Title: From the seismic cycle to long-term deformation: linking seismic coupling and Quaternary coastal geomorphology along the Andean megathrust
Abstract: TectonicsVolume 36, Issue 2 p. 241-256 Research ArticleFree Access From the seismic cycle to long-term deformation: linking seismic coupling and Quaternary coastal geomorphology along the Andean megathrust M. Saillard, Corresponding Author M. Saillard [email protected] orcid.org/0000-0001-5953-2640 Université Côte d'Azur, IRD, CNRS, Observatoire de la Côte d'Azur, Géoazur, Valbonne, France Correspondence to: M. Saillard, [email protected] for more papers by this authorL. Audin, L. Audin IRD, CNRS, Université Grenoble Alpes, ISTerre, Grenoble, FranceSearch for more papers by this authorB. Rousset, B. Rousset orcid.org/0000-0001-9304-0498 IRD, CNRS, Université Grenoble Alpes, ISTerre, Grenoble, FranceSearch for more papers by this authorJ.-P. Avouac, J.-P. Avouac orcid.org/0000-0002-3060-8442 Tectonics Observatory, Division of Geological and Planetary Sciences, California Institute of Technology, Pasadena, California, USASearch for more papers by this authorM. Chlieh, M. Chlieh orcid.org/0000-0003-2252-2187 Université Côte d'Azur, IRD, CNRS, Observatoire de la Côte d'Azur, Géoazur, Valbonne, FranceSearch for more papers by this authorS. R. Hall, S. R. Hall orcid.org/0000-0001-8258-4211 Earth Sciences Department, College of the Atlantic, Bar Harbor, Maine, USASearch for more papers by this authorL. Husson, L. Husson IRD, CNRS, Université Grenoble Alpes, ISTerre, Grenoble, FranceSearch for more papers by this authorD. L. Farber, D. L. Farber Earth and Planetary Sciences Department, University of California, Santa Cruz, California, USA Lawrence Livermore National Laboratory, Livermore, California, USASearch for more papers by this author M. Saillard, Corresponding Author M. Saillard [email protected] orcid.org/0000-0001-5953-2640 Université Côte d'Azur, IRD, CNRS, Observatoire de la Côte d'Azur, Géoazur, Valbonne, France Correspondence to: M. Saillard, [email protected] for more papers by this authorL. Audin, L. Audin IRD, CNRS, Université Grenoble Alpes, ISTerre, Grenoble, FranceSearch for more papers by this authorB. Rousset, B. Rousset orcid.org/0000-0001-9304-0498 IRD, CNRS, Université Grenoble Alpes, ISTerre, Grenoble, FranceSearch for more papers by this authorJ.-P. Avouac, J.-P. Avouac orcid.org/0000-0002-3060-8442 Tectonics Observatory, Division of Geological and Planetary Sciences, California Institute of Technology, Pasadena, California, USASearch for more papers by this authorM. Chlieh, M. Chlieh orcid.org/0000-0003-2252-2187 Université Côte d'Azur, IRD, CNRS, Observatoire de la Côte d'Azur, Géoazur, Valbonne, FranceSearch for more papers by this authorS. R. Hall, S. R. Hall orcid.org/0000-0001-8258-4211 Earth Sciences Department, College of the Atlantic, Bar Harbor, Maine, USASearch for more papers by this authorL. Husson, L. Husson IRD, CNRS, Université Grenoble Alpes, ISTerre, Grenoble, FranceSearch for more papers by this authorD. L. Farber, D. L. Farber Earth and Planetary Sciences Department, University of California, Santa Cruz, California, USA Lawrence Livermore National Laboratory, Livermore, California, USASearch for more papers by this author First published: 25 January 2017 https://doi.org/10.1002/2016TC004156Citations: 61AboutSectionsPDF ToolsRequest permissionExport citationAdd to favoritesTrack citation ShareShare Give accessShare full text accessShare full-text accessPlease review our Terms and Conditions of Use and check box below to share full-text version of article.I have read and accept the Wiley Online Library Terms and Conditions of UseShareable LinkUse the link below to share a full-text version of this article with your friends and colleagues. Learn more.Copy URL Share a linkShare onFacebookTwitterLinkedInRedditWechat Abstract Measurement of interseismic strain along subduction zones reveals the location of both locked asperities, which might rupture during megathrust earthquakes, and creeping zones, which tend to arrest such seismic ruptures. The heterogeneous pattern of interseismic coupling presumably relates to spatial variations of frictional properties along the subduction interface and may also show up in the fore-arc morphology. To investigate this hypothesis, we compiled information on the extent of earthquake ruptures for the last 500 years and uplift rates derived from dated marine terraces along the South American coastline from central Peru to southern Chile. We additionally calculated a new interseismic coupling model for that same area based on a compilation of GPS data. We show that the coastline geometry, characterized by the distance between the coast and the trench; the latitudinal variations of long-term uplift rates; and the spatial pattern of interseismic coupling are correlated. Zones of faster and long-term permanent coastal uplift, evidenced by uplifted marine terraces, coincide with peninsulas and also with areas of creep on the megathrust where slip is mostly aseismic and tend to arrest seismic ruptures. We conclude that spatial variations of frictional properties along the megathrust dictate the tectono-geomorphological evolution of the coastal zone and the extent of seismic ruptures along strike. Key Points Peninsulas are seismic barriers and show locally high Quaternary uplift rates (>0.4 m/ka) They tend to form above aseismically sliding areas and are permanent over multiple seismic cycles These dynamic features reflect anelastic deformation of the fore arc in response to the subduction seismic cycle 1 Introduction Over the last few decades, a number of M > 8.5 earthquakes occurred at locations along subduction zones where only much smaller earthquakes had been documented previously [Ammon et al., 2005; Lay, 2015]. This pattern suggests that a portion of a subduction zone can rupture either in more frequent smaller or, plausibly, less frequent larger earthquakes [e.g., Kanamori and McNally, 1982; Thatcher, 1990]. The observation is particularly evident in the historical record of earthquakes along the Peru-Chile megathrust, where the Nazca plate is subducting beneath South America (Figure 1). Despite along strike variability, stationarity seems to characterize the location of seismic rupture limits, suggesting that the subduction zone along the Peru-Chile margin is characterized by "segmentation" [Ando, 1975]. Segmentation is also shown in geodetic measurements of interseismic and postseismic strains in South America. As observed in other subduction zones, seismic ruptures (1) tend to correlate with asperity patches that are highly locked during the interseismic period and (2) tend to arrest at patches where slip is mostly aseismic, either due to interseismic or postseismic creep [e.g., Chlieh et al., 2008; Perfettini et al., 2010; Moreno et al., 2010; Métois et al., 2012; Protti et al., 2014]. This correlation is interpreted to result from the spatial variation of megathrust friction along and across strike [Burgmann et al., 2005; Kaneko et al., 2010; Hetland and Simons, 2010; Cubas et al., 2013; Schurr et al., 2012]. Figure 1Open in figure viewerPowerPoint Marine terrace uplift rates, trench-coast distance, and rupture length of historical earthquakes along the Andean margin. (a) Uplift rates of marine terraces reported in the literature (we present the average rate since terrace abandonment; Table S1 in the supporting information [Jara-Muñoz et al., 2015]). Each color corresponds to a marine terrace assigned to a marine isotopic stage (MIS). Gray dots are the uplift rates of the central Andean rasa estimated from a numerical model of landscape evolution [Melnick, 2016]. (b) The distance between the coast and the trench was measured parallel to the convergence direction [DeMets et al., 1994]. Main peninsulas are indicated with names and arrows. Horizontal blue bands are the areas where the coastline is less than 110 km (light blue) or 90 km (dark blue) from the trench. (c) Lateral extent of the rupture zone of historical megathrust earthquakes are color coded by magnitude from southern Chile to central Peru (reported in Table S2). Continuous lines indicate the rupture zones better constrained than those represented by dashed lines. (d) Geodynamic setting of the Andean margin (10°S–40°S) and location of major great historical megathrust earthquakes. Major bathymetric features, the coastline (blue line), and the Peru-Chile trench (thick black line) are indicated. Convergence directions and velocities (cm/yr) of the Nazca plate toward the South America plate are from DeMets et al. [1994]. Red line corresponds to the 40 km isodepth of the subducting slab [Hayes and Wald, 2009]. Spatial variations in fore-arc morphology also bear information regarding the frictional properties along the subduction interface that have been observed to correlate with the extent of megathrust ruptures [e.g., Marshall and Anderson, 1995; von Huene and Klaeschen, 1999; Fuller et al., 2006; Audin et al., 2008; Rehak et al., 2008; Rosenau and Oncken, 2009; Bejar-Pizarro et al., 2013; Cubas et al., 2013; van Dinther et al., 2013]. The margins of South America and Japan provide examples of this relationship between the fore-arc morphology and megathrust ruptures as in both locations the coastline seems to mark the downdip limit of the seismogenic portion of the megathrust [Ruff and Tichelaar, 1996; Khazaradze and Klotz, 2003]. During the 1960 Chile earthquake, most of the slip occurred offshore [e.g., Moreno et al., 2011]. Further, during the 2007 Pisco earthquake the locally contorted coastline separates the uplifted, offshore area from the subsided, onshore area, indicating that the rupture occurred updip of the coastline [Sladen et al., 2010]. This relationship is consistent with the observation that the downdip edge of the locked portion of the megathrust determined from modeling of interseismic strain follows the coastline quite closely [Chlieh et al., 2004; Bejar-Pizarro et al., 2013]. Another observation linking fore-arc morphology to the spatial occurrences of megathrust earthquake ruptures is that the ruptures tend to correlate with fore-arc basins and arrest at locations where there are trench-perpendicular gravity highs that correspond to local topographic fore-arc highs [Mogi, 1969; Nishenko and McCann, 1979; Song and Simons, 2003; Wells et al., 2003; Llenos and McGuire, 2007]. Indeed, along the coasts of Japan, Alaska, and South America, seismic ruptures tend to stop close to peninsulas or coastal promontories [Wells et al., 2003; Melnick et al., 2009; Cubas et al., 2013]. Taken together, these observations suggest that the morphology of fore arcs, and the location and shape of the coastline in particular, could be used to assess megathrust segmentation during large interplate earthquakes. In this study, we test the hypothesis that the coastline geometry and kinematics (uplift and subsidence) reflect the spatial variations of frictional properties of the megathrust and, in turn, the seismic segmentation of the megathrust offshore South America. In this study, we assess correlations among three types of quantitative data from central Peru to southern Chile (between ~12°S and ~40°S): (1) coastal uplift rates derived from the elevation of dated marine terraces (Figure 2 and Table S1 in the supporting information); (2) the distance between the trench and the coast in the convergence direction; and (3) the pattern of interseismic coupling, a quantity defined as the ratio of the deficit of slip rate during the interseismic period divided by the long-term slip rate [cf. Avouac, 2015]. This ratio quantifies the degree of locking along the plate interface. A null interseismic coupling means that the subduction interface is creeping at the long-term slip rate (no slip deficit is building up). An interseismic coupling of 1 means that the subduction interface is fully locked. We compare the spatial variations of these three data sets with the latitudinal distribution of historical earthquake rupture zones for the last 500 years (see Table S2 for references). Figure 2Open in figure viewerPowerPoint Panorama view of a stair-cased sequence of at least three marine terrace levels uplifted at ~55, ~200, and ~480 m of elevation at Punta Villa Señor (Chile, ~30.5°S). The oldest level is at the highest elevation and the youngest at the lowest elevation (photo by M. Saillard). 2 Long-Term Coastal Dynamics of the Andean Margin Along the western Andean margin, the Nazca plate is subducting eastward beneath South America at a rate varying from 7.7 in southern Peru to 8.1 cm/yr in southern Chile, respectively [DeMets et al., 1994] (Figure 1). A number of geomorphological studies have characterized coastal uplift from marine terraces (see Table S1 for references), which provide reliable long-term records of past sea level highstands (odd marine isotopic stage (MIS); Figure 2) [Lajoie, 1986; Chappell and Shackleton, 1986; Pedoja et al., 2011]. As present-day sea level is higher than the highest levels recorded since the early Pleistocene, any Pleistocene coastal features standing above the present sea level must reflect tectonic uplift [Lajoie, 1986]. Stair-cased marine terrace sequences are preserved discontinuously along the coast of Peru and Chile (Figure 1). Well-preserved wave-cut platforms are more commonly found on peninsulas and promontories, in particular at San Juan de Marcona, Ilo, Mejillones, Tongoy, and Arauco (Figure 1). Built marine terraces are also found in bays where the wave-cut planation surface is covered by deposits due to lower wave energy [Trenhaile, 1987, 2000; Sunamura, 1992; Bloom, 1998]. Marine terraces typically date from about 800 ka to <10 ka [Regard et al., 2010] (see Figure 1a and Table S1 for references). Recent efforts in surveying and dating these coastal features using in situ produced cosmogenic nuclide exposure dating have provided robust constraints on the rates and patterns of local coastal uplift along most of the coast of Peru and Chile [Marquardt et al., 2004; Quezada et al., 2007; Saillard, 2008; Melnick et al., 2009; Saillard et al., 2009, 2011; Cortès et al., 2012; Jara-Muñoz et al., 2015]. These data reveal spatiotemporally variable uplift rates along and across strike during the last million years [Saillard et al., 2009, 2012] (Figure 1a). Mean uplift rates are moderate to high and range between 0.1 and 1.7 m/ka (Figure 1a). The higher number and elevation of preserved marine terraces on peninsulas (i.e., San Juan de Marcona (~15.3°S), Mejillones (~23.5°S), and Arauco Peninsulas (~37.5°S)) suggest faster active uplift of these areas than other locations along the coastal fore arc (e.g., between Mejillones and Caldera; Figure 1). Where uplift has been more subdued (or potentially periods of subsidence), the individual terrace levels are more intermingled (less distinct) and often coalesce to form a wide, gently sloping oceanward wave-cut planation surface [Regard et al., 2010]. This typical "rasa" landform called the central Andean rasa lies at a relatively uniform elevation of ~110 m along the coast from 15°S to 33°S. This relatively simple cliff-face coastal morphology is interrupted by peninsulas such as Mejillones Peninsula, where some of the best preserved marine terraces sequences exist [Regard et al., 2010]. Calculated uplift rates of the central Andean rasa from a review of published chronological data of the central Andean rasa are about 0.25–0.3 m/ka [Regard et al., 2010]. This study argues that the whole Andean fore arc, except the peninsulas, has been uplifted relatively continuously for at least 400 ka (MIS 11), after a Pliocene period of quiescence or subsidence, associated with active Quaternary deformation [Hall et al., 2008; Regard et al., 2010; Rodriguez et al., 2013]. From a landscape evolution model simulating coastal erosion and formation of the central Andean rasa, Melnick [2016] estimated Quaternary coastal uplift rates of 0.13 ± 0.04 m/ka along the central Andean fore arc except in the peninsula areas. Coastal uplift rates calculated from marine terrace sequences in peninsula areas are up to about 0.7–0.8 m/ka and even locally up to 1.7 m/ka (Figure 1) [Saillard et al., 2009, 2011; Jara-Muñoz et al., 2015]. Modeled uplift rates derived by Melnick [2016] for the central Andean rasa are lower than uplift rates calculated by Regard et al. [2010] for the same rasa and are much lower than those derived in peninsula areas where the rasa is replaced by distinct marine terraces (Figure 1). As marine terraces are morphotectonic markers of past sea level variations and formed during past sea level highstands, individual marine terraces are assigned to a sea level highstand based on their measured exposure age and associated error (analytical and production rate errors) in the case of terrestrial cosmogenic dating. The age of the corresponding sea level highstand is then used to calculate an uplift rate with the uncertainty coming from the measurement of the terrace elevation and the chronology of the sea level curve [Siddall et al., 2006]. In contrast, Melnick [2016] models the topographic evolution of the coastal region given different uplift scenarios, not based on any direct measurements of uplift or chronometric markers, making this model strongly dependent on the initial topographic modeling of the shelf slope. While the uplift rates generated by this model are generally the same order of magnitude for many coastal regions, it fails to yield the directly measured uplift rates near peninsula areas. Additionally, uplift rates in peninsula areas are (1) variable from one MIS to another and (2) higher at the center of the peninsula than on both flanks (Figure 3). In particular, near Arauco, uplift rates symmetrically vary across the peninsula (Figure 3c). The absence of preserved coastal landforms in some locations suggests that coastal subsidence or very slow uplift could prevail in these regions. The coastal region is subsiding in central Peru [Le Roux et al., 2000; Hampel, 2002] (Figure 1) while very slowly uplifting in the Arica Bend [García et al., 2011] and along the coastal fore arc between the Arica Bend and the Mejillones Peninsula with the slow development of an ~1000 m high coastal scarp. Figure 3Open in figure viewerPowerPoint Uplift rates of marine terraces versus latitude in three peninsula areas. Each color corresponds to a marine terrace assigned to a marine isotopic stage (MIS). Zoom views of the three peninsulas illustrate the local N-S variability of marine terrace chronologies showing higher rates closer to the trench (i.e., near the center of the peninsula). (a) Mejillones Peninsula data set after Victor et al. [2011] (squares) and Regard et al. [2010] (triangles), and references therein. (b) Elevations of marine terraces in Tongoy Peninsula extracted from SRTM DEM and ages from Saillard et al. [2009]. (c) Maule area data set: Arauco Peninsula after Melnick et al. [2009] (circles) and Carranza and Topocalma Peninsulas after Jara-Muñoz et al. [2015] (squares). Diamonds correspond to the uplift rates from 10Be dating of marine terraces by Melnick et al. [2009]. In this study, we present a new data set, the Tongoy Peninsula data set (Figure 3b), that we compare to two available data sets (to the north (Figure 3a) and to the south (Figure 3c)). Along the Andean margin, marine terraces are preserved where coastal uplift has occurred, either as embayments (built marine terraces) or peninsulas (wave-cut platforms), regardless of the resistance to coastal erosion [Saillard, 2008]. While the trench-coast distance varies along the margin from ~50 km to more than 200 km, we observe that marine terraces are chiefly present and are being uplifted more rapidly than the rest of the coastal zone where the trench-coast distance is less than 110 km (Figures 1a and 1b). This pattern suggests a relationship between marine terrace formation and deeper processes related to slab subduction and the downdip extension of the seismogenic zone. 3 GPS-Derived Interseismic Coupling Model of the Andean Margin Interseismic coupling (ISC) is defined as (1)where (Vo − Vcreep) corresponds to the slip rate deficit and Vcreep to the creeping rate on the megathrust during the interseismic period. Vo is the long-term slip rate imposed by plate motion, in this case the Nazca/South America convergence rate taken from Kendrick et al. [2003]. An ISC of 0 would indicate a fully creeping megathrust (Vcreep = Vo), while an ISC of 1 would correspond to a full locking (Vcreep = 0). To retrieve the interseismic coupling pattern along the megathrust interface from the geodetic measurements, we define a megathrust geometry defined by trench axis on the seafloor and the average local dip of the subducting slab derived from the Slab 1.0 model [Hayes et al., 2012]. The megathrust is meshed into 20 × 20 km2 square elements embedded in an elastic half-space. The response of a finite fault at a specific location is expressed by summing the contributions of a regular grid of subfaults as (2)where u is the displacement at an arbitrary station, i is the ith subfault along strike, and j is the jth subfault along dip. Sij and Rij are the dislocation slip amplitude and rake angle. The terms Gsij and Gdij are the subfault Green's functions for a unit slip in the strike and dip directions. With this approach, a full representation of the fault response relies on two parameters: the dislocation amplitude and the rake angle that can be inverted by matching the modeled static displacements to the observed GPS measurements. The general forward problem is written as (3)where d represents the theoretical data values, G represents the Green function linking the observables to the model, and m are the parameters describing the model. The static displacements are computed following the equation of Okada [1992] with an average shear modulus of 40 GPa and a Poisson coefficient of 0.25. Using the back slip approach [Savage, 1983], we perform nonlinear inversions of the GPS data based on a simulated annealing algorithm to determine the interseismic coupling distribution [Chlieh et al., 2011]. The misfit between the observations and model predictions is computed using a classical weighted root-mean-square of the residuals (wrms) criterion. The cost function to minimize is defined as the summation of the weighted quadratic summation of the misfit to the data (wrms) and another term meant to control the roughness of the back slip distribution: (4)where Dc represents the differences in back slip rate between adjacent cells and λ1 control the roughness through a L1 + L2 norm [Ji et al., 2002; Chlieh et al., 2014]. We first search for the optimal smoothing factors by varying λ1 from 0.01 to 10. Figure 4b shows one of the best fitting interseismic models for λ1 = 0.5. The average misfit (wrms) for this model is 3.2 mm/yr. The GPS residuals are oriented in all directions and remain globally within their uncertainties (Figure S1 in the supporting information). A spatial resolution test indicates that the along-strike resolution is relatively high everywhere but drops significantly along-dip especially near the trench. There the boundary condition (no coupling) results in little coupling except where required by the data. The poorly resolved patches appear in coastal areas with sparse GPS measurements or in areas where the trench-coast distance is high, typically higher than 150 km (Figure S2). To avoid overinterpretation of the interseismic coupling distribution, we indicate the poorly resolved areas with gray shading in the coupling map (Figure 4b). We observe that areas where the coast is closer to the trench tend to have low coupling and are also relatively well resolved and infer that these areas of low coupling are a robust feature of our inversions. Figure 4Open in figure viewerPowerPoint Comparison between the uplift rates, interseismic coupling, major bathymetric features, and peninsulas along the Andean margin (10°S–40°S). (a) Uplift rates of marine terraces reported in the literature (we present the average rate since terrace abandonment; Table S1 in the supporting information [Jara-Muñoz et al., 2015]). Each color corresponds to a marine terrace assigned to a marine isotopic stage (MIS). Gray dots are the uplift rates of the central Andean rasa estimated from a numerical model of landscape evolution [Melnick, 2016]. (b) Major bathymetric features and peninsulas and pattern of interseismic coupling of the Andean margin from GPS data inversion (this study). Gray shaded areas correspond to the areas where the spatial resolution of inversion is low due to the poor density of GPS observations (see text and supporting information for more details). The Peru-Chile trench (thick black line), the coastline (thin black line), and the convergence direction (black arrows) are indicated. We superimposed the curve obtained by shifting the trench geometry eastward by 110 km (trench-coast distance of 110 km; blue line) with the curve reflecting the 40 km isodepth of the subducting slab (red line; Slab1.0 from Hayes and Wald [2009]), a depth which corresponds approximately with the downdip end of the locked portion of the Andean seismogenic zone (±10 km) [Ruff and Tichelaar, 1996; Khazaradze and Klotz, 2003; Chlieh et al., 2011; Ruegg et al., 2009; Moreno et al., 2011; Métois et al., 2012]. The two curves are spatially similar in the erosive part of the Chile margin (north of 34°S), whereas they diverge along the shallower slab geometry in the accretionary part of the Chile margin (south of 34°S), where the downdip end of the locked zone may be shallower (Figure 4b). Red arrows indicate the low interseismic coupling associated with peninsulas and marine terraces and evidence of aseismic afterslip (after Perfettini et al. [2010] below the Pisco-Nazca Peninsula; Pritchard and Simons [2006], Victor et al. [2011], Shirzaei et al. [2012], Bejar-Pizarro et al. [2013], and Métois et al. [2013] for the Mejillones Peninsula; Métois et al. [2012, 2014] below the Tongoy Peninsula; and Métois et al. [2012] and Lin et al. [2013] for the Arauco Peninsula). FZ: Fracture zone. Horizontal blue bands are the areas where coastline is less than 110 km (light blue) or 90 km (dark blue) from the trench (see Figure 1). 4 Permanent Coastal Uplift Compared to Interseismic Coupling While the depth of the downdip end of the locked zone may vary from margin to margin worldwide [e.g., Dixon and Moore, 2007], in South America, the 110 km trench-coast distance not only seems to be a threshold distance within which the coastal zone is being uplifted more rapidly and where marine terraces are best preserved. These areas often coincide with peninsulas (Figure 1). But also, the 110 km trench-coast distance curve seems to mark the downdip end of the locked portion of the Andean seismogenic zone (Figure 4b). We compare Quaternary coastal uplift rates with both the location of major peninsulas and bathymetric features as well as the pattern of interseismic coupling in this fore-arc region (Figure 4). Interseismic coupling varies widely along the coast: it is particularly high (i.e., >0.6) in central Peru, north of the Pisco-Nazca Peninsula (north of ~14°S), south of the Mejillones Peninsula (between ~24°S and ~27°S), and between the Tongoy Peninsula and the Arauco Peninsula (between ~32°S and 37°S). In contrast, it is low (i.e., <0.4) beneath the Pisco-Nazca Peninsula (14–15°S), Mejillones Peninsula (~23°S), Punta Choros-Chañaral Peninsula (~29°S), Tongoy Peninsula (~30°S), and the southern part of Arauco Peninsula (~37.5°S) (Figure 4). In order to investigate further the relationship between coupling and the morphology of the coastline, we compare the trench-coast distance with the lateral variations of the coupling model (Figure 5). The distance between the trench and topography is taken parallel to the measured convergence direction (Euler pole from NUVEL-1A; [DeMets et al., 1994]) because if ongoing subduction of both plates impacts the long-term morphology of the overriding plate, it will be in the direction of convergence. Coupling is integrated downdip in the direction of convergence. Because we anticipate that coupling variations are unrelated to the trench-coast distance at very long (>500 km) and very short (<100 km) wavelengths, we band passed the signals between these two values. In particular, the orientation of the coast of southern Peru, between Arica and San Juan de Marcona, is highly oblique to the direction of convergence over ~800 km. Therefore, only for geometrical reasons, this obliquity increases the trench-coast distance as calculated by us (in the direction of convergence) irrespective of the coupling. At the other end of the spectrum, wavelengths shorter than 100 km likely mirror local structural heterogeneities rather than coupling properties. Figure 5Open in figure viewerPowerPoint Comparison between short-term, GPS-derived interseismic coupling (red) and trench-coast distances (blue) integrated along the Benioff zone in the convergence direction (Euler pole from NUVEL-1A [DeMets et al., 1994]). Both signals have been band passed for wavelengths ranging between 100 km and 500 km (including the mean values and therefore represent the departures from average). Dashed lines show the high-pass filtered signal, leaving all wavelengths shorter than 500 km. Coupling is integrated downdip along the direction of convergence and projected on the slab geometry derived by Hayes et al. [2012]. Black arrows indicate the location of the main peninsulas. Color bars indicate the shortest width pairs (lateral shift) of minima (peninsulas and low coupling areas, yellow; embayments and high coupling areas, gray). Within the prescribed range, th