A Lacuna Sísmica Do Norte Do Chile: Estado Da Arte

! ! ! ! "#$%&'!&'#!()*+')*#!&*!(*)),!&*! ! -'.$*#!/0,)'#1!-21!&*!34536! ! ! ! ! ! ! !

! ! ! ! ! 7.89*)#8&,&*!&*!:;'!<,%0'!=7:<>!!*!!7.89*)#8&,&*!&*!?),#@08,!=7.?>! ! ! ! ! Centro de Sismologia da USP IAG / IEE !!!!/*.$)'!&*!:8#+'0'A8,!&,!7:! !

! :B:D7.?E!FG#*)9,$H)8'!:8#+'0HA8I'!&,!7.?! ! ! A lacunaJ*0,$H)8'!!&*!! sísmica55K4LK345L do! Norte do Chile: estado da arte ! "!

Hans Agurto-Detzel

São Paulo, 15-Setembro-2014 B05310 SCHOLZ AND CAMPOS: SEISMIC COUPLING B05310

Figure 4. The earthquake history and coupling of Chile. See text for sources of couplingScholz and Campos, 2012. data. JGR

Ward [1990] as far as 400 km downdip. Because we study concluded it was fully coupled on a 20 dipping surface only the seismogenic portion of the interface, for our cal- down to 35 km depth, below which the coupling linearly culations we use the geodetic estimates of the moment. tapered to zero at 55 km. By the same reasoning as discussed Lorenzo-Martin et al. [2006] found, from GPS measure- in the Cascadia case, we take the fully locked region to be ments, that c = 0.96 for A = 110 109m2 and an assumed the seismogenic area. This boundary seems to have been G c Â subduction velocity of 69 mm/yr, yielding P_ G = 7.3 F. defined by the Mw 7.7 Tocopilla earthquake of 2007, in [65] The only known preceding earthquake of size similar which its region of maximum slip was concentrated around to the 1960 event was one that occurred in 1575 [Cisternas 35 km, with some slip, possibly post-seismic, extending to 50 km depth [Bejar-Pizarro et al., 2010]. Hence with this et al., 2005]. Using that 385 yr recurrence time yields P_ S = assumption c = 1 and with v = 79 mm/yr on a 102 450 6.2 F, for a cS = 0.8. If we include in that period the 1837 G p Â earthquake, estimated to have been about half the size of km surface, P_ G = 3.6 F. Vargas et al. [2005] identified sec- tor-rupturing earthquakes prior to 1877 in 1768, 1543, and 1960 [Abe, 1979], we get P_ S = 9.3 F, larger than P_ G (this is an overestimate because a goodly portion of the 1837 earth- 1430, which indicate an average T of 149 years [Comte and quake occurred in the Central Chile segment). The tsunami Pardo, 1991; Nishenko, 1985]. Assuming Mw 8.8 for these record of this region indicates great tsunami-generating earthquakes and this value of T yields P_ S = 3.4 F and cS = events every 285 years on average [Cisternas et al., 2005]. 0.94. This consists of the less frequent 1960-type events that rupture the entire 850 km length of this segment and smaller 3.16. Central and Southern Peru earthquakes like that of 1837 that rupture portions of it. [67] The earthquake history and coupling of this region are [66] Northern Chile: This region, from 18 to 23S, is a shown in Figure 5 (modified after Dorbath et al. [1990]). 450 km long seismic gap that last ruptured in the Iquique The coupling of the northern two-thirds of this region was earthquake of 1877, with a magnitude estimated at Mw determined by Perfettini et al. [2010]. They found three 8.8–9.0 [Abe, 1979; Bejar-Pizarro et al., 2010; Comte and zones of high coupling, shown in Figure 5. A narrow zone of Pardo, 1991; Kausel and Campos, 1992]. Chlieh et al. low coupling separates the northern two. The southern two [2004] studied this area with GPS and INSAR and are separated by a much wider zone of low coupling that

12 of 22 LETTER doi:10.1038/nature13681

Gradual unlocking of plate boundary controlled initiation of the 2014 Iquique earthquake

Bernd Schurr1, Gu¨nter Asch1, Sebastian Hainzl1, Jonathan Bedford1, Andreas Hoechner1, Mauro Palo1, Rongjiang Wang1, Marcos Moreno1, Mitja Bartsch1, Yong Zhang2, Onno Oncken1, Frederik Tilmann1, Torsten Dahm1, Pia Victor1, Sergio Barrientos3 & Jean-Pierre Vilotte4

On 1 April 2014,Northern Chile wasstruck by a magnitude 8.1earth- rupture area was less than 2.6 m (refs 6, 7) leaving most of the past slip quake following a protracted series of foreshocks. The Integrated Plate deficit of ,8–9 m accumulated since 1877 unaffected8. On 1 April 2014, Boundary Observatory Chile monitored the entire sequence of events, the Mw 8.1 Iquique earthquake north of Iquique struck the central por- providing unprecedented resolution of the build-up to the main event tion of the gap. Using seismological and geodetic observations we here and its rupture evolution. Here we show that the Iquique earthquake analyse the rupture, its relationship toprevious locking of the plate inter- broke a central fraction! of the so-called northern Chile seismic gap, face, and the pre-seismic transients leading up to the earthquake. the last major segment of the South American plate boundary that 1,2 2 M density had not ruptured in the past century . Since July 2013 three seismic 0 0 M 0 1 1868 2001 8 clusters, each lasting a few weeks, hit this part of the plate boundary .4

1906 with earthquakes of increasing peak magnitudes. Starting with the –18º

1948 second cluster, geodetic observations show surface displacements that Arica can be associated with slip on the plate interface. These seismic clus- 1956 tersand their slip transients occupied a part of the plate interface that 1987 –19º

was transitional between a fully locked and a creeping portion. Leading 1945 up to this earthquake, the b value of the foreshocks gradually decreased 1877 2014 04 01 M 8.1 Pisagua 1869 1911 during the years before the earthquake, reversing its trend a few days 1933 –20º before the Iquique earthquake. The mainshock finally nucleated at Iquique

the northern end of the foreshock area, which skirted a locked patch, 1940

2014 04 03 M 7.6 Se 1871 and ruptured mainly downdip towards higher locking. Peak slip was ismic gap –21° 1906 attained immediately downdip of theforeshockregionandatthemar- 1905 gin of the locked patch. We conclude that gradual weakening of the –1 1967 central part of the seismic gap accentuated by the foreshock activity Latitude 6.7 cm yr in a zone of intermediate seismic coupling was instrumental in caus- 2007 –22° 1928 ing final failure, distinguishing the Iquique earthquake from most 2007 great earthquakes. Finally, only one-third of the gap was broken and 5 M 7.7 –23° 4 Mejillones the remaining locked segments now pose a significant, increased seis- 1947 peninsula 1995 3 slip mic hazard with the potentialto host anearthquake with a magnitude (m) of .8.5. 1998 2 Antofagasta 1987 –24° 1995 The northern Chile–southern Peru seismic gap last broke in 1877 in 1 M 8.1 1 1942 BB+SM a great earthquake (Mw ,8.8) that ruptured from south of Arica to the 100 km SM 1936 1910 1947 Mejillones peninsula (see Fig. 1). The reported historical recurrence inter- 1960 GPS 50 cm 1988 val for the past 500 years in this region has been estimated at 111 6 33 1900 2000 –73° –72° –71° –70° –69° 1 years , making it probably the most mature seismic gap along the South Time (years) Longitude American plate boundary. In the past two decades the two adjoining segments south and north broke in the M 8.1 Antofagasta earthquake Figure 1 | Map of Northern Chile and Southern Peru showing historical w earthquakes and instrumentally recorded megathrust ruptures. IPOC of 1995 (refs 3, 4) and the Mw 8.4 Arequipa earthquake of 2001 in south- 5 instruments used in the present study (BB, broadband; SM, strong motion) are ern Peru . In the previous cycle the southern Peru and northern Chile shown as blue symbols. Left: historical1,2 and instrumental earthquake record. segments broke within few years (Fig. 1), suggesting that they might be Centre: rupture length was calculated using the regression suggested in ref. 28, 1 coupled in time . The imminence of a large megathrust event in this with grey lines for earthquakes M .7 and red lines for Mw .8. The slip region motivated the setting up of an international monitoring effort distribution of the 2014 Iquique event and its largest aftershock derived in this with the Integrated Plate Boundary Observatory Chile (IPOC). Having study are colour coded, with contour intervals of 0.5 m. The green and black started in 2007, there now exists an exceptional database that monitors vectors are the observed and modelled horizontal surface displacements of the 4,5,7 the gradual plate boundary failure with various geophysical techniques. mainshock. The slip areas of the most recent other large ruptures are also . plotted. Right: moment deficit per kilometre along strike left along the plate Several major earthquakes (Mw 7) have occurred in this gap since boundary after the Iquique event for moment accumulated since 1877, 1850 (Fig. 1); the largest until now was the Mw 7.7 Tocopilla earthquake assuming current locking (Fig. 3a). The total accumulated moment since 1877 in 2007, which broke the southern rim of this segment beneath and north from 17u Sto25u S (red solid line) is 8.97; the remaining moment after of Mejillones peninsula along a total length of 150 km. Only the down- subtracting all earthquake events with Mw .7 (grey dotted line) is 8.91 for the dip end of the locked zone slipped in this event, and the total slip in the entire northern Chile–southern Peru seismic gap. Lay et al. (2014) GRL 1GFZ Helmholtz Centre Potsdam, German Research Centre for Geosciences, Telegrafenberg, 14473 Potsdam, Germany. 2School of Earth andSchurr Space Sciences, et al. (2014) Nature Peking University, Beijing 100871, China. 3Centro Sismolo´gico National, Universidad de Chile, Facultad de Ciencias Fı´sicas y Matema´ticas, Blanco Encalada 2002, Santiago, Chile. 4Institut de Physique du Globe de Paris, 1, rue Jussieu, 75238 Paris cedex 05, France.

Waveform inversion is widely used for constructing seismic source models; –17º however, models may differ substantially for 1868 M 8.8 Lower plate Interplate Upper plate the sameSlip earthquake (m) [e.g., Beresnev, 2003; –19º ﬁ Raza ndrakoto8 and Mai, 2014]. Waveform Slip (m) inversion is especially problematic for large 8 earthquakes with a long source time function, because6 seismic data for such –18º 6 2001 M 8.4 events are contaminated by various later phases, which are difﬁcult to calculate accurately because4 of the limited accuracy 4 of the Green’s function. To cope with this problem, we applied the newly developed –19º inversion method2 of Yagi and Fukahata 2 [2011a], in which the uncertainty of the Green’s function is taken into account. One 0 of the clear advantages0 of this method is that the effect of correlated modeling errors M 6.7 is naturally mitigated, therefore making –20º –20º possible to discuss the detailed rupture process, including the small initial rupture phase [Yagi and Fukahata, 2011b]. According to this method, the smoothness of slip distribution is objectively determined from the observed data, based on Akaike’s –21º Bayesian Information Criterion (ABIC) 1877 M 8.8[Akaike, 1980; Yabuki and Matsu’ura,1992] seismic gap seismic gap and the non-negative constraint for slip is

Northern Chile 1967 M 7.4 not needed. We calculated the theoretical Green’s –22º function using the method of Kikuchi and Kanamori [1991] with a sampling rate of 0.1 s. To explain the teleseismic waveforms, we slightly modified the structure model of –21º –71º –70º 2007 M 7.7Delouis et al. [2009]. We assumed that the faulting occurred on a single plane and –23º adopted a hypocenter (19.572S,Figure 70.908 2 W;| Source processes of events in the March–April 2014 Iquique depth = 22 km) and fault mechanismearthquake (strike sequence. RMTs of relocated earthquakes in this sequence are 350°, dip 15°) that were slightlyshown modifi anded coloured by their location with respect to the slab interface; those 1995 M 8.1 from the CSN hypocenter andinterpreted the Global as upper plate events are green, lower plate earthquakes are blue, Centroid-Moment-Tensor (GCMT) solution Figure 1. Total slip distribution, aftershock distribution, and moment- [http://www.globalcmt.org; lastand access interplate on events are red. Earthquakes are overlain on the preferred –24º rate function. (a) Map view of inverted total slip distribution of the 2014 fault-slip models for the M 8.2 and M 7.7 events (hypocentres are shown with –73ºoff Iquique earthquake. –72º Large and –71ºsmall stars indicate the –70º epicenter of –69º18 April 2014], respectively. –68º The fault area the main shock and the largest aftershock, respectively. Also shown are used for the source process inversionstars), with was 2-m contour intervals. Dot–dashed lines in the background are slab Figure 1the| Tectonic focal mechanism setting of main shock of 2014 determined Iquique in this study earthquake and the first sequence.taken as 225 kmRupture× 195 km, whichcontours was 13, plotted every 10 km. areas of large2 days aftershocks historical (black earthquakes circles), determined are by indicated the Centro Sismológico by grey (modelled)expanded into and bilinear white B-splines with an Nacional (CSN), the University of Chile. The seismic source area of 1877 off interval of 15 km. Recent studies show that (estimated)Iquique outlines. (M 8.8) earthquake Relocated [Chlieh 2014et al., 2011] earthquakes is indicated by area thick shown by colour: Hayes et al. (2014) Nature great earthquakes have a complexfrom rupture the mainshock asperity. Almost all 2014 events have been located foreshocksdotted in line. red; Topography those between and bathymetry the are mainshock from ETOPO1 [Amante (largest and orange circle) on Eakins, 2009]. (b) The moment-rate function of the main shock. Yagimanner et al. (2014) GRL including back propagatingup-dip of the rupture zones of the two biggest quakes. 1 April 2014 and the largest aftershock (M 7.7) on 3 April inrupture orange; and and slip reactivation more around the recent eventshypocenter in yellow. [e.g., IdeRupture et al., 2011]. areas Therefore, of the we tookM 8.2 a slip-rate and M duration7.7 events of 56 s on are each fault patch, whichRecent was megathrust earthquake sequences (see, for example, refs 21, 22) colouredexpanded and contoured into linear at B-splines 2.0-m with intervals. an interval The of 0.8 extent s. Based of on the preliminary northern analyses, Chile we assumedhave that demonstrated the the need for integrative real-time monitoring and seismic gaprupture is indicated front velocity with can be arrows. up to 3.4 Bathymetric km/s, which yielded data the are start taken time from of the thelinear B-splinesassessments at each that map seismic cycles into models of strain accumulation sub-fault. We also assumed no slip after 80 s from the initial rupture. GEBCO_08 grid30. near the source regions of large earthquakes. The Iquique sequence is an ideal case study involving the integration of geodetic, geodynamic and YAGI ET AL. ©2014. American Geophysical Union. All Rights Reserved. 2 Between the mainshock and this largest aftershock, earthquake loca- seismological constraints to improve the quantification and assessment tions migrated southwards, towards the epicentre of the M 7.7 event. of an earthquake sequence as it evolves. Organizing seismotectonic infor- Since then, aftershocks have been scattered up-dip of the rupture areas mation for major global plate boundaries2,23 is crucial for understanding of the two largest events. the spectrum of expected behaviour of a fault zone after a major event has Analyses of regional moment tensors19 (RMTs; Methods) of fore- occurred (and indeed beforehand), and is the foundation of any frame- shocks indicate that many represent thrust faulting on or near the plate work for better communication of time-dependent earthquake hazards interface (Fig. 2). However, about 20% have well-constrained depths too to affected communities24. shallow to involve interface slip, have focal mechanisms inconsistent To understand the scale of this earthquake in the context of the Chil- with thrust faulting, or have nodal planes severely rotated with respect ean subduction zone, it is useful to compare the Iquique sequence with to local slab structure. Although constraining the absolute depths of off- the 2010 M 8.8Mauleearthquake2, the Chilean margin’s last great mega- shore earthquakes in subduction zones is difficult, the spread of solu- thrust event. Although both occurred in recognized seismic gaps, their tions (which are all subject to similar uncertainties) and the systematic evolution, behaviour and characteristics were quite different (Extended rotation of many mechanisms with respect to slab geometry indicate Data Fig. 4). In contrast to Iquique, Maule did not have any recognized that upper plate faulting was involved in the foreshock sequence. This foreshocks. Its mainshock nucleated in the middle of the South Central is particularly true of the M 6.7 event on 16 March, whose depth (20 km, Chile Seismic Gap1 and ruptured bilaterally beyond the extent of the gap NEIC; 20.6 km, CSN; 12 km, global centroid moment tensor project into the 1985 M 8.0 and 1960 M 9.5 rupture zones to the north and south, (global CMT)20; 15.5 km, NEIC W-phase) is substantially shallower respectively. The Iquique earthquake nucleated within the Northern Chile than the slab, and whose shallownodalplaneisrotated 60u (globalCMT; Seismic Gap, but in a region that had slipped more recently than much of CSN W-phase) to 70u (NEIC W-phase) anticlockwise with respect to the gap, and it ruptured an area much shorter than the gap’s recognized the slab, perhaps indicating that this foreshock occurred above the plate extent5,6,11,25. Over the two months surrounding the Maule mainshock, interface along a splay fault. Aftershock RMTs show that the vast majority earthquakes demonstrated a typical Gutenberg–Richter relationship, with occurred along the megathrust interface (Fig. 2), surrounding regions a b value of ,0.85 (Extended Data Fig. 4). The Iquique sequence, on the of largest co-seismic slip. The M 7.7 aftershock ruptured a compact por- other hand, is deficient in moderate-to-large events (b < 0.73, in contrast tion of the seismogenic zone ,50 km south and directly along strike with b < 1.02 over the preceding year). Co-seismic2 and post-seismic26

Kinematic analysis of the Mw 8.1 Iquique rupture makes use of two mapped by both the HFSR and co-seismic slip agrees quite accurately complementary approaches. First, weperformedwaveformmodellingof with the downdip end interseismic coupling (Figs 2a, c and 3a). The local strong motion seismograms and teleseismic body waves to constrain accelerated downdip rupture propagation for both earthquakes closely the kinematic development of the rupture towards the final displacement followed the gradient towards higher locking. The Iquique event and in a joint inversion with continuous GPS data of static displacements its largest aftershock therefore seem to have broken the central, only partly (Figs 1 and 2a). Second, we used the backprojection technique applied locked segment of the northernChile–southern Peru seismic gap, releas- to stations in North America to map the radiation of high-frequency seis- ing part of the slip deficit accumulated here since 1877 (compare Fig. 1). mic waves (HFSR; 1–4 Hz)9,10.Thelattertechniqueisnotsensitivetoabso- The seismicity before the Iquique earthquake also concentrated inthis lute slip amplitudes, but rather to changes in slip and rupture velocity. zone of intermediate locking at the fringe of the highly locked, high-slip During the first 35–40 s the rupture propagated downdip with increas- patch (Fig. 3a). Starting in July 2013, three foreshock clusters with increasing velocity, nearly reaching the coastline (Fig. 2a, b). Towards the end ingly larger peak magnitudes and cumulative seismic moment occurred of the rupture, the area near the epicentre was reactivated. In spite of here (Figs 2c and 3a, c). The mainshock rupture started at the northern the relatively complicated kinematic history of the rupture, the cumu- end of the foreshock zone, inside the region of intermediate locking lative slip shows a simple ‘bullseye’ pattern with a peak co-seismic slip (Figs 2c and 3a). The second foreshock cluster (January 2014) is associ- of about 4.4 m (Fig. 1), consistent with the slip model in ref. 11 based ated with a weak transient deformation, whereas the third cluster (March on teleseismic and deep-water tsunami wave recordings. A slip patch of 2014) shows a very distinct transient signal. GPS displacement vectors similar size and magnitude has been found12 but with the highest slip calculated over the times spanning these foreshock clusters point towards placed ,40 km further south from teleseismic-only recordings; how- the cluster epicentres (Extended Data Fig. 4). Deformation for both tran- ever, this seems not to be compatible with onshore displacement mea- sients is entirely explained by the cumulative co-seismic displacement of surements from GPS. The largest aftershock so far was the Mw 7.6 event the respective foreshock clusters (Fig. 3d inset and Extended Data Fig. 4). of 3 April that, like the mainshock, ruptured initially downdip. Rupture The area affected by the foreshocks is then reactivated by the aftershock then propagated northeastwards, attaining a peak slip of 1.2 m after about sequence of the Iquique earthquake (Fig. 2c). Comparing this with the 20 s. Both earthquakes broke a total of 200 km of the margin. long-term deformation history of the margin, we found that the same The Iquique rupture affected an area shown from geodetic analysis area shows a high gradient of locking from weakly locked updip to fully to be a zone of intermediate interseismic coupling at 18.5u–21u Sinter- locked downdip (Fig. 3a). rupting the otherwise fully locked northern Chile–southern Peru gap8,13 Additionally, the analysis of the frequency–magnitude distributionof (Fig. 3a). The Iquique mainshock nucleated at the northwestern border the foreshocks reveals outstanding characteristics in both the spatialand of a locked patch and ruptured towards its centre (Figs 2a and 3a). The temporal vicinities of the rupture. The earthquake number N as a function downdip end of the mainshock and for the large Mw 7.6 aftershock rupture of the magnitude Mis found to follow the well-known Gutenberg–Richter

a –71º –70º c –71º –70º Mainshock Aftershock

–19º 04020 60 80 s 0 20 40 s –19º

–20º 3.0 km s–1 –20º

–1 2.0 km s 20140316 M 6.7

–21º –21º b 0

) 20 –1 4 Depth (km) 40

N m s 60 19 2 Days relative to mainshock Moment rate (10 80 2.5 × 1021 N m –400 –200 0 2 4 6 8 10 0 100 0 30 60 90 120 150 180 –71 –70 Time (s) Longitude

Figure 2 | Kinematic rupture development of the Mw 8.1 main and Mw 7.6 waveforms. b, Moment rate and time history of backprojected energy (arbitrary aftershock and the distribution of foreshocks and aftershocks. The absolute scale). Black solid line, mainshock kinematic source-time-function; nucleation point of each earthquake rupture is indicated by a coloured star. red dashed line, rescaled backprojection energy. c, Map (top) and longitudinal a, Arrows indicate the propagation of main energy release during the first 40 cross-section (bottom) of ,3,600 foreshocks and ,1,400 aftershocks coloured and 25 s for mainshock and aftershock, respectively. The contour lines according to their time of occurrence relative to the mainshock. The slips of represent isolines of slip rate for the mainshock from the kinematic inversion the mainshock and largest aftershock are contoured. The beachball depicts the 21 during different time intervals after rupture nucleation (0.05 m s intervals, double-couple of the largest Mw 6.7 foreshock that had a rupture geometry line thickness scaled by slip-rate). Coloured diamonds represent maxima of distinctively different from the mainshock and largest aftershock (Fig. 1) and a semblance scaled to the peak value of the emitted energy for mainshock and centroid depth of only 9 km (blue star) and that thus probably occurred in the aftershock for each time step based on the backprojection of teleseismic upper plate.

Fig. 2. Seismicity preceding the Iquique earthquake. (A) Gray dots show the foreshocks from 16 to 31 March; the intensity of gray indicates the depths of the events. The slip distribuon of Mw 8.1 and Mw 7.6 earthquakes inverted from far-ﬁeld broadband records of the FDSN network is shown with the color. (B) 15 km wide cross-secon along the line A- Aʹ shown in (A). (C) 15 km wide cross secon along the B-Bʹ line. In the vercal cross-secons we plot the focal mechanisms of events with Fig. 2. Seismicity preceding the Iquique earthquake. (A) Mw larger than 4.6. Mechanisms were computed by broad band moment tensor analysis. The Gray dots show the foreshocks from 16 to 31 March;gray curve shows the seismogenic contact the intensity of gray indicates the depths of the events. The slip according to (16). distribution of Mw 8.1 and Mw 7.6 earthquakes inverted from far-field broad-band records of the FDSN network is shownRuiz et al. (2014) Science with the color. (B) 15 km wide cross-section along the line A- $ƍVKRZQLQ $ C) 15 km wide cross section along the B-%ƍ line. In the vertical cross-sections we plot the focal mechanisms of events with Mw larger than 4.6. Mechanisms were computed by broad band moment tensor analysis. The gray curve shows the seismogenic contact according to (16).

/ http://www.sciencemag.org/content/early/recent / 24 July 2014 / Page 5 / 10.1126/science.1256074

Geophysical Research Letters 10.1002/2014GL061138

Mw8.1 N b a

Mw8.1

Mw6.7 Trench Paralell Distance (km)

S Jan. Feb. Mar. Apr. Date Mar. Apr.

Figure 4. (a) Space-time diagram of all the detected events before and after the 2014 Iquique, Chile Mw 8.1 earthquake. The blue and red circles denote the foreshocks and aftershocks, respectively. The red stars indicate the repeating earthquakes. The diagram shows the earthquake origin times and locations projected onto the strike of the fault plane. The black, yellow, and white stars denote the hypocenters of the Mw 8.1 main shock, the largest Mw 6.7 foreshock, and the largest aftershock Mw 7.7, respectively. Focal mechanisms (from the USGS) are plotted as green beach balls. While all available focal mechanisms are used from January to February, only focal mechanisms with Mw > 5.0 are selected from 1 March until the main shock origin. (b) Enlargement of Figure 4a showing the intensive foreshocks between 14 March and 1 April 2014 (blue circles scaled to magnitude). The red dashed lines represent the approximate locations of the fronts of earthquake migrations.

(GFZ) Potsdam [http://geofon.gfz-potsdam.de/]. We focus on the continuous waveforms retrieved between 1 January and 6 April 2014. Kato and Nakagawa (2014) GRL The waveform data (both continuous and template waveforms) were filtered with a passband from 1 to 8Hzanddecimatedfromanintervalof0.01sto0.04s.Astemplateevents,weused375available earthquakes (184 foreshocks and 191aftershocks)listedintheUSGSevent catalog occurring between 1 January and 6 April 2014 (Figure 1b). The magnitudes of most template events are >4.0 (Figure S1 in the supporting information). To obtain template event waveforms and target waveforms, we cut a 10s window of the waveform starting from 5.0 s prior to the synthetic S wave arrival time. The synthetic S wave arrival time from each template location to each station was calculated using a one-dimensional seismic velocity structure derived from a series of seismic surveys conducted near the source region (Table S1 in the supporting information) [e.g., Patzwahl et al., 1999; Andean Continental Research Project Working Group,2003]. We computed correlation coefficients between a template event waveform and a target waveform, shifting the 10 s window by increments of 0.04 s through the continuous waveforms. At each time step, we calculated the average correlation coefficients over the seismic network whenever the total number of available station channels was greater than 15. We set a threshold for event detection for each template as 10 times the median absolute deviation of the average correlation coefficients calculated throughout the day of interest. To remove multiple detections, we allocated the location of the detected event to that of the template event with the highest correlation coefficient within the ±5 s window. We identified 3348 events, which is nearly 10 times the number listed in the USGS catalog (375 events) (Figure S1 in the supporting information). As an example of a detected event missing from the catalog, our matched filter technique recovered a moderate magnitude event (M 4.3) at a time when no seismic events are reported in the USGS catalog (Figure 2). For more details of the matched filter technique, see Kato et al.[2013]. In addition, we extracted repeating earthquakes from the newly detected catalog using the waveform data for earthquakes of M 2 or greater. We selected earthquake pairs whose epicenter separations are up to 25 km and calculated cross-correlation coefficients using 0.5–2, 1–4, 2–8, and 4–16 Hz passband-filtered vertical seismograms [e.g., Igarashi, 2010; Kato and Igarashi, 2012]. The time window of each seismogram was from the P wave arrival to 30 s after the direct S wave arrival. We categorized repeating earthquake pairs as those whose cross-correlation coefficients (for the frequency range roughly corresponding to the corner frequency of

Fig. 3. Motion of coastal GPS stations preceding the Iquique earthquake. (A) North components and (B) East components relative to a linear evolution model with seasonal variations estimated since 2012 (14). The thick red line denotes the origin time of mainshock while the black dotted lines show the occurrence time of the Mw > 6 foreshocks. Ruiz et al. (2014) Science

/ http://www.sciencemag.org/content/early/recent / 24 July 2014 / Page 6 / 10.1126/science.1256074

LETTER RESEARCH

a b c 8

6 –19º

Magnitude 4 d 1,250 GPS signal 14 Mar. 8 Pisagua Pisagua 1,000 4

–20º 0 750 0 W-motion (mm) Iquique Iquique 500 –400 –300 –200 –100 0 13 July 250 Cumulative event number Days

–21º 14 Jan. 0 –250 e 1.0 1.0 1.2 0.9

0.8 1.0 0.8 0.6 –22º value value b 0.8 b 0.7 0.4 Locking 0.6 0.2 0.6

0 0.5 –2,000 –1,500 –1,000 –500 0 –23º –72º –71º –70º –72º –71º –70º Time relative to mainshock (days) Figure 3 | Maps of interseismic locking and b value, and time history of lines are fits of the estimated background (grey shaded area) for the four phases, seismicity and deformation. a, Geodetic interseismic locking and foreshocks. during each of which the background rate is almost constant (see the text). The July–August 2013, January and March 2014 foreshock clusters are marked. The inset shows measured GPS displacement time-series stacked from b, b-value map of the central portion of northern Chile gap for the last 2000 near-coast stations between 19u S and 21u S smoothed with a four-day moving days before the mainshock, where results are calculated for all M $3 foreshocks average and the modelled signal related to cumulative slip of the foreshock within 50 km if their number exceeds 100. The rectangular box encloses the area events. e, Time series of b value (means 6 s.d.) for the events inside the box used for the results in c–e. c, Magnitude–time plot. Arrows mark the July– (coloured and grey shaded area). Black points and bars refer to the results for August 2013, January and March 2014 clusters. d, Observed (black thick line) the events outside the box. and ETAS-modelled (red line) cumulative M $3 activity; the thin coloured

2bM relation, N < 10 . b valueshavebeenproposedtoactasastresssensor, in foreshock activity preceding subductionSchurr megathrust et al. (2014) Nature events has repeat- with low b values indicating high stresses14.Mappingtheb value in Fig. 3b edly been reported18,20,21. In contrast to, for example, the foreshocks of indicates significantly lower b values in the source area than in all other the 2011 M 9 Tohoku-Oki earthquake21, which started about three weeks regions where the b value can be resolved. A gradual decrease in the b value before the mainshock, the foreshock clusters described here have been from about 0.75 to below 0.6 is observed in the source region within the active since the start of our observations in 2007. However, only more three years before the Iquique earthquake (Fig. 3e). This only reverses recently have foreshocks with increasing magnitudes and thus more within the last days of the foreshock sequence. Similar decreases in b values fertile aftershock sequences resulted in a supercritical state with self- before large megathrust earthquakes have recently been documented, accelerating seismicity19.Thisultimatelyallowedanearthquaketonucle- in particular for the M 9 Tohoku event15. ate that was strong enough to break into the stronger locked and thus To reveal potential changes of background seismic activity related to aseismic part of the interface. From the spatial correlation of foreshock aseismic processes, we fit the foreshockseismicitybymeansofanEpidemic- activity and slip gradients with the Iquique mainshock rupture region, Type-Aftershock-Sequence (ETAS) model16, identifying only 42% of we infer gradual unlocking of the plate interface. Moreover, the start of pre-mainshock events as Omori-type aftershocks triggered by larger fore- the decrease in b value ,3 years ago, in spite of constant plate conver- shocks. Using the Akaike information criterion17, the remaining back- gence and loading rate, indicates a physical change at the plate interface. ground rate is found to be significantly time dependent: we identified From only intermediate locking the Iquique segment seems to havebeen four subsequent periods of almost constant background rates (Fig. 3d, e). dominated by mostly smaller locked asperities embedded in a condition- The mainshock preparation process seems to have been initiated by a ally stable environment, if we assume that rate and state frictional beha- relative seismic quiescence, which starts at the same time as b values are viour is controlling locking and creep22–24. Progressive rupturing of the found to drop and ends in July 2013, when the background rate returns smaller asperities by foreshocks will have loaded the remaining larger to approximatelythe pre-quiescence values;thefinal phase starts18 days asperities in this zone until their failure: this evolution may be seen as before the quake, when the background seismicity increases more than the culmination of a runaway process as the likely key mechanism lead- 35-fold in conjunction with the onset of the final transient GPS signal. ing up to the Iquique earthquake. Both the rupture direction and speed Such a sequence of seismic quiescence, recovery and acceleration of of the mainshock and its triggered large aftershock are controlled by background activity is expected inthe stress accumulation framework18. the stress gradient in the remaining asperities, corroborating theoretical On the basis of the inverted ETAS parameters and the low b values before analysis25.Thisindicatesthatnotonlythesizeoflargerasperitiesbutalso the mainshock, the seismicity is expected to accelerate with time as the their stress topography is important for understanding the propagation, result of a branching ratio—that is, the average number of daughter events acceleration and stopping of megathrust earthquakes. The Iquique event per earthquake—larger than one indicating a transient supercritical state19. broke a region of heterogeneous coupling, where rupture not only broke Hence, the period leading to the Iquique earthquake documents progres- amoderate-sizedasperitybutalsopenetratedintoaweaklycoupledzone, siveasperity failure, here observed in unprecedented detail. An increase possibly by dynamic weakening26.

a Figure 3 | Coulomb failure (DCFS) stress changes for foreshocks and –19º aftershocks. DCFS is resolved onto the subduction zone interface13 1 day before the mainshock (a) and after the largest aftershock (b). Earthquakes are coloured by the local DCFS resulting from previous earthquakes, at their time of occurrence; light grey symbols indicate negative stress changes, and dark grey symbols, positive. The slip models of the mainshock and the M 7.7 aftershock are shown with white contours (1 m intervals). The dashed box in b represents the spatial extent of the region in a.

slip during the Maule sequence indicated rupture of most of the seismogenic zone, resulting in aftershock mechanisms that spanned a broad –20º range of the faulting spectrum (Extended Data Fig. 4). In contrast, the Iquique sequence has not elicited a clear upper plate or outer rise response. Future studies of regional GPS data27 may reveal the extent of post-seismic interface slip and its relation to the 2014 Iquique sequence. We can use Coulomb failure stress (DCFS) analysis28 to assess whether the 2014 Iquique earthquake sequence followed a spatial and temporal migration pattern dictated by the stress changes caused by previous earthquakes. DCFS calculations (Fig.3) showthat 18 of the20 foreshocks with associated RMTs ruptured the megathrust interface where it had been positively stressed by previous events, and loaded the hypocentral –71º –70º region of the subsequent M 8.2 event by 0.04 MPa. It thus seems that b –17º the northward migration of foreshocks responded to cascading DCFS, ultimately leading to the mainshock. The hypocentral region of the M 7.7

20020 2252222522 15015 1751 aftershock was loaded 0.25 MPa by the mainshock and the first 27 h of 100 125 75 0 55 50 75 0 aftershocks. Aftershocks have generally nucleated in areas of increased 25 –18º DCFS, surrounding the main slip patches of the largest events. Overall, the hypocentres of ,70% of relocated aftershocks (94 of 138 events) occurred in areas of positive DCFS. If uncertainties in relocated hypocentres –19º (62–3 km) are considered (Fig. 3), more than 80% of aftershocks occurred in regions of positive DCFS, lending support to the real-time use of DCFS modelling as earthquake sequences unfold to aid in anticipating likely locations for subsequent events. –20º Analysis of moment accumulation and release along the plate margin (Fig. 4) shows that the main asperity ruptured in the M 8.2 earthquake partly filled a historical gap in moment relative to adjacent sections of the arc. The March 2014 foreshocks began near the northern extent of –21º the 1933 event, and moved northwards. Despite surrounding regions oflarger moment deficit, thesouthernportion oftheM 8.2 and the M 7.7 aftershock ruptured approximately the same region as the 1933 event. –22º Moment from both large 2014 earthquakes, and from all historical events, 100 km falls off rapidly to the south between 20u S and 21u S; here, a 50–80-km section of arc has seen little to no seismic activity over the past century –72º –71º –70º –69º –68º or more, until the northern extent of the 1967 M 7.4 earthquake is reached.

North South 20 10 100% 50% 19

10 Moment per km (N m)

1018

1017

1016

1015 0 100 200 300 400 500 Distance along trench (km) Figure 4 | Moment deficit along the northern Chile subduction zone. accurate rupture areas are used8. Red shows moment for historical earthquakes; Moment calculated for historical seismicity from the USGS Combined blue for 2014 seismicity (dark blue up to 3 April 2014; light blue since then); Catalog32 since 1900, resolved as moment per kilometre along strike. For each green represents all summed moment. Horizontal dashed lines represent earthquake, moment is divided evenly over the length of the rupture, calculated moment accumulation levels given constant coupling percentages of 50%, 75% using empirical relations14. For the largest earthquakes (M .7.5), more and 100%.

breaks down rapidly as the pests or patho- reducing unsustainable inputs such as irriga- In seeking new crops to sustainably feed gens evolve to evade or overcome the resis- tion water and nitrogen fertilizer (6 ). More an expanding world population, there is com- tance gene. Transgenic approaches allow the audacious goals include the transgenic engi- pelling need for a multipronged approach

simultaneous introduction of multiple differ- neering of rice (and other C3 crops) to intro- that includes traditional breeding, molecular ent resistance genes (stacking) into a single duce C4 metabolism—a more effi cient car- breeding, and genetic modifi cation. We need plant variety, which generates more durable bon fi xation pathway—and thereby increase to accelerate this new green revolution in the resistance in a desired variety. This strategy photosynthetic capacity (13 ). lab, in the fi eld, and through better communi- is analogous to the use of drug combinations Although the “intelligentReferências breeding” cation outside the scientifi c community if we (cocktails) when treating diseases like tuber- approach that is guided by DNA mark- are to address the nearly 3 billion chronically culosis and HIV infection, and has proven ers has been much less controversial, there undernourishedPUBLICATION people worldwide. S effective both in disease treatment and in remains considerable public opposition References LETTER doi:10.1038/nature13681 preventing the emergence of multidrug- to the deployment of genetically modifi ed Geophysical 1. H. C. J. Godfray et al., ResearchScience 327, 812 (2010). Letters resistant bacteria and viruses. organisms (GMOs), especially in the food 2. H. C. J. Godfray, T. Garnett, Philos. Trans. R. Soc. Lond. B 369 Transgenic modification also permits supply. One consequence has been the focus Biol. Sci. , 20120273 (2014). Gradual unlocking of plate boundary controlled RESEARCH 3. W. Lutz, S. K LETTERC, Philos. Trans. R. Soc.Rupture Lond. B Biol. process Sci. of the 2014 Iquique Chile Earthquake the alteration of specifi c biochemical path- of regulatory agencies on the technology to 10.1002/2014GL060274365, 2779 (2010). ways and genetic networks. Examples can develop the plant rather than on the proper- 4. D. Tilman, C. Balzer, J. Hill, B. L. Befort,in relation Proc. Natl. Acad. with the foreshock activity initiation of the 2014 Iquique earthquake Sci. U.S.A. 108, 20260 (2011). Key Points: Yuji Yagi1, Ryo Okuwaki2, Bogdan Enescu1, Shiro Hirano3, Yuta Yamagami2, Suguru Endo2, be found among efforts to enhance nutri- ties of the engineered plant itself. This regu- • 2014 5. Iquique C. K. Khoury earthquake et al rupture., Proc. Natl. Acad. Sci. U.S.A. 111, Bernd Schurr1, Gu¨nter Asch1, Sebastian Hainzl1, Jonathan Bedford1, Andreas Hoechner1, Mauro Palo1, Rongjiang Wang1, 2 process inverted applying and Takuya Komoro Marcostion, Moreno including1, Mitja Bartsch the1 ,genetic Yong Zhang modifi2, Onno cation Oncken1 ,of Frederik latory Tilmann morass1, Torsten has Dahm delayed1, Pia Victor the1, Sergio availability Barrientos 3of 4001 (2014). PUBLICATIONS a novel formulation 4 6. N. V. Fedoroff et al., Science 327,1 833 (2010). 2 & Jean-Pierre“golden Vilotte rice” to increase its beta-carotene golden rice more than a decade at a proba- • Stress release of foreshocks may Faculty of Life and Environmental Sciences, University of Tsukuba, Tsukuba, Japan, Graduate School of Life and 7. P. Hedden, Trends Genet. 19, 5 (2003). 3 explain the initial rupture phase Environmental Sciences, University of Tsukuba, Tsukuba, Japan, Faculty of Engineering, Information and Systems, content (11 ). The overexpression of a tran- ble cost of tens of millions of lives (6 , 14). 8. D. N. Duvick, Maydica 50, 193 (2005). of main shock University of Tsukuba, Tsukuba, Japan On 1 Aprilscription 2014,Northern factor Chile from wasstruck the by aflowering magnitude 8.1earth- plantrupture Efforts area was to less address than 2.6 m (refsconcerns 6, 7) leaving about most of the past safety slip • Foreshock 9. L. Cabrera-Bosquet activity and slow slipet al., J. Integr. Plant Biol. 54, 312 8 quakeArabidopsis following a protracted thaliana seriesGeophysical of foreshocks. in soybean The Integrated dramati- Plate deficit Researchof of ,GMOs8–9 m accumulated have chiefl since Letters 1877y focused unaffected .on On 1evidence- April 2014, contributed(2012). to the triggering Boundary Observatory Chile monitored the entire sequence of events, the Mw 8.1 Iquique earthquake north of Iquique struck the central por- 10.of the J. main D. G. shock Jones et al., Philos. Trans.Abstract R. Soc. Lond. TheB Biol. rupture process of the 2014 Iquique, Chile earthquake is inverted from teleseismic P wave providingcally unprecedented boosts resolutiongrain yield of the build-up by extending to the main event thetion ofbased the gap. refutation Using seismological of claims and geodetic of observationsreal and wepoten- here Sci. 369, 20130087 (2014). data applying a novel formulation that takes into account the uncertainty of Green’s function, which has and its rupture evolution. Here we show that the Iquique earthquake analyse the rupture, its relationship toprevious locking of the plate inter- 11. T. Johns, P. B. Eyzaguirre, Food Policy 32, 1 (2007). length of time of vegetative development. tial adverse outcomes ( 6), but have not alle- Supporting Information: 21 broke a central fraction of the so-called northern Chile seismic gap, face, and the pre-seismic transients leading up to the earthquake. 7 been a major error source in waveform inversion. The estimated seismic moment is 1.5 × 10 Nm (Mw = 8.1), RESEARCH LETTER The 1 April 2014 Iquique,• 12.Readme S. B. Preuss Chile, et al., PLOS MONE w, e307178.1 (2012). earthquake the lastThe major result segment is of a the larger South American plant capable plate boundary of thatsup- viated concerns. The scientifi c community associated with a 140 km long and 140 km wide fault rupture along the plate interface. The source process is 1,2 • 13.Figures S.S1 von and Caemmerer S2 et al., Science 336, 1671 (2012). 2 M density had not ruptured in the past century . Since July 2013 three seismic 0 0 10.1002/2014GL060238 M 0 fi 1 341 characterized by unilateral rupture propagation. During the rst 20 s, the dynamic rupture front propagated 1868 porting greater seed yield ( 12). Improved is understandably2001 8 reluctant to move beyond 14. B. Alberts et al., Science , 1320 (2013). clusters, each lasting a few weeks, hit this part of the plate boundary rupture.4 sequence with earthquakes of increasing peak magnitudes. Starting with the 1906 Correspondence 15. J. Bailey-Serres to: et al., Rice 3, 138from (2010). the hypocenter to the large asperity located about 50 km southward, crossing a remarkably active water-use and nitrogen-use effi ciencies are evidence-based–18º logic, but must explore new

1948 Y. Yagi, second cluster, geodetic observationsKey show Points: surface displacements that 1 Arica 1 1 2 foreshock area at high velocity (of about 3.0 km/s), but small and irregular seismic moment release rate. Our Thorne Lay , Han Yue , Emily E. [email protected], and Chao An 10.1126/science.1254135 can beattractive associated with targets slip on the to plate increase interface. These yields seismic while clus- approaches to advocate GMOs. result may suggest that the 20 s long initial phase was inﬂuenced by the stress drop due to the foreshock 1956 tersand their slip transients occupied• The a part 1 April of the 2014plate interfaceMw 8.1 that earthquake 1987 –19º 1 activity near the main shock hypocenter.2 Moreover, the 2 week long swarm-like foreshock activity migrating was transitional between a fully lockedruptured and a creeping about portion. 20% Leading of the 1877 1945 Department of Earth and Planetary Sciences,Citation: University of California, Santa Cruz, California, USA, School of Civil and up to this earthquake, the b value of the foreshocks gradually decreased 1877 2014 04 01 M 8.1 Pisagua roughly at 5 km/day toward the main shock hypocenter, and possibly associated slow slip, contributed to the seismic gap 1869 Yagi, Y., R. Okuwaki, B. Enescu, S. Hirano, 1911 duringGEOPHYSICS the years before the earthquake, reversing its trend a few days 1933 Environmental Engineering, Cornell University, Ithaca, New York, USAstress accumulation prior to the Mw 8.1 megaquake. The main shock initial rupture phase might have • The rupture was very localized and did –20º Iquique Y. Yamagami, S. Endo, and T. Komoro before the Iquique earthquake. The mainshock finally nucleated at LETTER doi:10.1038/nature13677 (2014), Rupture process of the 2014 triggered the rupture of the large asperity, which had large fracture energy.

the northern end of the foreshock area,not which rupture skirted to athe locked trench patch, 1940

2014 04 03 M 7.6 Se Iquique Chile Earthquake in relation 1871 and ruptured mainly downdip towards• higher locking. Peak slip was ismic gap The northern and southern ends of the –21° withSeismic the foreshock activity activity, precedingGeophys. recent large 1906 attained immediately downdip of theforeshockregionandatthemar- 1905 Res. Lett., 41, doi:10.1002/ Recognizing1877 gap have nowForeshocks had similar ruptures Abstract from–1 On 1 April 2014, a great (Mw 8.1) interplate thrust earthquake ruptured in the northern gin of the locked patch. We conclude that gradual weakening of the earthquakes, including the 1 April 2014 1967 2014GL060274. Latitude 1. Introduction central part of the seismic gap accentuated by the foreshock activity 6.7 cm yr

portion2007 of the 1877 earthquakeContinuing seismic gap in northern megathrust Chile. The sequence earthquake commenced on potential 16 in Chile –22° earthquake in Chile, hints that some large in a zone of intermediate seismic coupling was instrumental in caus- On the night of 1 April 2014, a great thrust earthquake (Mw 8.1) occurred off Iquique, northern Chile. The 1928 Received 18 APR 2014 ing final failure, distinguishing the Iquique earthquake from most March 2014 with2007 a magnitude 6.7earthquakes thrust event, may followed potentially by be thrust-faultingpredictable. aftershocks that migrated the 1 AprilSupporting 2014 Information: Chile Earthquake5 Accepted 2 JUN 2014 earthquake information determined by the Centro Sismológico Nacional (CSN), the University of Chile M 7.7 after the 2014 Iquique earthquake great earthquakes. Finally, only one-third of the gap was broken and Accepted article online 5 JUN 2014 • Readme –23° 4 Mejillones [http://www.sismologia.cl; last access on 18 April 2014] is as follows: origin time = 1/4/2014 23:46:45 (UT); the remaining locked segments now pose a significant, increased seis- 1947 northward ~40kmpeninsula over 2weeks to near the main shock hypocenter. Guided by short-period teleseismic 1995 3 slip 1 2 epicenter = 19.572S, 70.9081 W; depth = 38.92 km; Mw = 8.2. In the off3 Iquique region, the1 Nazca Plate is mic hazardEmily with E. theBrodsky potentialto and host• Figure Thorne anearthquake S1 Lay with a magnitude (m) Gavin P. Hayes , Matthew W. Herman , William D. Barnhart , Kevin P. Furlong , Seba´stian Riquelme , Harley M. Benz ,

1998 ﬁ of .8.5. P wave2 backprojectionsAntofagasta and inversion4 of deepwater tsunami3 wave1 recordings, a5 nite-fault inversion of • Figure S2 1987 Eric Bergman , Sergio Barrientos , Paul S.subducting Earle & Sergey at a rate Samsonov of about 63 mm/y beneath the South American Plate along the Peru-Chile trench –24° 1995 The northern Chile–southern Peru seismic gap last broke in 1877 in 1 M 8.1 1 • Figure S3 1942 teleseismic P and SHBB+SMwaves using a geometry consistent with[Kendrick long-period et al., 2003], where seismic several waves great interplate resolves earthquakes a have been observed since the dawn of a great earthquake (Mw ,8.8) that ruptured from south of Arica to the re there measurable, distinctive pre- probability of100 any km given earthquakeSM being a by similar sequences extending to months 1936 1910 1947 history. The coupling rate on the plate interface has been estimated from the Global Positioning System (GPS) Mejillones peninsula (see Fig. 1). The• reFigureported historical S4 recurrence inter- 1960 GPS 50 cm

spatially1988 compact large-slip (~2–6.7m) zone1 located ~30km downdip and ~30km along-strike south of cursors that can warn us in advance6 foreshock is low, because small earthquakesTheseismic prior gap to theory mainshocksidentifies regions ( 8). It of dataseems elevated [e.g., that hazardChlieh some et based al., 2011;a seismicMoreno hazardet al., 2011]. perspective Figure is 1 thatshows the the fault aftershock can host activity another during event of the first 2 days val for the past 500 years in this region• has been estimated at 111 33 1900 2000 –73° –72° –71° –70° –69° 1 Figure S5 21 years , making it probablyof thePUBLICATION the planet’s most mature largest seismic gap earthquakes? along theS South areTime often (years) notthe followed hypocenter, byLon glargeitude downdip ones. ofon the a lack foreshocklarge of recent earthquakes seismicity sequence. in comparison might The be main with afterpredictable. other the shock megaquake portions seismic of (hypocenters a a moment similar magnitude.determined is 1.7 Although by× CSN)10 and aNm great-sized the focal earthquake mechanism here of the had main been shock • 10 American A plate boundary. In the pastFigure two decades S6 the two adjoining fault.It has successfully explained past earthquakes(see, for example, expected, it is possible that this event was not it . 16 Foreshocks have long been considered theFigure 1 | MapData of Northernfor the Chile 2011 and Southern Mw 9.0 Peru Tohoku, showing historical Japan, This newfound optimism– determined is tempered in the presentby study. The main shock mechanism is consistent with the tectonic stress buildup!"#$%&' at ( segments south and north broke in• theFigureMw 8.1 S7 Antofagasta earthquake with a fault dip of 18°, radiatedref. 2) seismic and is useful energy for qualitatively of 4.5 describing8.4 × 10 whereJ, and large earth- static stressSections drop of this of subduction ~2.5 MPa. zone have ruptured since 1877 (Fig. 1), earthquakes and instrumentally recorded megathrust ruptures. IPOC the plate boundary. It is particularly notable that this megathrust event occurred within a plate boundary of 1995most (refs 3, promising 4) and the Mw 8.4 Arequipacandidates earthquake for of 2001predicting in south- instruments earthquake used in the present suggest study (BB, broadband; that detectable SM, strong motion) pre-quakes are mighttwo occur.major A scientifi large earthquake c and practical had been expected concerns. in the most notably in 1967, in a M 7.4 event between ,21.5 Sand22 S, and 5 Most of the 1877 gap remains unbroken and hazardous. u u ernGeophysical Peru . In the previous cycle the Research southern Peru and northern Letters Chile 1,2 segment3–6 that has not been ruptured since the great 1877 event, with an estimated magnitude of about 8.8 earthquakes. At least half of large earth-showncursory as blue symbols. processes Left: historical and may instrumental occur earthquake for record. somesubduction First, zone it adjacent remains to northern diffi cult Chile to distinguish, which had not ele- rup- inthe 2007 M 7.7 Tocopillaearthquake between ,22u Sand23.5u S.Slip segments broke within few years (Fig.Correspondence 1), suggesting that they to: might be Centre: rupture length was calculated using the regression suggested in ref. 28, [Chlieh et al., 2011; Figure 1]. 1 tured in a megathrust earthquake since a M 8.8 event in 1877. On during these events was limited to the deeper extent of the seismogenic coupled in time . The imminence of a large megathrust event in this with grey lines for earthquakes M .7 and red lines for M .8. The slip quakes have foreshocks,T. Lay, but these fore- large plate boundary earthquakes.w Twenty-1 April 2014vated a M earthquake8.2 earthquake activity occurred withinprior this to seismica main- gap. zone, leaving shallower regions unruptured6,11. Farther south, the 1995 regionRESEARCH motivated the settingLETTER up of an internationalMultiple monitoring effortslow-slipdistribution of events the 2014 Iquique during event and its largest a foreshock aftershock derived in this In this study we estimate the detailed rupture process of the 2014 off Iquique earthquake from teleseismicary 2014, when an Mw 5.7 interplate shocks are diffi [email protected] or even impossible tostudy three are colour days coded, withbefore contour intervalsthe earthquake, of 0.5 m. The green anda series blackHere we presentshockan from assessment earthquake of the seismotectonics swarms that of the do March– not M 8.1 Antofagasta earthquake broke the seismogenic zone immediately with10.1002/2014GL061138 the Integrated Plate Boundary Observatory Chile (IPOC). Having P wave data, using a newly developed inversion method [Yagi and Fukahata, 2011a] that takes into thrust event took place at (20.69°S, started in 2007, there now exists an exceptional database that monitors vectors are the observed and modelled horizontal surface displacements ofApril the 2014 Iquique sequence, includingIntense analyses of earthquake foreshocks reloca- south of the Mejillones and Peninsula, a featurea slow argued to be aslip persistent bar- distinguish from non-precursorysequence seismic of theof 2014smaller1. Iquique, earthquakes Introduction Chile began, migrating4,5,7 culminate in a major eventaccount (9 ). theThe uncertainty fore- of Green’s function, which has been a major error source in waveform inversion.70.80°W) on the southern edge of the the gradual plate boundary failure with various geophysical techniques. mainshock. The slip areas of the most recent other large ruptures are alsotions, moment tensors, finite fault models, moment deficit calculations rier to rupture propagation11,12. Adjacent to the southern coast of Peru, Key Points:activity. The foreshocks. for the 1 April 2014plotted. toward Right: moment the deficit future per kilometre mainshock along strike left hypocentral along the plate shocks of Tohoku may haveBased been on a the random rupture process of the main shock and characteristic foreshock activity, we propose a Iquique LCZ (5). On 8 January, another Several major earthquakes (Mw 7) have occurredMw 8.1 in this gap earthquake since boundary after the Iquique event for moment accumulated since 1877, and cumulative Coulomb stress transfer. This ensemble of informa- a seismic gap associated with the 1868 M 8.8 rupture was partly filled by • Foreshocks migrated both along strikeCitation: Mw 5.7 event occurred in the same area. 1850 (Fig. 1); the largest until now was the Mw 7.7 Tocopilla earthquake assuming current lockingNorthern (Fig. 3a). TheChile total accumulated experienced moment since 1877ation great allows subduction us to place the zone sequence megathrust withinevent thescenario context earthquake topreceded explainof regional the on foreshock-main 9 May 1877the shock with interaction.2014 an Iquique5 Mw andChile downdip event at rates of and 2–10 km/dother recent large1,2 earth- region at a2 rate of several kilometers per cluster that then triggered the mainshock the 2001 M 8.4 Arequipa earthquake. Coupling models indicate that in 2007, which broke the southern rimLay, of this T., segment H.Aitaro Yue, beneath KatoE. E. and Brodsky,and north Shigekifrom and 17u NakagawaSto25u S (red solid line) is 8.97; the remaining moment after The automatic location system of CSN • Slow-slipquakes events suggest occurred updip that of observable precursors day; then,estimated 2 days before seismic the mainshock, magnitude aseismicity M of8.7 –( 108.9 and). to [ ComteIf identify so, the areasand predictive of Pardo remaining, 1991] value and/or and of elevated the a tsunami hazard.fore- magnitudestrain accumulationMt of may 9.0. remain Recent to the southeast of the Arequipa event, of Mejillones peninsula along a totalC. length An (2014), of 150 km. The Only 1 the April down- 2014subtracting Iquique, all earthquake events with Mw .7 (grey dotted line) is 8.91 for the w identified 30 events in a smaller region the largest slip patch of the main 1 Our results constrain the size and spatial8.1 extent ofearthquake rupture, and indi- towards the northernmost edge of the 2014 rupture in Chile (a section dip endmay of the exist locked before zone slipped large in this earthquakes. event,Earthquake and the total and slip inVolcano the entire Research7.3 northern earthquake Center, Chile–southern Graduategeodetic Perustruck School seismic measurements of within gap. Environmental 10 km Studies, of of eastward Nagoyathe University,shocks deformation Nagoya,is limited Japan, of because the upper such2. plate Inversion random indicate clus- Analysis that most of the 1877 rupture of 15 km ! 15 km, in the period from 4 shock rupture Chile, Mw28.1 earthquake rupture cate that this was not the earthquake that had been anticipated. Signifi- ,200 km long). However, within that zone adjacent to Arica, at the pro- 1•GFZFinal Helmholtz slow-slip Centre event Potsdam, migrated German Research to Centre for Geosciences,Earthquake Telegrafenberg, Research 14473 Institute, Potsdam, Germany. University2School of of Earth Tokyo, and Space Tokyo, Sciences, Japan Peking University, Beijing 100871, China. to 24 January 2014. Then on 12 Janu- 3 Statistical modelssequence, of interactingGeophys. Res. earth- Lett., 41, mainshock 4nucleation site (see the fi gure,cant sectionsters of will the northern often Chilenot trigger subduction large zone) havemainshocks. not ruptured0 nounced bend in7 the subduction) zone and Peru–Chile; Trench, coupling> Centro Sismolo´gico National, Universidad de Chile, Facultad de Ciencias Fı´sicas y Matema´ticas, Blanco Encalada 2002, Santiago, Chile.zoneInstitut (Figure de Physique du Globe 1)from de Paris, 1, rue19°S Jussieu, to 75238 23°S Paris has a high coupling coef!"ficient,#$%&'To( imagealbeit*#+"# the+,-.&/ with rupture some( process#1"#2%,3'45&64 patchiness of the off Iquique( along#8"#$%&' earthquake, strike( #2" and#9,:-.3 we inverted( #$" teleseismic#<=436&3P wave( #?" data# recorded at the initiation point of the main in almost 150 years, so it is likely that future megathrustB earthquakesB is low and may) not support throughgoing rupture. ary, a small cluster of 6 identified cedex 05,quakes France. suggest that3818 –big3825, earthquakes doi:10.1002/2014GL060238. are panel A). Data from geodetic instruments However, the slow slip@&A3: inferred48( # broadband$" from#+464=&4A4 the network sea-( stations,#8"#?4CD./ which were# selected to ensure adequate azimuthal coverage and data shock rupture along dip [e.g., Béjar-Pizarro etwill al. occur, 2013; to theMétois south and et potentially al., 2013]. to the This north region of the 2014 has Iqui- been identiThe Nationalfied as Earthquake the north Information Center (NEIC) and Centro events occurred north of the previous Abstract To obtain a precise record21 of AUGUST the foreshock 2014 | VOL sequence 512 | NATURE before | the 299 2014 Iquique, Chile Mw 8.1 1 quality. The observed waveforms were shifted on the basis of their first arrival time and then converted into most likely to happen when regional earth- recovered from the sea fl oor after the Mw que9.0 sequence. fl oor geodetic data isDepartamento a more widespread, de Geofísica, FacultadSismolo de´gico Ciencias Nacional Físicas (CSN) y W-phaseMatemáticas, centroid Universidad moment tensor de Chile, (CMT) Chile. ones at (19.7°S, 71.0°W). During Feb- ©2014earthquake,Macmillan Publishers we applied Limited. a matchedAll rights reservedfilterChilean technique seismic to continuous gap seismograms [Kelleher, 1972; recordedNishenko near the source, 1985] based2 on thevelocity lack waveforms of large with earthquakes a sampling interval for of the 0.8 s. 137 Formitigating years the effect of aliasing3 and13 low-frequency Supportingquake Information: activity is alreadyReceived high 15 APR ( 1 2014– 4). How- earthquake showed that the fault had slippedOn 1 Apriland 2014,hence a M potentially8.2 earthquake moreIstituto ruptured predictable,Nazionale a portion di of Geofisica the pre- sub- e solutionsVulcanologia, for the Centro 1 April Nazionale 2014 earthquake Terremoti, align Rome, with the Italy. slab interfaceSchool of ruary 2014, another small cluster oc- fi 4 21 • Readme Acceptedregion. 21 MAY We 2014 newly detected about 10 times the number of seismic events listedduction in the zone routinely in northern constructed Chile offshore ofEnvironmental the citynoise, of Iquique, we Sciences, applied a major aUniversity port Butterworthand of indicate Li band-passverpool, a seismic UK.lter momentCentro between Sismológico of 0.001 (1.00–2.35) and 0.36Nacional,3 10 Hz beforeNm( FacultadM re-sampling.w 5 de8.07– Ciencias curred near (19.4°S, 71.0°W), where 16 ever, the same models also indicate that the slowly duringover the which foreshock the Nazca sequence plate ( has5, 6). been cursory underthrusting process. Unfortunately, South America the at sea-fl about oor 65 mm/yr [DeMets5 et al., 2010]. • Figure S1 earthquake catalog and identified multiple sequences of earthquake migrationsand hub at speedsfor the country’s of 2–10 copper km/d, miningFísicas industry. y Matemáticas, Peak shaking Universidad inten- 8.18). de OurChile, finite Chile. fault Institut solution de14 Physique(Methods du and Globe Fig. 2) de describes Paris, Sorbonne a rup- events with ML 2.4 to 4.0 were identi- Accepted article online 24 MAY 2014 fi 6 • Table S1 No additionalOn the slow order slip ofwas 6 detected to 9 m of on slip sitiesthe de reachedcitdata may MMI are VIII havetoo on sparse land, accumulated and to aParis tsunamiallow Cité, since,a Univ2unique m highParis 1877. hit inter-Diderot, coastal The UMR earthquaketure 7154 area inCNRS, the history deeper Paris, portion France. prior of thetoLaboratoire seismogenic de zone,Geologie, with a UM peakR8538 slip fied (Fig. 1, fig. S4, and table S1). On Publishedboth online along 6 JUN strike 2014 and downdip on the fault plane, updip of the main shock area. In addition, we found ©2014. American Geophysical Union. All Rights Reserved. fault immediately before the mainshock towns(7 ). inYAGI southernpretation ET AL. Peru of and the northern apparent Chile.CNRS Sixslip. Ecole fatalities Normale were attributed Superieure,of Paris,,8 m France. to the southeast of the hypocentre at depths of ,30–40 km. 115 March, the area was re-activated Department of Earth and Planetaryout Sciences, repeating University earthquakes of from the newly1877 detected is uncertain events, [likelyNishenko indicating, 1985; aseismicComte slip alongand Pardo the plate, 1991], so it is unclear whether the region regularly fails in Correspondence to: Bouchon et al. have recently suggested thatto the event, andThe at leastscarcity 13,000 homes of ocean were*Corresponding damaged floor or destroyed.author.measure- E- Pre-mail: [email protected] slip to the north is not well resolved but may account for the with 11 events of ML 2.6 to 4.6, fol- A. Kato,California, Santa Cruz, CA 95064, boundaryUSA. E-mail: fault brodsky@ during the foreshock sequence.huge single These ruptures observations or suggest intermittentlyliminary the occurrence estimates in larger suggestof multiple total ruptures economic then losses close sequences to US$100 million of smaller7. generation ruptures, of a local as tsunami. is the Slip case extended only ,50 km along the inter- fi lowed on 16 March by the first big [email protected]; [email protected] slow-slip events updip of the~70% main of shock interplate area. The earthquakesnal slow-slip are event preceded migratedA megathrust towardments the earthquakehighlights main shock in thisthe region second, was notpractical unexpected; bar- 230 face from the hypocentre, a very compact rupture area for an earth- along the Ecuador-Colombia coastline [Kanamori and McNallyThe subduction, 1982]. The zone rupture in Northern zone ofChile the15 is 1868 a well Peru identified seismic gap that last foreshock of Mw 6.7, about 50 km nucleation point. We interpret that several parts of the plate boundary fault perhapsM .3.5 earthquakes experienced occurred slow slip, offshore of Iquique between August 2013 quake of this size . The location of peak slip in this model is consistent ruptured in 1877. The M 8.1 Iquique earthquake of 1 April 2014 broke a highly South and with a slightly deeper cen- and March 2014, a 950% increase in the rate from January to July 2013w with W-phase CMT inversions (Extended Data Fig. 1), slab geometry Citation: causing stress loading on the prospectiveearthquake, largest slip patchwith of an the estimated main shock seismic rupture. magnitude of 8.5–8.8, partly reruptured in the 23 June 2011 Mw 8.4 troid (fig. S4). This event triggered a 700 16 MAY 2014 VOL 344 SCIENCE www.sciencemag.org(ref. 8). Over the three weeks before thecoupled event, there portion were more of than this 80 gap.(Extended To understand Data Fig. 2) andthe theseismicity centroid location preceding of an updated this event, CMT we Kato, A., and S. Nakagawa (2014), Peru earthquake (Figure 1), leaving an ~100 km long regionstudied offshore the location of southeastern and mechanisms Peru just of the north foreshocks of the and computed GPS time persistent precursory seismicity, includ- Multiple slow-slip events during a Published by AAAS earthquakes between M 4.0 and M 6.7 (Fig. 1). Before the recent sequence, (Methods). series at stations located on-shore. Seismicity off-shore Iquique started to increase ing an Mw 6.3 earthquake on 22 March, foreshock sequence of the 2014 Iquique, 1877 gap that may also havethis large subduction slip de zoneficit (between [e.g.,,Loveless19.5u S and et 21u al.S), had 2010]. been identified This model also matches inversions ofregionalgeodeticdata(Methods 3,4 which slowly moved northwards and Chile Mw 8.1 earthquake, Geophys. Res. 1. Introduction as a seismic gap , last rupturing in a Min, J8.8anuary earthquake 2014. in 1877. After The 16 Marchand Extended several Data M Fig.w 3),> 6 which events do not occurred uniquely resolve near slipthe up-dip low of Lett., 41, 5420–5427, doi:10.1002/ lasted until the occurrence of the 1 1 April event was followed by a vigorouscoupled aftershock zone. sequence These with more eventsthe migrated hypocentre butnorthward place strong for constraints about on50 thekm location until the and extent1 April on July 29, 2014 2014GL061138. The 1 April 2014 Iquique, Chile Mw 8.1 earthquake occurred at 23:46 (UTC) along a megathrust fault off On 16 March 2014, a Mw 6.7than thrust 100 M $ event4earthquakes,includinga occurred onearthquakeM or7.7 nearaftershock the occurred. near megathrust the south- On 16of slip aboutMarch between on 60km the-shore hypocentre north- cGPS and stations the coast, anddetected on its down-dip a westward edge. April 2014 Mw 8.1 event (Fig. 2 and fig. northernmost Chile [An et al., 2014; Laynorthwest et al., 2014], where of Iquique, the NazcaChile, plate is subductingernmost and was extent beneath followed of the M the8.2 South byrupture. two motionweeks that of thrustwe model aftershocks as aGeodetic slow thatslip models event slowly show situated that migrated slip during in the the mainshocksame area ended where west of the the S4). Received 7 JUL 2014 American plate at a convergence rate of ~8 cm/yr [Angermann et al., 1999]. TheThe plate seismic convergence moment has of all 2014 earthquakesmainshock to date occurred. equates to an coastline, inagreement with (perhaps slightly up-dip of) seismic models. Overall, these foreshocks delineate Accepted 15 JUL 2014 produced a series of large subduction(~20km/week) earthquakes along the northward Peru-Chile Trench alongevent (Figure of the just 1a).M megathrust, Northernmost8.3, much smaller from than 20.2°S the estimated to 19.6°S. size of the The 1877 locationTsunami models of this16 place sequence better constraint in on shallow slip, and show little a region that spans ~150 km along the Accepted article online 18 JUL 2014 5,6 Chile has been quiescent for 136 years since the 1877 M 8.6 earthquake, resultingearthquake in a current and of“North the potential Chile event that could fill the seismic gap motion up-dip and west of the hypocentre. strike of the subduction zone. We relo- Published online 1 AUG 2014 the northern portion of the 1877 seismic gap focused attentionSince the giant on M the 8.8 region, mega-thrust and earthquake on 1 April of 1877 2014, (1±4 a WKHMw seismic gap” stretching for ~450 km length along the strike of the plate boundary(Methods). [e.g., Khazaradze The earthquake and Klotz sequence, spans a section of thew subduction Earthquakes in this sequence were relocated bycated using athis multiple-event, precursory seismicity and computed regional seismic moment 8.1 interplate thrust earthquakezone about initiated one-third at of the the size n oforthern the inferredkm long end 1877 region rupture of thestretching9. It foreshock remains from hypocentroidalArica sequence (18°S) to decomposition the (19.642°S, Mejillones17 algorithm, Peninsula using seismictensor phaseusing dataa linear from time-domain, broad-band waveform inverse method 2003; Chlieh et al., 2004]. Recent geodetic measurements show that the North Chileunknown seismic how gap subduction comprises zones behave(24.5°S, over multiple Fig. 1) seismichas had cycles relativelylocal, few regional major and seismic global stations.events with This only allows us to interpret locations (15) (Fig. 2). The centroid of the Mw 6.7 precursor of 16 March was only at least two highly locked segments bounded70.817°W, by narrow 23:46:46 areas of UTC relatively [U.S. weak Geological coupling offshore Survey from (USGS) National Earthquake Information Center (NEIC): 2,18 and whether any given section can be associatedmoderate M withw < a 8 characteristic events in 1933,within 1967, a and consistent, 2007 (4 regionally±9). Based anchored on recent and absolute10 km deep, framework in an area. where the seismogenic interface is at about 20 km www.sciencemag.org Iquique [Métois et al., 2013; Béjar-Pizarrohttp://earthquake.usgs.gov/regio et al., 2013]. The 2014 Iquique, Chileearthquake, earthquakenal/neic/]). making occurred it unclear The in global whethergeodetic thiscentroid seismic data gap mome(5 should, 10, 11nt behave), tensorthe degreeBeginning (gCMT) of interseismic on16 solution March coupling with aforM 6.7² thisthe earthquake, ratio thedepth foreshock (16) (Fig. sequence 2 and fig. S5). This event had a reverse focal mechanism this area of relatively weak coupling within the seismic gap, where the ratein of the background twenty-first seismicity century as it is did in thebetween nineteenth. the Observationsinterseismic sug-slip rategenerated and the plate more convergence than 80 M $4 velocity earthquakes²in (Fig.with 2), showinga strike aof north- 277° that is at a sharp angle with respect to the trench event [http://www.globalcmt.org/CMTsearch.htmgest that enough strain has accumulatedl] indicates along this an plate almost boundary purely seg- ward double-couple migration towards faultingtheepicentreofthe1AprilM 8.2 event (Extended higher than in the surrounding highly locked segments. the area shows two distinct highly-coupled segments (Loa and Cama- (Fig. 2A). During a week, a persistent seismicity occurred in a zone of ment to host an earthquake close to M 9 (see, for example, ref. 5), and Data Fig. 4). Over the following week, the NEIC recorded 140 M $4after- geometry with strike 357°, dip 18°, and rake 109° at arones).centroid They are depth separated of 21.9kmby a low coupling and centroid zone (LCZ) location offshore Iqui- 10 km radius (Fig. 2C), most of these events were located inside the According to the U.S. Geological Survey (USGS) catalog (USGS National Earthquakeearthquakes Information of this size Center: have occurred in the past. The expectation from shocks, including a M 7.7 event on 3 April, 27 h after the mainshock. south of the hypocenter (19.77°S, 70.98°W), with a centroidque (5). On time 1 April shift 2014, of a 42.5s ~150 km and long seismic portion of moment the gap broke of in an shallow South-American plate and had very diverse focal mechanisms. http://earthquake.usgs.gov/regional/neic/), the main shock was preceded by1 intensiveNational Earthquake foreshock Information sequence Center, United StatesM Geological 8.1 Survey,earthquake Golden, Colorado after 80401,a strong USA. 2 Departmentprecursory of Geosciences, activity Pennsylvania that started State University, on 16 University Park, Pennsylvania 16802, w On 22 March a new Mw 6.3 foreshock occurred about 30 km north of the 21 3 ´ 4 5 lasting around 15 days (Figure 1b) [Brodsky1.69 and× 10 Lay, 2014;Nm(LayM etw al8.1)., 2014; (FigureYagiUSA. etCentro al1).., Sismolo 2014]. A substantialgico The Nacional, largest Universidad aftersho de Chile,March Blancock Encalada(12 sequence, 13 2002,). Understanding Santiago ensued, 8370449, Chile. the theGlobal complex Seismological largest nucleation Services, of which Golden, phase Colorado preceding 80401, USA. Canada16 March Centre forforeshock. Mapping and After this event precursory seismicity moved to the Earth Observation, Natural Resources Canada, Ottawa, Ontario K1A 0E4, Canada. foreshock (Mw 6.7), which had a focal mechanism similar to that of the main shock, occurred on 16 March the Mw 8.1 mainshock and its coseismic rupture should provide insights vicinity of the 22 March foreshock with depths and mechanisms indicat- Downloaded from occurred on 3 April 2014 with Mw 7.7 (02:43:14 UTC, 20.518°S, 70.498°W, centroid depth 31.3 km), 49 km 2014 at ~40 km south of the main shock hypocenter. The foreshock sequence in the USGS catalog shows a into the seismic gap history, present-day seismic hazard,21 AUGUST and nucleation 2014 | VOL ing 512 that | NATURE it occurred | 295 at the plate interface (Fig. 2B). migration or movement toward the mainsouthwest shock hypocenter of Iquique. after 16 The March, region and the extending foreshocks were from updipprocess of©2014 the ofMacmillan megathrust 3 April Publishers earthquakes. event Limited. southward All rights reserved to ~23°S, updip From 16 March until the 1 April mainshock (i.e., for 17 days), all the located outside of the largest slip patch of the main shock rupture (Figure 1b). Such a spatiotemporal Seismic activity preceding the Iquique earthquake initiated in a re- cGPS stations located along the coast between Iquique and Pisagua of the 2007 Mw 7.7 Tocopilla earthquake rupture zonegion [e.g., betweenBéjar-Pizarro 19.5°S and et21°S, al. ,where 2010; couplingSchurr ranges et al. from, 2012], 0.2 to 0.5 evolution of the foreshock sequence provides a unique opportunity to investigate the nucleation process started to move trench ward (Fig. 3). This slowly increasing westward (5) ; that is the plates are slowly creeping past each other at a fraction of of the main shock rupture [e.g., Bouchonremains et al., 2011; strainedKato et aland., 2012; hasBouchon potential et al., for 2013]. either However, an the ~Mw 8.5 event or several smaller great events (Figure 2). motion contrasts with the usual inland-directed inter-seismic motion. plate rate [when the plates are fully locked, coupling is 1.0]. This region The displacements measured during this period were quite large (more detailed evolution of the foreshock sequence is unclear and poorly understood, as many of the seismic events was actively monitored because its seismic activity had steadily in- than 5 mm for the stations located between Iquique and Pisagua and are likely missing from the existing USGS catalog. Such missed events arise because small-magnitude creased since 2008 with repeated interplate thrust events of magnitude about a cm at the PSGA station). Only a fraction of this cumulative dis- ©2014. American Geophysical Union. All Rights Reserved. LAY ETearthquakes AL. are masked by overlapping arrivals from many earthquakes, or because successive earthquakes smaller than 4.0. Several seismic clusters were located in this3818 region by placement (up to 20%) can be attributed to the largest foreshock of 16 occur over short time intervals [e.g., Enescu et al., 2007; Peng and Zhao, 2009; Kato et al., 2013]. the National Seismological Center of Chile (CSN) (Fig. 1 and fig. S1); March (Mw 6.7), suggesting that slow aseismic slip was taking place some of these clusters were associated with persistent micro-seismicity To obtain a precise record of the foreshock sequence before the 2014 Iquique, Chile earthquake, we applied a offshore, concurrently with the development of the precursory seismici- observed by the nearest seismological stations (fig. S2). Global catalogs ty. Whether this motion started slightly before, or coincided with, the 16 cross-correlation detector technique [e.g., Kato et al., 2013] to continuous seismograms recorded near the reveal an increase in seismicity near Iquique since 2005, compared with March fore-shock is beyond the current GPS resolution. We inverted for source region. The newly detected events clearly illustrate multiple sequences of earthquake migrations the previous 10 years (fig. S3). In order to understand the seismicity that the slip distributions on the subduction interface that best reproduce the during the foreshock sequence, both along strike and downdip on the fault plane. In addition, we extracted preceded the 1 April mainshock we used the events listed in the CSN observed displacements using Okada's formulas for an elastic half-space catalog to relocalize and estimate the focal mechanism of several fore- (14) (Fig. 4). We found a slip of ~0.8 m for the Mw 6.7 events of 16 shocks of this sequence. Simultaneously, we calculated the GPS time March, located in a narrow area in the vicinity of the CSN epicenter. We ©2014. American Geophysical Union. All Rights Reserved. KATO AND NAKAGAWA 5420 series of the closest permanent stations up to the date of the mainshock then computed the aseismic slip for the period from 10 March to the and inverted for the slip rate on the plate interface (14). mainshock, that extends over an area of 70 km ! 20 km located between The most recent seismic activity in northern Chile started on 4 Janu- the fore-shocks and the coast (Fig. 4).

/ http://www.sciencemag.org/content/early/recent / 24 July 2014 / Page 1 / 10.1126/science.1256074