation measured
Lee, W.H.K., Valdes, C., 1985. HYPO71PC: a personal computer version of the by GPS in the seismic gap between Constitución and Concepción in Chile. Phys.
HYPO71 earthquake location program. U.S. Geol. Surv. Open File Rep., 85–749. Earth Planet. Inter. 175, 78–85. http://dx.doi.org/10.1016/j.pepi.2008.02.015.
Lin, Y.N., Sladen, A., Ortega-Culaciati, F., Simons, M., Avouac, J.-P., Fielding, E.J., Ryder, I., Rietbrock, A., Kelson, K., Bürgmann, R., Floyd, M., Socquet, A., Vigny, C.,
Brooks, B.A., Bevis, M., Genrich, J., Rietbrock, A., 2013. Coseismic and post- Carrizo, D., 2012. Large extensional aftershocks in the continental forearc trig-
seismic slip associated with the 2010 Maule Earthquake, Chile: characteriz- gered by the 2010 Maule earthquake, Chile. Geophys. J. Int. 188, 879–890.
ing the Arauco Peninsula barrier effect. J. Geophys. Res. 118 (6), 3142–3159. http://dx.doi.org/10.1111/j.1365-246X.2011.05321.x.
http://dx.doi.org/10.1002/jgrb.50207. Saffer, D.M., Marone, C., 2003. Comparison of smectite- and illite-rich gouge fric-
Lucassen, F., Trumbull, R., Franz, G., Creixell, C., Vásquez, P., Romer, R.L., Figueroa, tional properties: application to the updip limit of the seismogenic zone along
O., 2004. Distinguishing crustal recycling and juvenile additions at active conti- subduction megathrusts. Earth Planet. Sci. Lett. 215, 219–235. http://dx.doi.org/
nental margins: the Paleozoic to recent compositional evolution of the Chilean 10.1016/S0012-821X(03)00424-2.
Pacific margin (36–41◦ S). J. South Am. Earth Sci. 17, 103–119. http://dx.doi.org/ SERNAGEOMIN, 2003. Mapa Geológico de Chile: versión digital. Servicio Nacional de
10.1016/j.jsames.2004.04.002. Geología y Minería, Santiago, Chile.
Madariaga, R., 1977. High-frequency radiation from crack (stress drop) models Sobiesiak, M., Meyer, U., Schmidt, S., Götze, H.J., Krawczyk, C.M., 2007. Asperity gen-
of earthquake faulting. Geophys. J. Int. 51, 625–651. http://dx.doi.org/10.1111/ erating upper crustal sources revealed by b value and isostatic residual anomaly
j.1365-246X.1977.tb04211.x. grids in the area of Antofagasta, Chile. J. Geophys. Res. 112, B12308. http://
Martin, M.W., Kato, T.T., Rodriguez, C., Godoy, E., Duhart, P., McDonough, M., Cam- dx.doi.org/10.1029/2006JB004796.
pos, A., 1999. Evolution of the late Paleozoic accretionary complex and overly- Song, T., Simons, M., 2003. Large trench-parallel gravity variations predict seis-
ing forearc-magmatic arc, south central Chile (38◦ –41◦ S): constraints for the mogenic behavior in subduction zones. Science 301, 630. http://dx.doi.org/
tectonic setting along the southwestern margin of Gondwana. Tectonics 18, 10.1126/science.1085557.
582–605. http://dx.doi.org/10.1029/1999TC900021. Spinelli, G.A., Saffer, D.M., Underwood, M.B., 2006. Hydrogeologic responses to
Melnick, D., Echtler, H.P., 2006. Morphotectonic and geologic digital map compila- three-dimensional temperature variability, Costa Rica subduction margin. J. Geo-
tions of the South-Central Andes (36◦ –42◦ S). In: The Andes. Springer, Berlin, phys. Res. 111, B04403. http://dx.doi.org/10.1029/2004JB003436.
Heidelberg, pp. 565–568. Tassara, A., 2010. Control of forearc density structure on megathrust shear
Métois, M., Socquet, A., Vigny, C., 2012. Interseismic coupling, segmentation and strength along the Chilean subduction zone. Tectonophysics 495, 34–47.
mechanical behavior of the central Chile subduction zone. J. Geophys. Res. 117, http://dx.doi.org/10.1016/j.tecto.2010.06.004.
B03406. http://dx.doi.org/10.1029/2011JB008736. Thurber, C.H., 1983. Earthquake locations and three-dimensional crustal structure
Moore, J.C., Saffer, D., 2001. Updip limit of the seismogenic zone beneath the ac- in the Coyote Lake area, central California. J. Geophys. Res. 88, 8226–8236.
cretionary prism of southwest Japan: an effect of diagenetic to low-grade meta- http://dx.doi.org/10.1029/JB088iB10p08226.
morphic processes and increasing effective stress. Geology 29, 183–186. http:// Thurber, C., Eberhart Phillips, D., 1999. Local earthquake tomography with
dx.doi.org/10.1130/0091-7613(2001)029<0183:ULOTSZ>2.0.CO;2. flexible gridding. Comput. Geosci. 25, 809–818. http://dx.doi.org/10.1016/
Moreno, M., Haberland, C., Oncken, O., Rietbrock, A., Angiboust, S., Heidbach, O., S0098-3004(99)00007-2.
2014. Locking of the Chile subduction zone controlled by fluid pressure be- Toomey, D.R., Foulger, G.R., 1989. Tomographic inversion of local earthquake data
fore the 2010 earthquake. Nat. Geosci. 7, 292–296. http://dx.doi.org/10.1038/ from the Hengill-Grensdalur Central Volcano Complex, Iceland. J. Geophys.
ngeo2102. Res. 94, 17497–17510. http://dx.doi.org/10.1029/JB094iB12p17497.
Moreno, M., Melnick, D., Rosenau, M., Baez, J., Klotz, J., Oncken, O., Tassara, A., Chen, Tsuji, T., Tokuyama, H., Costa Pisani, P., Moore, G., 2008. Effective stress and pore
J., Bataille, K., Bevis, M., Socquet, A., Bolte, J., Vigny, C., Brooks, B., Ryder, I., pressure in the Nankai accretionary prism off the Muroto Peninsula, southwest-
Grund, V., Smalley, B., Carrizo, D., Bartsch, M., Hase, H., 2012. Toward under- ern Japan. J. Geophys. Res. 113, B11401. http://dx.doi.org/10.1029/2007JB005002.
standing tectonic control on the M w 8.8 2010 Maule Chile earthquake. Earth Vásquez, P., Franz, G., 2008. The Triassic Cobquecura Pluton (Central Chile): an ex-
Planet. Sci. Lett. 321–322, 152–165. http://dx.doi.org/10.1016/j.epsl.2012.01.006. ample of a fayalite-bearing A-type intrusive massif at a continental margin.
Moreno, M., Rosenau, M., Oncken, O., 2010. 2010 Maule earthquake slip corre- Tectonophysics 459, 66–84. http://dx.doi.org/10.1016/j.tecto.2007.11.067.
lates with pre-seismic locking of Andean subduction zone. Nature 467, 198–202. Vásquez, P., Glodny, J., Franz, G., Frei, D., Romer, R.L., 2011. Early Mesozoic plutonism
http://dx.doi.org/10.1038/nature09349. of the Cordillera de la Costa (34◦ –37◦ S), Chile: constraints on the onset of the
Moscoso, E., Grevemeyer, I., Contreras-Reyes, E., Flueh, E.R., Dzierma, Y., Rabbel, W., Andean Orogeny. J. Geol. 119, 159–184. http://dx.doi.org/10.1086/658296.
Thorwart, M., 2011. Revealing the deep structure and rupture plane of the 2010 Völker, D., Grevemeyer, I., Stipp, M., Wang, K., He, J., 2011. Thermal control of
Maule, Chile earthquake (Mw = 8.8) using wide angle seismic data. Earth Planet. the seismogenic zone of southern central Chile. J. Geophys. Res. 116, B10305.
Sci. Lett. 307, 147–155. http://dx.doi.org/10.1016/j.epsl.2011.04.025. http://dx.doi.org/10.1029/2011JB008247.