LETTERS

PUBLISHED ONLINE: 30 JANUARY 2011 | DOI: 10.1038/NGEO1073

Limited overlap between the seismic gap and coseismic slip of the great 2010 Chile earthquake
S. Lorito1 *, F. Romano1 , S. Atzori1 , X. Tong2 , A. Avallone1 , J. McCloskey3 , M. Cocco1 , E. Boschi1 and A. Piatanesi1
The Mw 8.8 mega-thrust earthquake and tsunami that occurred on 27 February 2010 offshore the Maule region, Chile, was not unexpected. A clearly identied seismic gap113 existed in an area where tectonic loading has been accumulating since the great 1835 earthquake14 . Here we jointly invert tsunami and geodetic data to derive a robust model for the coseismic slip distribution and induced coseismic stress changes. We compare these with past earthquakes and the preseismic locking distribution13 , to assess if the Maule earthquake has lled the seismic gap. We nd that the main slip patch is located to the north of the gap, overlapping the rupture zone of the Mw 8.0 earthquake that occurred in 1928, with a secondary concentration of slip to the south. The seismic gap was only partially lled and a zone of high preseismic locking remains unbroken, inconsistent with the assumption that distributions of seismic rupture might be correlated with preseismic locking. Moreover, we conclude that increased stress on the unbroken patch may in turn have increased the probability of another major to great earthquake there in the near future. On 27 February 2010 at 06:34:14 utc an earthquake of moment magnitude Mw 8.8 occurred offshore the Maule region, Chile, some 360 km southwest of Santiago, with epicentre at 36.12 S, 72.90 W (http://earthquake.usgs.gov/earthquakes/eqarchives/epic/). The ensuing tsunami caused severe damage along the adjacent coasts, with reported maximum wave heights of more than 10 m at Constitucin (http://www.ngdc.noaa.gov/hazard/tsu.shtml), and a devastating inundation in Juan Fernndez islands, some 600 km offshore. The earthquake and tsunami claimed together more than 500 lives. The 2010 Maule earthquake ruptured the rapidly converging (6.8 cm yr1 ; ref. 10) interface between the Nazca and South American plates and was the largest mega-thrust earthquake recorded along this segment since the Mw 9.5 1960 Chile shock (Fig. 1). The source zone of the Maule earthquake features heterogeneous but locally high plate coupling13 . Tectonic stress has probably accumulated there continuously since the most recent major (Mw 8.5; ref. 4) subduction earthquake, experienced and described by Darwin during the voyage of the Beagle in 1835 (ref. 14), storing more than 12 m of slip where the interface is fully locked. The more recent Mw 7.9 1939 earthquake was in fact an intra-plate event14 , and the segment between the Mw 9.5 1960 earthquake to the south, and the Mw 8.0 (ref. 4) 1928 earthquake to the north, then contains a well defined seismic gap113 , referred to here as the Darwin gap. The 2010 Maule rupture nucleated close to the coast, producing a significant displacement both on the sea floor and onshore.
32

1985
0.25

8,000 7,000 6,000 5,000 4,000 3,000 2,000

0.75

Santiago

34
0.85

yr 6.8 cm

1928
Constitucion
`

Latitude ( S)

0.75

Altitude (m)

36

1939

1,000 0 1,000 2,000 3,000

1835

0.85 Concepcion
`

0.50

38

1960

Mw 8.8 27 Feb 2010

4,000 5,000 6,000 7,000 8,000

0 40 76 74 72 Longitude (W)

100 (km) 70

Figure 1 | Location map of the 2010 Maule earthquake and the seismic gap. Geographic location of the 27 February 2010 Mw 8.8 Maule earthquake, with epicentre (red star) between Concepcin and Constitucin in South Central Chile. The green and white beach ball is the United States Geological Survey centroid moment tensor. Yellow stars are the epicentres of 1928, 1939, 1960 and 1985 earthquakes, with their approximate source zones4,7,16,19 , which are shaded for thrust inter-plate events. The 1960 source zone includes both 21 May Mw 7.9 and 22 May Mw 9.5 earthquakes19 . Orange lines are contours of the preseismic locking distribution13 . The segment that probably contains the source zone of the 1835 earthquake (indicated by the black dashed line with arrows) is the zone of the Darwin14 gap, where a major earthquake was expected12 . The red line with triangles is the trench between the Nazca and South America Plates26 . The white arrow indicates the approximate convergence direction of the plates along with its estimated velocity10 .

Tsunami observations provide an indirect measure of the seafloor displacement, whereas geodetic observations (Interferometric Synthetic Aperture Radar (InSAR), Global Positioning System

1 Istituto

Nazionale di Geosica e Vulcanologia, Via di Vigna Murata 605, 00143 Rome, Italy, 2 Institute of Geophysics and Planetary Physics, Scripps Institution of Oceanography, University of California, San Diego, La Jolla, California 92093-0225, USA, 3 Environmental Sciences Research Institute, School of Environmental Sciences, University of Ulster, Coleraine BT52 1SA, Northern Ireland. *e-mail: stefano.lorito@ingv.it.
NATURE GEOSCIENCE | ADVANCE ONLINE PUBLICATION | www.nature.com/naturegeoscience

2011 Macmillan Publishers Limited. All rights reserved.

LETTERS
32

NATURE GEOSCIENCE DOI: 10.1038/NGEO1073

are mapped in Supplementary Fig. S1, values in Supplementary Table S1). The major slip patch is north of the epicentre between 36 and 34 S, with a longitudinally elongated area of some 200 80 km, and a slip concentration centred on 3535.5 S with peaks of 1819 m. The highest slip is confined to about 60110 km down-dip from the trench, corresponding to a depth of 2540 km. The southern slip patch, which measures some 130 60 km, is centred on about 37.5 S and is characterized by lower overall slip, up to about 910 m at a similar depth to the northern patch. The slip direction is grossly consistent with plate convergence, with a dominant thrust and minor right-lateral components. Our slip model corresponds to a seismic moment of 1.55 1022 N m and magnitude Mw = 8.8 using a rigidity value of 30 GPa. The combined tsunami and geodetic data set (Fig. 3) provided good control on the slip distribution (Supplementary Fig. S2). Both amplitudes and timing of tsunami data are well predicted by the retrieved model, InSAR residuals are within 10% over most of the covered area (see also Supplementary Fig. S3) and GPS vectors are well reproduced, particularly in the epicentral region. Land-level change data are predicted within errors almost everywhere as well as the hinge-line for the coseismic displacement (Fig. 3b,c; see also Fig. 2), which should run approximately along the coastline15 . Two other published models share first-order features with ours16,17 , and a third shares the main slip concentration north of the epicentre18 . A comparison between their resolution analyses16,17 and ours (Supplementary Fig. S2) shows that by including tsunami and further geodetic data we achieve better resolution. We moreover use variable strike and dip over the fault, which better captures the geometrical complexity of the plate interface19 . A detailed comparison with all of the published and preliminary online slip models is reported in Supplementary Information. In Fig. 4a we compare our slip model with past earthquakes and the preseismic locking distribution13 . The gap zone roughly corresponds to the 1835 source zone, where the locking is very high (locally >85%), and no significant earthquakes have occurred since then. The rupture was bilateral and propagated north and south from the hypocentre at an average velocity of 22.6 km s1 (refs 16, 18; fixed at 2.25 km s1 in our model). The Maule earthquake nucleated where locking is >75%, and propagated principally to the north, releasing the maximum slip where the 1928 event had partially relaxed the fault, and where the locking was lower around 35 S. The rupture then propagated into a zone of higher locking farther north and stopped near the southern termination of the 1985 event, which could be expected to have low prestress. The southward-propagating rupture, conversely, initiated with lower slip in a relatively weakly locked zone (50%75%) before releasing more slip in a strongly locked area (>75%) around 37 S, and overlapping the 1960 source zone. The northern section of the 1960 source region is known to have experienced relatively small slip and the termination of the Maule rupture coincides reasonably well with the northern 1960 asperity at about 38.5 S (ref. 19). The slip distribution of the Maule earthquake is then not consistent with simple earthquake recurrence models and our results contradict the recent work, on the basis of preliminary slip distributions, that the slip distribution and preseismic locking are strongly correlated for this event13 . These preliminary slip models, however, were based on teleseismic data that, if used alone, seem to have poor resolution on the slip-distribution details16 . Further studies, including joint inversions of seismological, tsunami and geodetic data, might be necessary for better understanding differences between our and published seismological models18 . Similarly complex relationships between locking and slip can be nevertheless observed in other, well instrumented, recent subduction earthquakes20 . It is then not sufficient to consider previous earthquakes, plate interface coupling and approximately uniform and constant loading with strain accumulated in the

1985

34

6.8 cm

yr

1928

Latitude ( S)

36

38 1960

1835

0 40 76 74 Longitude ( W) 72

(km)

100 70

10 Slip (m)

15

20

Figure 2 | Slip distribution of the 2010 Maule earthquake. Slip distribution for the 2010 Mw 8.8 Maule earthquake obtained from the joint inversion of tsunami and geodetic data, represented by colours according to the scale at the bottom. White arrows represent the slip direction (rake). Thin black contours indicate the associated surface vertical displacement (1-m-interval solid lines for uplift, 20-cm-interval dashed lines for subsidence). Epicentres and source zones are plotted only for major thrust earthquakes (compare Fig. 1).

(GPS) and land-level changes15 ) enable the estimation of the onshore displacement field. The two data sets possess almost complementary resolving power on the coseismic deformation, from which the characteristics of the earthquake rupture can be inferred. Here we carry out a joint inversion of these data to derive a robust model for the slip distribution, and the subsequent Coulomb stress change over the fault plane. It is usually assumed that high prestress coincides with maxima of locking and elapsed time since the most recent failure. It is also sometimes proposed that the slip distribution of a future earthquake might be correlated with the distribution of prestress13 . We then compare coseismic slip and stress change during the Maule earthquake with the regional prestress distribution as inferred by preseismic locking and slip in past earthquakes. Our purpose is twofold. We assess (1) the extent to which the slip distribution matches the inferred prestress distribution, to test the fundamental hypothesis that the two might be correlated, and (2) the extent to which the Maule event closed the Darwin gap, reducing the seismic hazard in the region. The slip distribution model obtained with the joint inversion of tsunami and geodetic data is shown in Fig. 2 (slip uncertainties
2

NATURE GEOSCIENCE | ADVANCE ONLINE PUBLICATION | www.nature.com/naturegeoscience 2011 Macmillan Publishers Limited. All rights reserved.

NATURE GEOSCIENCE DOI: 10.1038/NGEO1073

a 1.6
ANC 0.9 ARI 0.9 CAL 1.9 COR 0.3 IQU 2.0 TAL 1.3 WAI 0.9 OWE 0.3 D51406 0.05 D54401 0.4 ANT 0.5 CLD 1.3 COQ 0.3

LETTERS
b
32
Obs Pred 20 cm 1m

34

Latitude ( S)

D32412 0.6 SFE 1.7

36

VAL 0.3 RIK 0.2 EAS 0.04 D51426

38

Observed Predicted 40

100 (km) 70

60 min

74 72 Longitude ( W) Residuals

c
Observed Predicted

d
32

34

Ascending

100 90 80

36

38

70
0 100

Relative error (%)

Latitude ( S)

40 32

(km)

60 50

Descending

40 30 20 10 0
0 100

34

36

38
(km)

1 M

40

74 72 70 Longitude ( W)

Figure 3 | Comparison between observed and predicted data sets. a, Observed and predicted tsunami time series. Peak value (m) is indicated for each station. Abbreviated station names are as in Supplementary Table S2. b, Observed and predicted GPS vectors. Contour lines of predicted vertical displacement as in Fig. 2. Yellow squares indicate positions of land-level-change measurements. c, Observed and predicted land-level changes plotted against latitude scale of b. Error bars for observed data are experimental uncertainties15 , whereas for predicted data they are calculated adding 1 errors (Supplementary Table S1) to the average slip model. d, Residuals between observed and predicted InSAR LOS displacement expressed as percentages of the observed data values.

interseismic period on strongly coupled asperities and released during the earthquake, generating the regions of highest slip. The lobe of strong coupling, which extends from 38.5 S to about 36 S, remained largely unbroken despite its location in the centre of the Darwin gap, whereas the area that experienced the maximum slip of nearly 20 m in the 2010 event overlaps the source region of the 1928 earthquake, although it could only have accumulated some 56 m of slip since then. It is true that most of the regions of high slip in Fig. 4a correspond to regions of relatively high (>50%) coupling, which probably controls the long-term distribution of seismic moment release; however, coupling by

itself provides a poor forecast of the slip and in this case would have even failed to forecast the location experiencing the highest slip. These results suggest that strong coupling is a necessary but not sufficient condition for high slip in a single event. Slip must additionally be controlled by other factors such as friction and fault rheology, structural heterogeneity, and the history of slip on the fault, including aseismic slip, probably over the past several seismic cycles. The Coulomb stress distribution induced on the fault plane by this model of slip during the Maule earthquake is shown in Fig. 4b. The increase in Coulomb stress deeper than 45 km or
3

NATURE GEOSCIENCE | ADVANCE ONLINE PUBLICATION | www.nature.com/naturegeoscience

2011 Macmillan Publishers Limited. All rights reserved.

LETTERS
a
32 1985 34 1928 34

NATURE GEOSCIENCE DOI: 10.1038/NGEO1073

b
32 1985

1928

Latitude ( S)

36

Latitude ( S)

38

1835

1960

38

183 5
1960

36

0 40 76 74 72

100 (km) 70

0 40 76 74 72

100 (km) 70

Longitude ( W) 0 5 10 Slip (m) 15 20

Longitude ( W) 5 4 3 2 1 0 1 2 3 Coulomb stress change (MPa)

Figure 4 | Comparison of the Maule earthquake slip distribution and coseismic stress variation to preseismic locking and past earthquakes. a, Slip distribution for the Maule earthquake compared with the estimated position of the Darwin gap, in the segment where the 1835 earthquake probably occurred. Source zones of past thrust earthquakes as in Fig. 1. White lines are preseismic locking contours as in Fig. 1. The Darwin gap was only partially lled and a zone of high preseismic locking remains unbroken. b, Coulomb stress changes associated with the Maule earthquake, resolved on the mega-thrust surface. Positive stress changes may favour a future rupture. An increase of stress occurring after the Maule earthquake in the Darwin gap might have increased the probability of a future earthquake there.

shallower than about 15 km in Fig. 4b is unlikely to have strong consequences for seismic and tsunami hazard. The deeper region might be beyond the seismogenic zone limit17 , and large shallow slow earthquakes are unlikely at accretionary margins21 . It might be expected that some afterslip has occurred in both regions but this has not been detected so far. Ongoing geodetic studies will help to address this issue. However, our model shows that not only did the Darwin gap not rupture completely, but the Maule earthquake also produced a strong increase in stress on the highly coupled lobe extending northward from about 37 S to 36 S. Such stress increases have been frequently shown to trigger further large earthquakes22 . In this case, failure of the unbroken areas of the gap, if they are highly coupled, might produce an event of magnitude Mw 7.58. In summary, the Maule earthquake was not the characteristic earthquake that was expected to close the Darwin gap. Whilst the rupture does seem to terminate against the source regions of the 1960 and 1985 earthquakes and the high slip is broadly confined to areas of relatively high coupling, zones of very high coupling in the Darwin gap remain unbroken. Moreover, the highest slip occurs in an area that failed in 1928 that is coupled only at the 5075% level. It is then hard to envisage that preseismic locking might be used for anticipating future slip in seismic gaps. The area of strong plate locking in the Darwin gap that did not fail in this event experienced some 2 MPa of further loading by it. Rather than relaxing accumulated stress in the Darwin gap, reducing near-future seismic hazard there, this strong stress interaction might have increased the probability of another major to great earthquake in the Darwin gap in the near future.

Methods
We use sea-level recordings at 15 coastal tide-gauges around the Pacific Ocean, and at four deep-sea bottom pressure gauges in the open sea (Supplementary Fig. S4). Tide-gauge data are from Intergovernmental Oceanographic Commission/United Nations Educational, Scientific and Cultural Organization (http://www.ioc-sealevelmonitoring.org/) and University
4

of Hawaii (http://ilikai.soest.hawaii.edu/uhslc/background.html) sea-level data centres. Data providers for each station are listed on these websites, and in Supplementary Table S2. Deep-ocean Assessment and Reporting of Tsunami sensor data are from the National Oceanic and Atmospheric Administration data centre (http://www.ndbc.noaa.gov/). We use 25 stations data from continuous GPS sites managed in South America by the International Global Navigation Satellite Systems Service (http://igscb.jpl.nasa.gov/) and three stations data distributed by the French Centre National de la Recherche ScientifiqueInstitut National des Sciences de lUnivers (INSU, France) through the GPSCOPE portal (https://gpscope.dt.insu.cnrs.fr/), hosted at the Division TechniqueINSU. Using GIPSY-OASIS II v.5.0 software23 , we applied a robust procedure24 to analyse GPS data, consisting of a precise-point-positioning strategy followed by network ambiguity resolution, and, then, by alignment of the daily solutions to ITRF 2005 (ref. 25). For each site, static coseismic offsets are estimated by comparing the average position calculated in the 78 days before the earthquake with the position obtained by the daily solution of the day of the earthquake. The far-field stations are used to define a stable reference frame. In the inversion, we use only the six GPS data closest to the source (Supplementary Table S3). InSAR enables imaging of the relative change of the ground deformation with high resolution and wide coverage soon after the event. For repeat-pass interferometry shorter time span between the satellite acquisitions is preferred to isolate the coseismic signal from the preseismic and postseismic deformation. In the present case the acquisition date ranges from May 2007 to February 2010 before the earthquake, and March 2010 to May 2010 after the earthquake17 . We use ascending and descending L-band images acquired by ALOS PALSAR, a satellite of Japan Aerospace Exploration Agency (http://www.jaxa.jp/index_e.html), and this provides two Line-Of-Sight (LOS) components of the three-dimensional displacement field (Supplementary Fig. S3). We calibrated InSAR LOS data by means of the GPS data set described above (see also http://supersites.unavco.org/chile.php and ftp://topex.ucsd.edu/pub/chile_eq). We use 34 land-level-change data observed mostly at coastal sites and some at estuarine valleys15 . These data are an estimate, at several sites, of the coseismic displacement produced by the 27 February 2010 earthquake, spanning almost the whole source length (Fig. 3c). Our fault model has variable dip and strike. We sample the cross-section of the subduction zone geometry analysis carried out by Gavin Hayes at the United States Geological Survey (http://earthquake.usgs.gov/earthquakes/eqarchives/subduction_ zone/us2010tfan/), and then develop it along strike, using trench coordinates provided by Peter Bird26 . The fault plane is divided into 200 subfaults of 25 25 km (Supplementary Fig. S5 and Table S1).

NATURE GEOSCIENCE | ADVANCE ONLINE PUBLICATION | www.nature.com/naturegeoscience 2011 Macmillan Publishers Limited. All rights reserved.

NATURE GEOSCIENCE DOI: 10.1038/NGEO1073

The vertical displacement associated with each of the subfaults is calculated with Okadas formulas27 , and transferred to the sea surface as the initial condition for tsunami propagation. The propagation is calculated with the Cornell Multi-Grid Coupled Tsunami Model code (http://ceeserver.cee.cornell.edu/pll-group/comcot.htm), based on shallow water equations in spherical coordinates. We use total reflecting boundaries at the coastlines and open boundaries elsewhere. Linear equations are solved for the propagation in the open sea, using a grid spacing of 2 arcmin. For coastal propagation we use nonlinear equations, and finer nested grids with 30 arcsec spacing. We use the SRTM30_PLUS (http://topex.ucsd.edu/WWW_html/srtm30_plus.html) bathymetric model. Okadas formulas are also used for calculating subfault contributions at GPS, land-level data sites and InSAR, for a set of points subsampled from the ALOS PALSAR interferograms. A variable LOS based on the satellite state vectors is used for the projection into the ground-satellite direction. The inversion method28 is based on a global search technique, simulated annealing and the linear superposition of Greens functions. We add smoothing and seismic-moment-minimization constraints to the slip distribution. The rupture front is assumed to be circular and propagating at constant speed, fixed at 2.25 km s1 (ref. 18), as we had poor resolution on this parameter. The slip vector (slip, rake) for each subfault is the observable retrieved with the inversion (Fig. 2). The rake is kept constant over three large blocks (whose centres are the starting points of the white arrows in Fig. 2). We also make checkerboard tests for assessing the resolution of the single data sets and of both data sets together in the joint inversion (Supplementary Fig. S2). Coulomb stress values are computed by means of COULOMB 3.1 (refs 29,30). Coulomb stress change is defined as CFF = + , where is the change in shear stress on the failure plane (positive in the slip direction), is the change in normal stress (positive when the fault is unclamped) and is the apparent friction coefficient. The friction coefficient used in this study is = 0.4.

LETTERS
14. Darwin, C. Journal of the Researches into the Natural History and Geology of the Countries Visited During the Voyage of the HMS Beagle Round the World 2nd edn (John Murray, 1845). 15. Faras, M. et al. Land-level changes produced by the Mw 8.8 2010 Chilean earthquake. Science 329, 916 (2010). 16. Delouis, B., Nocquet, J-M. & Valle, M. Slip distribution of the February 27, 2010 Mw = 8.8 Maule earthquake, central Chile, from static and high-rate GPS, InSAR, and broadband teleseismic data. Geophys. Res. Lett. 37, L17305 (2010). 17. Tong, X. et al. The 2010 Maule, Chile earthquake: Downdip rupture limit revealed by space geodesy. Geophys. Res. Lett. 37, L24311 (2010). 18. Lay, T. et al. Teleseismic inversion for rupture process of the 27 February 2010 Chile (Mw 8.8) earthquake. Geophys. Res. Lett. 37, L13301 (2010). 19. Moreno, M. S., Bolte, J., Klotz, J. & Melnick, D. Impact of megathrust geometry on inversion of coseismic slip from geodetic data: Application to the 1960 Chile earthquake. Geophys. Res. Lett. 36, L16310 (2009). 20. Konca, O. et al. Partial rupture of a locked patch of the Sumatra megathrust during the 2007 earthquake sequence. Nature 456, 631635 (2008). 21. Bilek, S. L. The role of subduction erosion on seismicity. Geology 38, 479480 (2010). 22. Nalbant, S. S., Steacy, S., Sieh, K., Natawidjaja, D. & McCloskey, J. Earthquake risk on the Sunda trench. Nature 435, 756757 (2005). 23. Lichten, S. & Borders, J. Strategies for high-precision Global Positioning System orbit determination. J. Geophys. Res. 92, 1275112762 (1987). 24. DAgostino, N. et al. Active tectonics of the adriatic region from GPS and earthquake slip vectors. J. Geophys. Res. 113, B12413 (2008). 25. Altamimi, Z., Collilieux, X., Legrand, J., Garayt, B. & Boucher, C. ITRF2005: A new release of the international terrestrial reference frame based on time series of station positions and earth orientation parameters. J. Geophys. Res. 112, B09401 (2007). 26. Bird, P. An updated digital model of plate boundaries. Geochem. Geophys. Geosyst. 4, 1027 (2003). 27. Okada, Y. Surface deformation due to shear and tensile faults in a half-space. Bull. Seismol. Soc. Am. 75, 11351154 (1985). 28. Lorito, S., Piatanesi, A., Cannelli, V., Romano, F. & Melini, D. Kinematics and source zone properties of the 2004 SumatraAndaman earthquake and tsunami: Nonlinear joint inversion of tide gauge, satellite altimetry, and GPS data. J. Geophys. Res. 115, B02304 (2010). 29. Lin, J. & Stein, R. S. Stress triggering in thrust and subduction earthquakes, and stress interaction between the southern San Andreas and nearby thrust and strike-slip faults. J. Geophys. Res. 109, B02303 (2004). 30. Toda, S., Stein, R. S., Richards-Dinger, K. & Bozkurt, S. Forecasting the evolution of seismicity in southern California: Animations built on earthquake stress transfer. J. Geophys. Res. 110, B05S16 (2005).

Received 27 July 2010; accepted 31 December 2010; published online 30 January 2011

References
1. Barrientos, S. Is the PichilemuTalcahuano (Chile) a seismic gap? Seismol. Res. Lett. 61, 43 (1990). 2. Campos, J. & Kausel, E. The large 1939 intraplate earthquake of Southern Chile. Seismol. Res. Lett. 61, 43 (1990). 3. Madariaga, R. La Seismicidad de Chile, Fisica de la Tierra, Vol. 10 221258 (Ediciones de la Universidad Complutense de Madrid, 1998). 4. Beck, S., Barrientos, S., Kausel, E. & Reyes, M. Source characteristics of historic earthquakes along the central Chile subduction zone. J. South Am. Earth Sci. 11, 115129 (1998). 5. Klotz, J. et al. Earthquake cycle dominates contemporary crustal deformation in Central and Southern Andes. Earth Planet. Sci. Lett. 193, 437446 (2001). 6. Ruegg, J. C. et al. Interseismic strain accumulation in south central Chile from GPS measurements, 19961999. Geophys. Res. Lett. 29, 15171520 (2002). 7. Campos, J. et al. A seismological study of the 1835 seismic gap in South Central Chile. Phys. Earth Planet. Iner. 132, 177195 (2002). 8. Brooks, B. A. et al. Crustal motion in the Southern Andes (26 36 S): Do the Andes behave like a microplate? Geochem. Geophys. Geosyst. 4, 1085 (2003). 9. Moreno, M. S., Klotz, J., Melnick, D, Echtler, H. & Bataille, K. Active faulting and heterogeneous deformation across a megathrust segment boundary from GPS data, south central Chile (3639 S). Geochem. Geophys. Geosyst. 9, Q12024 (2008). 10. Vigny, C. et al. Upper plate deformation measured by GPS in the Coquimbo Gap, Chile. Phys. Earth Planet. Iner. 175, 8695 (2009). 11. Ruegg, J. C. et al. Interseismic strain accumulation measured by GPS in the seismic gap between Constitucin and Concepcin in Chile. Phys. Earth Planet. Iner. 175, 7885 (2009). 12. Madariaga, R., Mtois, M., Vigny, C. & Campos, J. Central chile finally breaks. Science 328, 181182 (2010). 13. Moreno, M., Rosenau, M. & Oncken, O. 2010 Maule earthquake slip correlates with pre-seismic locking of Andean subduction zone. Nature 467, 198204 (2010).

Acknowledgements
We acknowledge discussions with our colleagues E. Tinti and A. Herrero about seismic rupture properties, and with S. Nalbant about coseismic stress. We also acknowledge C. Vigny, leader for the acquisition of the GPSCOPE GPS data in Chile. We thank N. DAgostino and E. DAnastasio, who set up and implemented the GPS data processing strategy. We appreciate the effort of our colleagues at Cornell University who developed the tsunami-modelling package. We moreover wish to thank all of the data providers who made this study possible. Some figures were drawn with generic mapping tools (http://gmt.soest.hawaii.edu/). J.M. acknowledges support from the UK NERC under grant numbers NE/F01161X/1 and NE/H008519/1.

Author contributions
S.L., F.R. and A.P. were involved in all of the phases of this study. S.A., X.T. and A.A. processed, modelled and analysed geodetic data, and wrote part of the Methods. J.M. and M.C. contributed to result interpretation and paper writing. E.B. promoted the experiment, contributed to result interpretation and supported the project.

Additional information
The authors declare no competing financial interests. Supplementary information accompanies this paper on www.nature.com/naturegeoscience. Reprints and permissions information is available online at http://npg.nature.com/reprintsandpermissions. Correspondence and requests for materials should be addressed to S.L.

NATURE GEOSCIENCE | ADVANCE ONLINE PUBLICATION | www.nature.com/naturegeoscience

2011 Macmillan Publishers Limited. All rights reserved.