Relationship Between Subduction Erosion and the Up‐Dip Limit of the 2014 Mw 8.1 Iquique Earthquake

The aftershock distribution of the 2014 Mw 8.1 Iquique earthquake offshore northern Chile, identified from a long‐term deployment of ocean bottom seismometers installed eight months after the mainshock, in conjunction with seismic reflection imaging, provides insights into the processes regulating the updip limit of coseismic rupture propagation. Aftershocks updip of the mainshock hypocenter frequently occur in the upper plate and are associated with normal faults identified from seismic reflection data. We propose that aftershock seismicity near the plate boundary documents subduction erosion that removes mass from the base of the wedge and results in normal faulting in the upper plate. The combination of very little or no sediment accretion and subduction erosion over millions of years has resulted in a very weak and aseismic frontal wedge. Our observations thus link the shallow subduction zone seismicity to subduction erosion processes that control the evolution of the overriding plate.

landward migration of the trench and the volcanic arc (Rutland, 1971) have been linked to subduction erosion (Clift & Vannucchi, 2004). Two modes of subduction erosion have been identified: (1) frontal erosion as commonly caused by the underthrusting of rough seafloor topography, like bending related horst-andgrabens or seamounts colliding with the lower slope (Ranero & von Huene, 2000;Sallarès & Ranero, 2005) and (2) basal erosion by the continuous removal of material from the base of the overriding plate (von Huene et al., 2004) The latter directly impacts the structural development of the plate boundary and might be essential to understand the updip limit of seismic rupture during megathrust earthquakes and the onset of seismicity at the seismogenic updip end (Byrne et al., 1988;Wang et al., 2010). The offshore location of the shallow subduction interface complicates the recording of earthquake-related processes by geophysical data, causing a gap in our understanding of the structural configuration of the plate boundary, overriding plate and subducting plate. Hence, the slip-rate deficit (kinematic coupling ratio) and the detailed seismicity are not accurately resolved during the seismic cycle. However, the extent and termination of coseismic slip are frequently revealed through the spatial distribution of seismicity after the mainshock (Mendoza & Hartzell, 1988), therefore, the postseismic period offers the opportunity to study postseismic processes (Husen et al., 1999;Tilmann et al., 2010).
To study the processes that govern the updip limit of seismic rupture in the northern Chilean subduction zone, we combine data from amphibious seismic observations from the postseismic phase of the 2014 Iquique earthquake with multi-channel seismic reflection data of the marine forearc acquired in 2016. We suggest that the updip limit of the 2014 Iquique earthquake activates subduction erosion at the updip limit of the seismogenic zone during the postseismic and possibly the coseismic phase, which leads to extensive faulting of the upper plate, thereby manifesting the location of the updip limit over many earthquake cycles.

Data and Methods
We installed two temporary passive seismic networks on the marine forearc. The first deployment was an array of 15 short period ocean bottom seismometers (OBS) 8 months after the April 1, 2014 Iquique mainshock, using the Chilean Navy vessel OVP Toro. After 1 year, the seismic network was recovered and 14 OBS were redeployed during R/V SONNE cruise SO244/2 in December 2015 and subsequently with a denser station distribution to focus on the updip aftershock distribution (Figure 1). Furthermore, we used waveform data of the permanent seismic network of the Centro Seismológico Nacional (CSN, www.sismologia.cl) and the Integrated Plate Boundary Observatory Chile (IPOC, www.ipoc-network.org).
Earthquakes were detected with the scanloc module of SeisComP3 (GFZ & gempa GmbH, 2008) using a cluster search algorithm to associate phase detections to one or many potential earthquake sources. Source scanning was done with the local 1-D velocity model from Husen et al. (1999). Next, P-phases (Aldersons, 2004;Lange et al., 2012) and S-phases were picked automatically (Diehl et al., 2009), however, automated phase picking on OBS data proved to be unsatisfactory on small magnitude events due to the elevated noise level on some waveform data. Therefore, we manually revised the P-and S-phase onsets of all offshore events.
We calculated a minimum 1D velocity model (Husen et al., 1999;Kissling et al., 1995) and a local 2D earthquake tomography for v p and v s across the South American margin at 20°S from 71.5°W to 69.4°W using SIMUL2000 (Thurber, 1983(Thurber, , 1992. Subsequently, we estimated absolute locations based on the 2D velocity model with NonLinLoc (Lomax et al., 2000). The velocity model used was constructed from the 2D model transposing it across the whole study region by following the geometry of the trench. Relative hypocenter locations were estimated using a double-difference location scheme (Waldhuser & Ellsworth, 2000). After the final relocation, seismicity in the updip area forms distinct clusters compared to previous processing steps.
Finally, we calculated moment magnitudes (Ottemoller & Havskov, 2003) and local magnitudes (Hutton & Boore, 1987), followed by the calculation of focal mechanisms for 98 stronger events (Reasenberg PETERSEN ET AL.   (Duputel et al., 2015). The 1,877 M-9 North Chilean earthquake is displayed as an dashed ellipse (Comte & Pardo, 1991), the rupture area of the M8.1 1995 Antofagasta (Chlieh et al., 2004) and the 2007 M7.7 Tocopilla earthquake (Schurr et al., 2012) are outlined by 0.5 m slip contours. Depths of the plate interface (Slab2; Hayes et al., 2018) are shown with dashed gray lines and 20 km depth intervals. Plate convergence rate is shown as a black arrow (Angermann et al., 1999). Bathymetry from Geersen, Ranero, Klauke, et al. (2018) , 1985). We used focal mechanisms from the Global Centroid Moment Tensor catalog (Dziewonski et al., 1981;Ekström et al., 2012) for the largest events as those overloaded our data loggers, in turn, causing problems in identifying first motion polarities. Additional details on the seismological data processing are given in the supporting information and Figures S1-S12.
To provide additional structural context for the seismicity analysis, we use a multi-channel reflection seismic (MCS) line MC23 acquired during the R/V Marcus Langseth cruise MGL1610 in November 2016 (Tréhu et al., 2017). The along-dip MCS profile is processed up to pre-stack depth migration (Ma et al., 2020) and images the structure around the Iquique mainshock down to 35 km depth ( Figure 2a). The MCS data were collected with an 8-km long, 640 channel streamer and 6,600 cubic inch airgun array and resolve the plate boundary and the internal structure of the marine forearc in greater detail and to a greater depth than available from previous seismic reflection studies of the northern Chilean margin (Coulbourn & Moberly, 1977;Moberly et al., 1982;von Huene et al., 1999;von Huene & Ranero, 2003;Geersen et al., 2015).

Aftershock Distribution in the Marine Forearc
Our final seismicity catalog spans 23 months and starts eight months after the mainshock occurred (Figure 1). Since we focus on the marine forearc, our local seismic catalog covers the region between 72° W and 69.5°W in longitude and between 22°S and 19°S in latitude, comprising 1,778 local earthquakes ( Figure S1). Aftershocks outside this region and deeper than 40 km will be excluded from the following discussion. Generally, the majority of hypocenters are widely distributed, occurring along the plate interface, in the subducting slab, and the overriding plate ( Figure 2). However, a significant number of events with higher magnitudes occurred updip of the mainshock and the largest aftershock. These occurred in two major clusters, marked as C1 and C2 in Figure 1, which are separated by a zone of low aftershock activity. C1 forms an NNW-SSE trending band, whereas C2 trends west to east, forming a less focused cluster of smaller magnitude earthquakes. Similar elevated aftershock seismicity updip of the mainshock area has been reported for other subduction zones (e.g., Tilmann et al., 2010) and has elsewhere been correlated to changes in the slope or subducting plate topography interacting with the upper plate (Wang & Bilek, 2014). The elevated Iquique aftershock activity in the shallow marine forearc was previously described in studies using land stations only (León-Ríos et al., 2016;Schurr et al., 2020;Sippl et al., 2018;Soto et al., 2019). However, the seismicity of the marine forearc occurs far outside the land network, resulting in increased uncertainties and a systematic bias in hypocenters for offshore earthquakes. We compare our seismicity with the catalog from Soto et al. (2019). Both catalogs have 23 days and 425 events in common. Soto et al. (2019) observe several west-east striking seismicity streaks interpreted as markers of surrounding aseismic creep along the plate interface. We do not observe the east-west striking seismicity clusters, and offshore our events tend to be located in the continental crust. In general, the horizontal location discrepancies increase with increasing distance from the coast and are largest at the seismogenic updip limit, where the OBS are located (see Figure S8; Soto et al. (2019); Figure S2). We explain the difference between the catalogs by the much better coverage of the forearc seismicity with the OBS stations. Furthermore, we use a more accurate 2D velocity model derived from the offshore seismicity. The depth uncertainties, estimated by absolute locations, of events in C1, on which the following discussion focusses, shows a range between 0.5 and 2 km ( Figure S3) and thus indicates to be a smaller depth error compared to previous seismicity studies offshore Iquique (Sippl et al., 2018;Soto et al., 2019). The observed seismicity in the forearc cross-section of Figure 2 will be described from west to east following the subduction direction. Beneath the outer rise, westward of the trench, no significant seismic activity was detected during the 23 months of OBS recording. Elsewhere outer rise aftershock seismicity has been correlated to slip during large earthquakes that extend to the trench (Sladen & Trevisan, 2018). East of the trench, a ∼35 km wide zone with very sparse seismicity is observed. Further East of the observed aseismic zone (>35 km), a large number of events are located within 5 km distance of the plate boundary, that is derived from the multi-channel seismic data. Below the plate interface, seismicity occurred between 20 and 30 km depth, indicating earthquakes are located in the lower plate. Above the plate interface, several earthquakes occurred in the upper plate during the entire observations period. The local earthquake tomography (LET) reveals an elevated v p /v s ratio in the upper crust that decreases toward the inner forearc (Figures 2c and Figure S5). East of the mainshock, the aftershocks are diffusely distributed in the upper crust with local magnitudes mostly lower than 3. The vast majority of focal mechanisms at the seismogenic updip end indicate thrust faulting with one focal plane oriented subparallel to slab dip ( Figure S10).

Marine Forearc Structure and Active Tectonics in the 2014 Iquique Earthquake Region
The multi-channel seismic profile MC23 crosses the epicentral region of the 2014 Iquique earthquake (Figure 1) and images the structure of the incoming and subducting oceanic plate and the marine forearc (Figure 2a). On the incoming plate, the crust of the oceanic Nazca Plate is repeatedly offset by up to 500 m along bending related landward and seaward dipping normal faults (Geersen, Ranero, Klauke, et al., 2018). The trench is characterized by less than 500 m of sediment cover. Landward of the trench, a series of shallow, landward-dipping reflections indicates a ca. 7 km wide frontal prism (Figure 2a). Below the frontal prism, the top of the subducting oceanic basement has a landward dip of ∼12°. The high reflectivity of the oceanic basement under the marine forearc can be observed down to depths of ∼17 km at 35 km from the trench, where aftershock seismicity commences. Between 50 and 70 km from the trench, a series of pronounced seaward dipping normal faults cut from the seafloor into the framework rock of the upper plate (dashed lines; Figure 2b). Their locations correlate to some of the aftershock hypocenters in the upper plate (also compare Reginato et al., 2020).

Discussion
The combined analysis of the 2014 Iquique aftershocks and the seismic reflection image of the marine forearc within the rupture area offers the possibility to link short term deformation associated with a single seismic cycle to the permanent deformation history of an erosive convergent margin. Previous studies of the marine forearc structure of the 2014 Iquique earthquake related the updip aftershock seismicity to postseismic processes, including postseismic relaxation or afterslip (Cesca et al., 2016;León-Ríos et al., 2016;Soto et al., 2019). In contrast to Soto et al. (2019), our aftershock catalog, which is based on 23 months of amphibious and deep crustal MCS data, does not resolve any E-W elongated streaks of seismicity. Instead, we find a broad band of seismicity with individual earthquake clusters updip of the coseismic rupture in the upper and lower plate, besides the plate interface related seismicity. Similar lower plate aftershocks related to the updip limit were observed in Costa Rica (Bilek & Lithgow-Bertelloni, 2005) and Japan (Obana et al., 2013). Obana et al. (2013) related oceanic upper crustal events to bending of the incoming plate since they observed normal faulting events in the oceanic plate. Since we observe few thrust mechanisms at the Iquique updip limit and below the plate interface (in the oceanic plate) this might be similar to thrust faulting in the downgoing plate as observed by Nippress and Rietbrock (2007) after the 1995 Antofagasta earthquake (Figure 1). The thrust faulting in the oceanic crust and mantle was suggested to be associated with the re-activation of horst and graben structures from the plate bending at the outer rise bend (Nippress & Rietbrock, 2007). Above the lower plate seismicity, we document an updip limit of aftershock seismicity that is correlated with crustal reflectivity (Figures 1 and 2b), indicative of long-term along-dip seismo-tectonic segmentation of the subduction zone.
If the updip limit of plate-boundary seismicity is stable in space over many earthquake cycles, it can induce permanent forearc deformation expressed in first-order topographic changes (Rosenau & Oncken, 2009). Indeed, at other subduction zones, similar intense updip clusters of seismicity often correlate with structural or topographic changes of the forearc (Lange et al., 2007;Lieser et al., 2014;Tilmann et al., 2010;Tréhu et al., 2019). Furthermore, such updip seismicity occurs predominantly along the plate boundary. aftershocks occur in a narrow band roughly updip of the main coseismic rupture (upper panel in Figure 2b), implying postseismic deformation at the seismic-to-aseismic transition. These aftershocks are, however, not associated with a structural or topographic change of the forearc (Figure 2). They are further not limited to the megathrust fault but also occur in the upper overriding and lower subducting plates (Figure 2b).
We interpret the striking aftershock sequence at the updip limit of seismic rupture following the 2014 Iquique earthquake as an indication of active subduction erosion during the coseismic and early postseismic phase. From fossil subduction zones, we have learned that a wide shear zone, as exemplified by the different depths of the 2014 Iquique aftershocks, is characteristic for margins dominated by subduction erosion (Vannucchi et al., 2008). In the concept of subduction erosion, fracturing at the base of the upper plate starts at the updip limit of the seismogenic zone and increases toward the shallow plate interface up to the frontal prism (von Huene et al., 2004). This basal erosion induced by the 2014 Iquique earthquake is indicated by the high number of aftershocks that occur at or slightly above the plate boundary (Figure 2b). Such abrasion of material from the underside of the upper plate is expected to be in a dynamic equilibrium between the removal of material and steepening of forearc slope (von Huene et al., 2004). The associated deformation of the entire upper plate is illustrated by the overall fewer but still significant number of aftershocks above the updip limit of seismic rupture (Figure 2b). These aftershocks seem to occur along seaward dipping planes that match the location of seismically imaged normal faults within the upper plate (Figure 2a). This implies that the faults have moved in the postseismic phase of the 2014 Iquique earthquake. Although the absolute depths of the normal faults are not fully resolved by the seismic data, data from other erosive convergent margins suggest that they may cut to the plate boundary (Contreras-Reyes et al., 2015;Kodaira et al., 2012;Ranero et al., 2008).
While the aftershocks of the 2014 Iquique earthquake represent forearc deformation during the early postseismic phase of one earthquake, the seismically imaged structure of the marine forearc is a result of forearc deformation from hundreds to thousands of seismic cycles (Scholz, 1998). Repeated coseismic and postseismic deformation and associated subduction erosion force extensive faulting and pervasive fracturing at the updip limit of seismic rupture. Deformation is not limited to the plate boundary but active throughout the entire upper plate from the plate boundary to the seafloor (compare Geersen, Ranero, Klauke, et al., 2018). This is consistent with earlier studies based on numerical and analog modeling and conceptional considerations on the relationship between long-term forearc deformation of the overriding plate and earthquake behavior of subduction zone forearcs (Rosenau & Oncken, 2009;Wang & Hu, 2006). Northern Chile is an end-member margin in terms of trench sediment thickness, and subduction erosion is likely the dominant tectonic mode since at least Mesozoic times (Rutland, 1971). This is manifested in about 250 km loss of overriding continental plate since 150 Ma (Scheuber & Reutter, 1992) and associated eastward migration of the trench and volcanic arc. Over time, tectonic erosion of the upper plate (von Huene & Ranero, 2003) causes the updip limit to migrate landwards, mimicking the migration of the volcanic arc and the trench.
In the Iquique region of the 2014 Iquique earthquake, long-term upper plate faulting and fracturing at the updip limit of the seismogenic zone are manifested in the permanent deformation pattern of the marine forearc seaward of the 2014 Iquique aftershocks. This part of the marine forearc between the trench and the current updip limit (40 km distance) has migrated through the updip limit over the last millions of years. It is heavily faulted as indicated by discontinuous seismic reflections. The overall high reflection amplitudes in this region further support a high degree of fracturing and possibly fluids within the fractures. The elevated v p /v s ratio in the upper plate recognized from local earthquake tomography (Figure 2c) further supports the presence of fluids in the highly fractured outermost marine forearc (Popp & Kern, 1994;Wang et al., 2012). The fractured and fluid-rich outermost marine forearc, seaward of the 2014 Iquique earthquake rupture is likely too weak (and heavily deformed) to store sufficient elastic energy to nucleate a great earthquake. Further down-dip, the decrease in faulting and fluid content in the marine forearc, together with the onset of aftershock seismicity, indicates an increase in strength of the overriding plate that allows the storing of elastic energy (Figure 3). A similar down-dip segmentation of the North Chilean forearc that can build up elastic energy and rupture during great earthquakes has been previously suggested based on gravity data and seismic velocity structure Maksymowicz et al., 2018;Sallarès & Ranero, 2005). The segmentation is in-line with our amphibious aftershock observation of the 2014 Iquique earthquake rupture. It is further supported by the decrease of coseismic rupture at the transition from the heavily deformed outermost marine forearc to the less deformed section of the forearc between the coast and the updip limit of the seismic rupture (Duputel et al., 2015;Jara et al., 2018;Schurr et al., 2014).

Conclusions
Combining 2 years of local seismicity observations following the 2014 Iquique earthquake and structural constraints on forearc architecture derived from MCS data, we provide evidence for the interplay of plate boundary rupture and upper plate deformation in the context of long-term subduction erosion. The majority of aftershocks of the 2014 Iquique earthquake occurred updip of the coseismic slip and ∼32-60 km landward of the trench. Although most of the seismicity was within ∼5 km of the plate boundary, earthquakes extended through the upper plate, defining a seaward dipping zone that coincides with seaward dipping normal faults imaged in the MCS data. The updip band of aftershock seismicity separates a pervasively fractured and likely fluid-filled marine forearc farther seaward from a less deformed section of the upper plate forearc. At the transition, active subduction erosion during the postseismic and possibly coseismic phases leads to basal abrasion of the upper plate and associated extensional faulting of the upper plate at the updip end of the seismogenic zone. Landward migration of the updip end of the seismogenic zone, at similar rates compared to the trench and the volcanic arc, preconditions the structural setting of the heavily fractured, fluid-filled and therefore weak and aseismic outermost marine forearc.