Subducting seamounts control interplate coupling and seismic rupture in the 2014 Iquique earthquake area

To date, the parameters that determine the rupture area of great subduction zone earthquakes remain contentious. On 1 April 2014, the Mw 8.1 Iquique earthquake ruptured a portion of the well-recognized northern Chile seismic gap but left large highly coupled areas un-ruptured. Marine seismic reflection and swath bathymetric data indicate that structural variations in the subducting Nazca Plate control regional-scale plate-coupling variations, and the limited extent of the 2014 earthquake. Several under-thrusting seamounts correlate to the southward and up-dip arrest of seismic rupture during the 2014 Iquique earthquake, thus supporting a causal link. By fracturing of the overriding plate, the subducting seamounts are likely further responsible for reduced plate-coupling in the shallow subduction zone and in a lowly coupled region around 20.5°S. Our data support that structural variations in the lower plate influence coupling and seismic rupture offshore Northern Chile, whereas the structure of the upper plate plays a minor role.

T he Chilean subduction zone can be divided into different seismo-tectonic segments that are known to have repeatedly ruptured over the last centuries during great and giant megathrust earthquakes every B100-150 years 1 . This pattern of recurring earthquakes defines the northern Chile seismic gap (B18°S-24°S; Fig. 1a) with a last great event about 137 years ago (BMw 8.6 Iquique earthquake in 1877) 1 . The anticipation of an impending great earthquake has motivated numerous international field campaigns onshore, making the area one of the best-studied seismic gaps worldwide. Along its B550 km, the seismic gap is characterized by a heterogeneous plate-coupling distribution (Fig. 1b) [2][3][4] . South of 21°S a highly coupled patch is presented with moderate to high coupling (40.5) extending into the shallow-most part of the subduction zone. In contrast, in the central part of the seismic gap (19)(20).5°S) high coupling is only met in a small patch (460 km landward from the deformation front) and decreases to values o0.5 towards the trench. The two patches are separated by a lowly coupled region around 20.5°S. North of 19°S coupling is less well constrained due to the large distance of the coastline [2][3][4] . Similar to most subduction zones, the vast majority of stress accumulation across the North Chilean plate boundary occurs offshore, where plate coupling is highest. Understanding deformation patterns offshore is thus important for studying seismogenesis at the margin. In January 2014, a series of earthquake swarms were followed by the 1 April 2014 Mw. 8.1 Iquique earthquake that, together with a Mw. 7.6 aftershock 2 days later, ruptured the moderately coupled central part of the seismic gap 2,5-9 (Fig. 1b). Different slip models have been published for the earthquake deviating in the absolute amplitude of peak slip (between B4 and B7 m) 2,5-9 . However, all slip models comply that earthquake rupture extended only over a B150-km-long stretch (between B19 and 20.5°S) and did not migrate farther along-strike into the highly coupled southern part of the seismic gap. The slip models also  -10 -5  0  5  10  15  20  25  30  35  40  45  50  55  60  65  70  75  ARTICLE agree on the fact that peak slip occurred down-dip of the hypocentre and that earthquake rupture diminished towards the shallow-most up-dip part of the subduction zone.
Here, we present an unpublished compilation map of swath bathymetry of the seafloor and marine multichannel seismic reflection images of the overriding plate and plate-boundary fault zone of the northern Chile seismic gap. The new data show multiple large seamounts on the oceanic plate seaward of the deformation and along the plate interface under the marine forearc. Those subducting seamounts likely exert a primary control on regional-scale plate-coupling variations and on the limited up-dip and southward extent of seismic rupture during the 2014 Iquique earthquake.

Results
Oceanic plate structure seaward of the deformation front. The moderately coupled central part of the seismic gap 2-4 that ruptured during the 2014 Iquique earthquake matches the width of the Iquique Ridge interpreted to be subducted under the marine forearc ( Fig. 1) 4,10 . The ridge, however, does not form a spatially continuous prominent swell in the seafloor of the incoming plate. Instead, the oceanic plate close to the subduction zone is characterized by a subdued topography, slightly shallower (yellowish colours) than the surrounding seafloor (green colours), that is overprinted by multiple (410) conical seamounts in the limited area with swath bathymetric coverage (Fig. 1b).
Deformation of the plate boundary and overriding plate. However, we find that seamounts are not only located seaward of the deformation front, but also on the downgoing plate under the marine forearc (Fig. 2). About 20 km landward of the deformation front, at the southern end of the area that ruptured during the 2014 Iquique earthquake, a large seamount (1.8 s two-waytraveltime (TWT) high and B20 km wide) is subducting under the marine forearc ( Fig. 2c; also compare black-green circles in Fig. 1b for projection of subducting seamounts on seafloor bathymetry). With an average seismic velocity of 4.5 km s À 1 , as inferred from wide-angle seismic data farther to the south 11,12 , the relief of the seamount corresponds to an absolute height of B4 km. The under-thrusting of the seamount results in manifold deformation of the upper plate (Fig. 2c). Above its seaward flank the continental slope steepens to B12°, indicating large-scale failure of the entire upper plate in the wake of the subducting seamount, a deformation pattern observed elsewhere in continental margins affected by seamount subduction 13 . Multiple small offsets of the seafloor and shallow strata above discontinuous deeper reflections indicate extensive faulting cutting across much of the upper plate. Down-dip of the seamount, the plate boundary is not well imaged due to the presence of the seafloor multiple. However, localized uplift and fracturing of the seafloor and upper-plate strata between kilometres 33-45 and 48-58 suggest the possible presence of two additional seamounts or similar extensive relief on the subducting plate at depth. Two other seamounts that induce a similar deformation pattern in the upper plate are imaged in seismic lines SO104-22 (Fig. 2a, 25-37 km) and SO104-26 (Fig. 2b, 16-31 km). Both seamounts are about B2.3 km (B1 s TWT) high, and B12-15 km wide.
Temporal and spatial extent of seamount subduction. The sparse seismic lines do not provide information on how far and widespread seamounts have been subducted, but kinematic reconstructions 10 support that initial ridge subduction occurred over the last 2 Ma. The black dots in Fig. 1b illustrate the downdip extent of a subducting seamount for a 2-Ma time period using a plate-dip of 25°and a convergence rate of 6.7 cm a À 1 . This shows that the seamounts imaged by the seismic data started to subduct over the last B1 Ma, which is in agreement with the kinematic reconstructions. However, from the seismic and swath bathymetric data, we cannot infer on the presence/or absence of additional seamounts in the deeper part of the subduction zone or between the seismic lines.

Discussion
The multiple seismically imaged subducting seamounts (Fig. 2) indicate that the moderately coupled central part of the northern Chile seismic gap, which ruptured during the 2014 Iquique earthquake, and the lowly coupled area around 20.5°S have a distinct tectonic evolution. Here, the physical state of the plateboundary fault zone is widely affected by the presence of excess relief on the subducting plate. Subducted seamounts and, more generally, variations in subducting seafloor roughness are believed to influence the characteristics of plate-coupling and seismic rupture at different plate boundaries [14][15][16][17] . Wang and Bilek 17 argue that a subducting seamount causes pervasive fracturing around the initial plate boundary, therewith creating favourable conditions for aseismic creep and small earthquakes and unfavourable conditions for large earthquake rupture to propagate across that area. The seismic reflection lines image the plate-boundary fault zone as a group of reflections, possibly formed by several sub-parallel surfaces containing fluids. From this we cannot identify the main decollement, which may be only a few centimetres thick. However, extensive localized deformation in the upper plate above and around the subducting seamounts, as imaged by the seismic data, supports the interpretation that in the area of the 2014 Iquique earthquake the entire overriding plate, including the plate-boundary fault zone, is widely fractured and faulted by the deformation caused by the subducting seamounts.
The deformation associated with 2-4-km-tall under-thrusting seamounts (Fig. 2), of the extension of the Iquique Ridge now located under the marine forearc, may cause the conditions that prevent the long-term accumulation of elastic energy in the shallow-most up-dip part of the subduction zone as well as in the lowly coupled region around 20.5°S (Fig. 1b). Furthermore, the seamounts imaged at the southern end of the rupture area of the 2014 Iquique earthquake that extend to the depth of peak slip (around 20.5°S; SO104-27, Fig. 2c) possibly physically prevented seismic rupture during the 2014 Iquique earthquake from migrating southwards, into the highly coupled southern area of the northern Chile seismic gap. Similar seamounts imaged at the shallow plate interface (Fig. 2a,b) might have further prevented seismic rupture from propagating into the shallow-most up-dip part of the subduction zone.
South of 21°S the oceanic plate facing the highly coupled southern segment of the northern Chile seismic gap lacks large seamounts. Instead, oceanic plate morphology at the trench is controlled by prominent (0.5-1.5 km vertical displacement) horst and grabens that result from bend-faulting of the incoming plate (Fig. 1b). A similar subducting plate morphology is observed offshore large areas of northern Japan, where seismic rupture during the giant 2011 Mw 9.0 Tohoku-Oki earthquake extended up-dip to the trench triggering the large tsunami 18,19 . In contrast, north of 19°S where plate-coupling is not well constrained, neither large seamounts nor bending-related horst and grabens are observed seaward of the deformation front (Fig. 1b). Instead, comparatively smaller bending-related half grabens are formed along the trench by reactivation of NW-SE trending spreading fabric formed at the palaeo-spreading centre 20,21 (Fig. 1b).
Our data support that the change to a complex threedimensional plate interface, which is affected by the subduction of large seamounts in the moderately coupled central part (19)(20).5°S) of the northern Chile seismic gap, possibly exerts a first-order control on plate-coupling and seismic rupture. During the 2014 Iquique earthquake, the subducted seamounts may have physically limited along-slope and up-dip migration of seismic rupture. Furthermore, the deformation associated with the underthrusting seamounts possibly creates favourable conditions for aseismic creep through extensive fracturing around the initial plate-boundary fault zone, therewith favouring the overall reduced coupling rates in the shallow-most up-dip part of the subduction zone as well as in the lowly coupled region around 20.5°S. Although our results support a segmentation of the seismic gap, a more cumulative earthquake record is necessary to understand how smaller events such as the 2014 Mw. 8.1 Iquique earthquake interact with bigger events that historically have ruptured the entire segment.

Methods
Data Collection. Seismic reflection lines off Northern Chile were collected in 1995 by the Bundesanstalt fuer Geowissenschaften und Rohstoffe during R/V Sonne cruise SO104 in the framework of the Crustal investigations off-and onshore Nazca/Central Andes (CINCA) project. Seismic data were collected with a 3-kmlong digital streamer with 25-m-long active channels, and a B3,124-cubic-inch (51.2 l), well-tuned source with two airgun strings shot every 50 m.
Data Processing. Processing of seismic data was conducted with the Claritas software package. After data editing, the traces were cleaned with a pre-stack wavelet-shaping statistical deconvolution. Later, shot point interpolation and offset regularization was done for two-dimensional filtering. Semblance-type velocity analyses were interpreted every 5 km. The water-layer multiple was attenuated with parabolic radon filtering on super gathers after normal-moveout correction with stacking velocities. A pre-stack time migration was done on receiver gathers with time and space varying velocities based on the velocity analysis. After pre-stack time migration, new velocity analyses were interpreted for common-mid-point stacking. After stacking, the data were post-stack time migrated using a finite difference algorithm, with velocities following the geological structure. For the shallow geology, seismic velocities were based on velocity analysis, whereas deeper in the section, seismic velocities were based on wide-angle seismic profiles from the margin 11,12 . After post-stack time migration, the data were bandpass filtered with time-and space-varying frequencies, based on geological structure. For display purposes, the data are presented with an automatic gain control normalization based on the mean amplitude within a sliding window of 1,000 ms.