Extension at the coast of the Makran subduction zone (Iran)

In the Makran subduction zone, earthquake focal mechanisms and geodetic data indicate that the deforming prism currently experiences N–S compression. However, palaeostress inversions performed on normal faults observed along the coast reveal local stress components consistent with N‐S extension. Previously proposed mechanisms such as gravitational collapse are not favoured by N–S compression and surface uplift. We propose that the observed kinematics result from transient stress reversals following large earthquakes. During the interseismic period (now), the region experiences N–S compression. However, following a large reverse rupture on the subduction interface, stresses in the inner wedge relax, enabling a brief period of extensional faulting before a compressive stress state is re‐established. This mechanism, also observed in other subduction zones, requires low overall stresses in the upper plate and that the margin ruptures in large megathrust earthquakes that result in nearly complete stress drops.


| INTRODUC TI ON
The converging movement of tectonic plates at subduction plate boundaries typically induces compressional stress states, leading to folding and thrusting in the overriding plate. However, as observations accumulated, it has become increasingly clear that this picture is oversimplified for many active margins. For example, several subduction zones have normal faults in the upper plate, indicating extension in the direction of subduction (e.g. Adam & Reuther, 2000).
In this manuscript, we describe field evidence for widespread normal faulting in the Makran subduction zone that indicates margin-normal extension in the upper plate. These kinematics oppose those in the accretionary prism located south of the studied region, where numerous margin-parallel thrusts are observed (Grando & McClay, 2007;Kopp et al., 2000;Smith, McNeill, Henstock, & Bull, 2012). They also oppose the kinematics of regions situated to the north, which are currently experiencing shortening in the direction of convergence, accommodated by folding and thrusting (Burg, Dolati, Bernoulli, & Smit, 2013;Haghipour et al., 2012). Our investigation focuses on the coastal strip of the Makran where normal faults are commonly observed (e.g. Burg et al., 2013;Ellouz-Zimmermann et al., 2007;Grando & McClay, 2007;Hosseini-Barzi & Talbot, 2003;Normand, Simpson, Herman, et al., 2019;Ruh, 2017). We measured more than 200 normal faults, which we use to reconstruct the palaeostress conditions at the time of their formation. We compare these conditions with those implied by GPS measurements and earthquake focal mechanisms in an attempt to explain their origin.

| G EOLOG I C AL S E T TING
The Makran is an east-west trending belt located in southern Iran and Pakistan involving northward underthrusting of an oceanic portion of the Arabian plate below the continental plate of Eurasia ( Figure 1a). This region hosts one of the widest and thickest modern accretionary prisms on Earth, extending from the trench located 80-to 150-km offshore to 250-km inland. The offshore part of the prism accommodates most of the relative compression between the two plates, where seismic profiles reveal the presence of numerous thrusts and folds (Grando & McClay, 2007;Smith et al., 2012).
However, some convergence is also being accommodated within the emerged part of the prism, as evidenced by the presence of deformed late Quaternary fluvial and marine terraces (Haghipour et al., 2012;Normand, Simpson, Herman, et al., 2019). In addition to this compressional deformation, a narrow region near the coast is characterised by the presence of normal faults (e.g. Burg et al., 2013;, Hosseini-Barzi & Talbot, 2003 and seismic sections reveal the presence of south-dipping listric normal faults about 50-km offshore of the study area (Ellouz-Zimmermann et al., 2007;Grando & McClay, 2007). Both numerical modelling and sandbox experiments suggest that these major listric faults root into the main decollement level and are caused by gravitational collapse of the prism (Ellouz-Zimmermann et al., 2007;Ruh, 2017).
The Makran has experienced relatively few large historical earthquakes in comparison to other subduction zones. The last great thrust event in the region was a Mw 8.1 in 1945, which ruptured the plate interface beneath the Pakistan coast ( Figure 1b) (Byrne, Sykes, & Davis, 1992). The Iranian segment of the Makran has not ruptured in the last ~500 years (Heidarzadeh et al., 2008;Musson, 2009).

| NORMAL FAULTING IN THE S TUDY ARE A
This study focuses on structural observations in the vicinity of Chabahar, a 30 × 15 km headland where Pliocene rocks locally host uplifted Pleistocene terraces and associated sediments ( Figure 3).
The Chabahar headland is bound by a series of NE-SW and NW-SE striking normal faults that dip towards the headland (Figures 3 and   4a). Thus, even though the headland stands above the surrounding topography and is regionally uplifted, it is actually a downthrown fault block. The reason it stands high is due to exposure of resistant sandstones, whereas the surrounding footwall is comprised of easily erodible marls. The age of the offset units (Upper Miocene footwall vs. Pliocene hanging wall) (Samadian et al., 1996)

| S TRE SS INVER S I ON
We used fault geometry measurements made on Chabahar headland  to invert for the palaeostress at the time faults were formed using the Win-Tensor 5.8.8 software ( Figure 5) (Delvaux & Sperner, 2003) (see Data S1). Assuming all faults are normal, our results show that ~85% of the faults are consistent with a stress state whereby σ 1 is vertical, σ 3 is approximately horizontal and oriented ~20°N and σ 2 is parallel to the trench ( Figure 5c). This stress state is exactly the opposite of that inferred from GPS ( Figure 1b) and earthquake focal mechanisms (Figure 2).
This suggests that σ 2 and σ 3 could locally switch, which might explain the occurrence of a minor subset of normal faults with different orientations (Figures 3 and 5a). The downthrown fault block forming the Chabahar headland could also be an orthorhombic fault system consistent with N-S extension (e.g. Hosseini- Barzi & Talbot, 2003;Reches, 1978).
F I G U R E 5 Geometry and palaeostress state for normal faults on the Chabahar headland. All stereonets are equal area, lower hemisphere projection (Schmidt). Palaeostress inversion was performed with Win-Tensor 5.8.5 software (Delvaux & Sperner, 2003). (a) Poles of all 213 fault planes with density contours, plotted with Stereonet 9.5 (Allmendinger, Cardozo, & Fisher, 2013;

| D ISCUSS I ON
We have shown that despite GPS data ( Figure 1) and focal mechanisms ( Figure 2) suggesting that the coastal Makran is currently experiencing N-S compression (along with exposed marine terraces that show it is undergoing regional surface uplift), normal faults observed onshore near the coastline indicate N-S extension. Here, we discuss possible explanations for these opposing kinematics.
According to the critical wedge theory, the stress state in a deforming frictional prism adjusts itself to balance the applied tectonic stresses, shear resistance to sliding on the decollement and gravity (Dahlen, 1984;Davis, Suppe, & Dahlen, 1983). Except for very high values of fluid overpressure (λ > 0.98, see Dahlen, 1984), the critical wedge theory predicts a horizontally compressive stress state for the low-tapered (c. 5°) Makran wedge. Although this is consistent with deformation near the toe of the prism, it is not in line with extensional faulting observed at the coast, especially northwards of the shelf break, where the taper is further reduced (Figure 1c). One possible explanation for this extensional deformation is to advocate a change in conditions with time, which might cause the wedge to locally readjust (Dahlen, 1984). For example, if the basal decollement under the coastal region suddenly experienced a decrease in frictional resistance (e.g. due to an increase in fluid overpressure), the local wedge taper would be too steep for the new conditions, which is predicted to drive the wedge into extension to reduce the slope of the topography and achieve a new equilibrium (see figs. 12 and 13 of Dahlen, 1984). Extensional faulting could also result from sediment underplating, causing steepening of the taper followed by gravitational collapse (Platt et al., 1985), facilitated by an increase in fluid overpressure with depth (Ruh, 2017) and/or by sediment loading (Ellouz-Zimmermann et al., 2007). Numerical models and sandbox experiments have shown that these phenomena could be the cause of the large listric normal faults observed near the Makran shelf break with seismic reflection (Ellouz-Zimmermann et al., 2007;Ruh, 2017). However, if these mechanisms were the origin for the recently active normal faults observed onshore in this study, we would expect GPS measurements and coastal morphologies to indicate N-S extension and subsidence (Ruh, 2017), rather than the compression and uplift that is observed.
One way to reconcile the apparently conflicting kinematics observed in the Makran coastal strip is to consider the potential influence of large megathrust earthquakes (Wang & Hu, 2006), such as the Mw 8.1 event in the eastern Makran in 1945 (Byrne et al., 1992).
During great earthquakes, the inner part of the wedge, situated directly above the ruptured interface (i.e. the coastal region in the case of the Makran, Figure 1b), is expected to experience coseismic stress release (Figure 6b) (Wang & Hu, 2006;Wang, Hu, & He, 2012). If friction on the subduction interface during rupture drops to a sufficiently low value, the portion of the wedge situated directly above the ruptured area (the inner wedge) may be driven into a state of critical extension, leading to normal faulting (Figure 6b). After an earthquake, the inner wedge will progressively revert to compression due to locking of the subduction interface (Wang et al., 2012) (Figure 6a).
Even so, the stress state in this part of the wedge will depend on the amount of stress drop during the rupture and it may remain in the stable domain (i.e. be insufficient to induce thrust faulting) (Wang & Hu, 2006). Indeed, no true reverse faults were observed in the coastal Makran and episodes of normal fault reversal have only been interpreted (Grando & McClay, 2007;Hosseini-Barzi & Talbot, 2003).  (Dewey et al., 2007;Farías, Comte, Roecker, Carrizo, & Pardo, 2011;Hardebeck, 2012;Kato, Sakai, & Obara, 2011). The observation of coseismic rotation of the principal stress axes implies that the mentioned earthquakes caused a near-complete stress drop, which was sufficient to drive the surrounding crust into extension (Hardebeck, 2012;Hasegawa, Yoshida, & Okada, 2011).
Our mechanism proposed to explain the kinematics in the inner wedge requires that the western Makran does indeed experience occasional large megathrust earthquakes, as is known to occur in the eastern Makran. This mechanism also requires extremely low frictional resistance on the subduction interface during rupture and a low overall stress state in the inner wedge (Wang, 2000). Both of these features are consistent with numerical experiments (Ruh, 2017) and evidence for high-fluid pressures in the Makran wedge such as an overall low wedge taper (Smith et al., 2012;White & Louden, 1982), the presence of mud volcanos (e.g. Delisle, 2004;Snead, 1964), overpressured shale layers in the prism (Ruh, Vergés, & Burg, 2018)

| CON CLUS IONS
Our field investigation in the Makran subduction zone has revealed widespread evidence for normal faults oriented parallel to the coast, implying extension of the upper plate in the direction of subduction. These kinematics are the opposite from those indicated by GPS measurements and earthquake focal mechanisms, which indicate present-day N-S compression. We postulate that these kinematics reflect stress changes linked to major earthquakes on the subduction interface. According to this view, the coastal part of the wedge experiences N-S compression and is mostly stable during the interseismic period, whereas it experiences a brief period of extension as stresses are released during large ruptures on the subduction interface. This is permitted if near-complete stress drop occurs during the megathrust event, implying a weak subduction interface, which is supported by evidence for high pore-fluid pressures in the Makran prism.

ACK N OWLED G EM ENTS
This work was funded by the Swiss National Science Foundation, project no. 200021_155904. We are grateful to Reza Ensani, Feisal

Arjomandi, Nurrudin Mazarzehi, Yousef Adeeb and Gholamreza
Hosseinyar for helping us with logistics in Iran and accompanying us in the field. The authors declare that they have no conflict of interest.

DATA AVA I L A B I L I T Y S TAT E M E N T
All data collected in this study are freely available in the following data repository: https ://doi.org/10.5281/zenodo.2559480 .