The 2010–2011 South Rigan (Baluchestan) earthquake sequence and its implications for distributed deformation and earthquake hazard in southeast Iran

R. T. Walker,1 E. A. Bergman,2 J. R. Elliott,1 E. J. Fielding,3 A.-R. Ghods,4 M. Ghoraishi,5 J. Jackson,6 H. Nazari,5 M. Nemati,7,8 B. Oveisi,8 M. Talebian5 and R. J. Walters1 1Department of Earth Sciences, University of Oxford, South Parks Road, Oxford, OX1 3AN, UK. E-mail: Richard.Walker@earth.ox.ac.uk 2Center for Imaging the Earth’s Interior, Department of Physics, University of Colorado at Boulder, 390 UCB, CO 80309, USA 3Jet Propulsion Laboratory, California Institute of Technology, 4800 Oak Grove Drive, Pasadena, CA 91109, USA 4Department of Earth Sciences, Institute for Advanced Studies in Basic Sciences (IASBS), Zanjan, Iran 5Research Institute for Earth Sciences, Geological Survey of Iran, Azadi Square, Meraj Avenue, P.O. Box 11365–4563, Iran 6Bullard Laboratories, Cambridge University, Madingley Road, Cambridge, CB1 3EZ, UK 7Geology Department, Science Faculty, Bahonar University of Kerman, P.O. Box 76169143, Iran 8Seismotectonics Department, Geological Survey of Iran, Azadi Square, Meraj Avenue, P.O. Box 11365–4563, Iran


S U M M A R Y
We investigate the source processes and tectonic significance of two earthquakes that occurred on 2010 December 20 (M w 6.5) and 2011 January 27 (M w 6.2) within a desert region south of the town of Rigan, SE Iran.The two earthquakes, which we refer to as the South Rigan events, occurred close to one another at the northern margin of the Shahsavaran mountains: a mainly volcanic chain in which the potential for active faulting has not previously been considered in detail.Surface displacements mapped using SAR interferometry, multiple-event relocation analysis of epicentres, body-waveform modelling and field measurements of surface rupture together reveal that the 2010 December 20 earthquake involved an average of ∼1.3 m right-lateral slip on a vertical fault trending ∼210 • whereas the 2011 January 27 resulted from ∼0.6 m of slip on a conjugate left-lateral fault striking ∼310 • , parallel to the trend of the Shahsavaran mountains and confined within a zone of increased Coulomb stress from the earlier main shock.The main slip for the 2010 and 2011 main shocks failed to reach the surface though minor cracks and en-echelon fissures were mapped following both events.Some of the surface cracks may have been enhanced during a period of minor afterslip in the days following the 2010 main shock.Using the insights gained from our investigation of the two South Rigan earthquakes we perform a regional reconnaissance of the active faulting using SPOT5 (2.5 m) satellite imagery.We show that distributed ∼N-S right-lateral faulting is widely distributed north of the Shahsavaran mountains.We also show evidence for left-lateral strike-slip faulting parallel to the Shahsavaran mountains, with a component of extension in the east and shortening in the west, which is likely to accommodate regional N-S right-lateral shearing by clockwise rotation about a vertical axis.The distributed strike-slip faulting is closely associated with the distribution of towns and villages and constitutes a continuing hazard to local populations.

I N T RO D U C T I O N
In the evening of 2010 December 20, the rural desert region south of Rigan village in southeast Iran was shaken by an earthquake of M w 6.5.Despite the very sparse habitation the earthquake caused four deaths in the small settlement of Chah Qanbar.A second moderately-sized earthquake (M w 6.2) struck nearby on 2011 January 27.What is distinguishing about these two events is that they occurred in an area with little previously recorded seismicity (Ambraseys & Melville 1982), where few active faults have previously been mapped, and deep within what is often regarded as a non-deforming block within the Arabia-Eurasia collision zone.
The 2010 and 2011 earthquakes are not the only destructive events to have occurred on previously unknown faults within the deserts of eastern Iran.Although an active fault was known in the vicinity of Bam (Berberian 1976), the devastating 2003 December 26th Bam earthquake (M w 6.6), which was responsible for ∼30,000 deaths, occurred on a previously unmapped strike-slip fault in open desert south of the city (e.g.Talebian et al. 2004;Berberian 2005;Fialko et al. 2005;Jackson et al. 2006).The constraints on faulting provided by the South Rigan and Bam earthquakes offer the potential to learn about the distribution of active faults in this relatively unexplored region, as well as to examine the kinematics of active deformation in the southeastern boundary of the Arabia-Eurasia continental collision.The study thus gives us the potential to learn about the general kinematic processes involved in zones of distributed strike-slip faulting.
In this paper, we present a combined seismotectonic analysis of the South Rigan earthquakes and their role in the tectonics of southeast Iran.We use seismological, InSAR and field observations to constrain the source parameters of the two earthquakes and to locate their causative faults.We determine the source parameters of the two main shocks through body-wave waveform modelling.The location and orientation of the two fault sources are constrained through synthetic aperture radar (SAR) interferometry using advance land observation satellite (ALOS), TerraSAR-X and COSMO-SkyMed radar imagery.Improved epicentral locations for 93 earthquake events within the region provide constraints on the rupture propagation directions, the locations of the larger aftershocks and their depth range.The InSAR and seismological constraints are then combined with the available field observations of rupturing and fault geomorphology, both within the immediate epicentral zone and the wider region, to address the wider subject of active deformation within southeast Iran.

T E C T O N I C A N D G E O L O G I C A L S E T T I N G
The active tectonics of Iran are controlled by the northward motion of Arabia, at a rate of ∼25 mm yr −1 at a longitude of 56 • E, relative to Eurasia (Fig. 1; Vernant et al. 2004).The GPS velocities of points relative to Eurasia decrease to zero at both the northern and eastern borders of Iran indicating that the major part of the continental shortening is confined within the political borders of the country, with the majority of the deformation concentrated in the Zagros mountains of southern Iran (Z in Fig. 1a), and in the Alborz and Kopeh Dagh mountains in the north (A and K in Fig. 1a).
The northward motion of central Iran relative to western Afghanistan results in regional scale right-lateral shear across the eastern border of Iran, which is accommodated south of latitude 34 • N on N-S right-lateral faults that surround the Dasht-e-Lut (e.g.Walker & Jackson 2004;Meyer & Le Dortz 2007; see Fig. 1b).The 2010 and 2011 South Rigan earthquakes occurred in the low-lying and sparsely populated Narmashir desert region south of the Dasht-e-Lut desert.The Narmashir desert is bordered to the south by the Shahsavaran mountains mountain chain (Fig. 2).
In the Shahsavaran mountains chain, a sequence of Palaeozoic sedimentary rocks and granitic plutons of either Cretaceous or Palaeozoic age are exposed on the central and southern slopes (Berberian 1990;Nogol-e-Sadat 1996).The extent of outcrop of these Palaeozoic-Mesozoic rocks is outlined on Fig. 2. The northern slopes of the Shahsavaran mountains are formed from volcanic rocks of Eocene to recent age (Berberian 1990;Aghanabati 1992).The main active volcanic centre is Bazman volcano, though smaller vents and lava flows are scattered both along the northern slopes of the mountains and in the alluvial plain to the north of it.The epicentral zone of the 2010 and 2011 South Rigan earthquakes is centred on a small alluvial plain; the Saif Al-Dini plain.Saif Al-Dini is bounded to the south by Pliocene andesites and to the north by a volcanic centre exposing mainly basaltic andesites of Plio-Quaternary age, Quaternary olivine basalts and scoria cones, and minor outcrops of Miocene andesites and dacites (Aghanabati 1992; Fig. 3).

Overview and macroseismic effects
The 2010 December 20 South Rigan earthquake (M w 6.5) occurred at 22:12 local time (18:42 GMT) in a remote desert region ∼60 km south of the small town of Rigan in Kerman province of southeast Iran (Fig. 2).Initial epicentral locations placed the 2010 December 20 earthquake at 28.41N 59.18E (USGS PDE) and 28.10N 59.11E (Global CMT).The earthquake was felt in Bam, Iranshahr, Khash and Zahedan.The four casualties of the earthquake were children from a single family in the small hamlet of Chah-Qanbar (Fig. 3).The family were the only inhabitants occupying a masonry house.The remainder of the semi-nomadic population, who were living in tents (of a type locally called 'palas') at the time of the earthquake, were unharmed.Two small masonry houses at 28.15N 59.06E, ∼3 km west of Konarak spring (28.14N 59.09E), were destroyed, trapping a local woman for several hours.The population of Seyfeddini village (28.11N 59.07E), and at the temporary encampment at Take Farhad (28.136N 59.184E), live in palas tents and no casualties were reported at either location.
A second significant earthquake (M w 6.2) struck the same desert region at 12:09 local time (08:39 GMT) on the 2011 January 27.We refer to this event, which may be thought of as a large aftershock of the 2010 main shock, as the 2011 January 27 South Rigan earthquake.The USGS PDE epicentre was at 28.19N 59.01E and the Global CMT epicentre at 28.02N 59.02E.No casualties were reported from the 2011 event.Jackson et al. 2006;and this paper), the Global CMT catalogue (dark grey; http://www.globalcmt.org/), and from first-motions (light grey; listed in McKenzie 1972).Earthquakes of the Bam and South Rigan sequences are plotted at the locations listed in Table 1.All other focal mechanisms are plotted at their Global CMT locations (Ekstrom et al. 2012).Fault slip-rates are from Regard et al. (2005) and Walker et al. (2010).This map is in a Mercator projection.

Calibrated relocation of epicentres
Systematic location errors of up to 10-15 km have been reported in Iran (e.g.Berberian 1979;Ambraseys 2001;Walker et al. 2011).Given that the active faults in Iran are often separated by only 10-15 km, the uncertainty in epicentral location often precludes a confident identification of the causative fault using standard catalogue locations.To improve the location of events in the South Rigan sequence we use a relocation method based on the hypocentroidal decomposition (HDC) method for multiple event relocation (Jordan & Sverdrup 1981), to determine earthquake locations for which scatter and bias have both been minimized.The method has been applied in a number of recent studies, including several in Iran (Ritzwoller et al. 2003;Bondar et al. 2004;Walker et al. 2005;Biggs et al. 2006;Parsons et al. 2006;Tatar et al. 2007;Bondar et al. 2008;Nissen et al. 2010;Walker et al. 2011).The algorithm separates the relocation procedure into two steps, solving for the relative location of each event in the cluster with respect to the 'hypocentroid' or geometrical centre of the cluster and solving for the absolute location of the hypocentroid.For the cluster of earthquakes used in this study, we perform an 'indirect' calibration of the hypocentroid location, shifting the events such that the epicentres of the two South Rigan main shocks-and the epicentre of the 2003 December 26 Bam main shock, which we include in our analysis-lie on the bestfitting faults obtained from elastic dislocation modelling of InSAR data (see Section 3.5).
We use a cluster of 93 instrumentally recorded earthquakes in the HDC analysis.The calibrated locations of all 93 events are presented in Table 1.Many of the events in the cluster were extracted from a database of earthquake locations in the Iran region, located with the EHB single-event location algorithm of Engdahl et al. (1998).Except for a small number of events that were excluded from the HDC analysis (because of poor azimuthal coverage, too few arrival time data or inconsitent arrival time data), the cluster includes all those associated with the 2003 Bam earthquake, the 2010-2011 South Rigan sequence, along with scattered seismicity elsewhere within the region.
Although including events from Bam expands the geographical size of the cluster, and hence increases the possibility that heterogeneity in the velocity structure will introduce bias into the  2. Epicentres of smaller earthquakes are shown as red circles.All epicentral locations are those derived from HDC analysis and are displayed with their 90 per cent confidence ellipses for relative location (see Section 3.2; listed in Table 1).The locations of observed surface breaks from the first earthquake are shown as blue circles and those from the second earthquake as green circles.Note that the central section of ruptures originated from both the 2010 and 2011 earthquakes, and the blue circles are partly obscured by the overlying green symbols.Faults suspected as active in the region are shown as dashed black lines.The region is dominated by Quaternary and late Tertiary volcanism.Small basins filled with alluvial fans and small salt pans are sited between volcanic outcrops.Volcanic vents are distinguished by oxidised scoria, which appears red in the band combination used in the figure.This map is in a local zone UTM projection.calibrated locations, the overall cluster location is improved by adding a third a priori constraint (the 2003 Bam main shock).Including seismicity from Bam also provides more data with which to carry out the 'cleaning' process of identifying and flagging outlier readings, and improves the estimates of empirical reading errors.It is especially helpful for cleaning the arrival time data of the two South Rigan main shocks to have the Bam event included, as there are many readings that are only in common between the largest events.There is also improved constraint on origin time calibration from near-source readings associated with the Bam sequence.
The relative locations of all events in the cluster are constrained by phase readings at regional and teleseismic distances.We include arrival time data from the two main permanent seismic networks in Iran, the Iran National Seismograph Network (INSN), operated by the International Institute for Seismology and Earthquake Engineering (IIEES) and the Iran Telemetered Seismograph Network (ITSN), operated by the Iran Seismological Research Center (IRSC) at the Institute of Geophysics of Tehran University.The regional distance phase readings of the INSN and ITSN stations allow us to add many more events to the cluster and to better constrain the relative locations of the events already represented with teleseismic data.
Table 1.Calibrated epicentres of earthquakes in the study region determined with the methods described in Section 3.2.A1 and A2 are the azimuths in degrees, clockwise from north, of the semi-axes of the 90 per cent confidence ellipse for the epicentre.L1 and L2 represent the lengths of the corresponding semi-axes in km.Area is the area in km 2 of the 90 per cent confidence ellipse.Those events for which body-waveform or CMT solutions are available ( They do not help constrain the absolute location of the hypocentroid, however, because the nearest station is about 2 • epicentral distance away.This distance is too far, given the uncertainty about the crustal velocity structure and also because the azimuthal distribution of stations is poor, with the majority of stations located in the northwest quadrant relative to the epicentral region.Many of the events in the INSN and ITSN catalogues are very small and poorly recorded.We include in the cluster all events for which there was ITSN data through 2011 January 30, then added the corresponding INSN readings.The INSN stations are generally too few in number to allow reliable locations in this region by themselves, without being supplemented by the ITSN stations.The phase data sets of the teleseismic events in the cluster were supplemented with all available INSN and ITSN readings. The location calibration of the 2010-2011 South Rigan sequence is based on fitting the pattern of relative locations for the 2003 December 26, 2010 December 20 and 2011 January 27 main shocks to the set of near-orthogonal fault traces inferred from InSAR analysis, field work and additional seismological constraint in the case of the 2003 Bam event.The arrangement of events and location constraints is shown in Fig. 4. The 2003 Bam event (Fig. 4a) is strongly constrained to lie near a buried fault extending southwards from the city of Bam.The location along the fault is constrained by an S-P datum from an ISMN station in the city (e.g.Jackson et al. 2006).The 2010 December 20, South Rigan main shock is constrained by InSAR analysis to lie along the trend of the fault trace trending NNE and the 2011 January 27, main shock is constrained by In-SAR data to lie along the trend of the inferred fault trending WNW (Fig. 4b, see Section 3.5).All three constraints can be satisfied to high accuracy.The combined uncertainty in the location calibration of the entire cluster is about 1.6 km.This uncertainty is added to the uncertainty in relative location to obtain the absolute uncertainty in the calibrated location for each event.The increase in the size of each confidence ellipse is less than 1.6 km, because of the nature of covariance matrix addition.All relocated epicentres in the South Rigan region are shown, along with their 90 per cent confidence ellipses, in Fig. 4(b).

Origin times and depth distribution of events
Readings from three temporary seismograph stations installed about 50 km from the epicentral region near the town of Rigan (Fig. 2) by the Geological Survey of Iran (GSI) provide substantial constraint on the relative locations of many of the South Rigan sequence aftershocks, and also provide a basis for calibrating the origin time of the cluster.The earliest readings are for an event on 2010 December 28.Additional constraints on origin time were provided by localdistance readings from the earlier Bam earthquake sequence.Calibration of origin time is performed by comparing observed arrival times with the predicted times from a local velocity model.Given that the locations of the earthquakes are constrained by other data, the velocity model can be adjusted to best fit the observed slope of observed readings.The resulting model is a two-layer crustal model, with a 12-km thick surficial layer of 5.6 km s −1 overlaying a 34-km thick crustal layer of 6.3 km s −1 .The Moho depth of 46 km matches  1 was calibrated by fixing the three main shock epicentres (in yellow, with their 90 per cent confidence ellipses) to lie on the faults-shown as green lines-modelled from uniform-slip inversion of InSAR data for the Bam (from Jackson et al. 2006) and South Rigan earthquakes (Section 3.5).The Bam fault model of Jackson et al. (2006) includes a subevent on a west-dipping reverse fault projecting to the surface at the Baravat escarpment.The epicentre of the 2010 South Rigan main shock lies ∼4 km north of the modelled fault for that earthquake.This apparent mislocation is likely to result from rupture extending north of the limit that is not present in the uniform-slip model (see variable-slip model in Fig. 9).(b) Relocated events in the South Rigan sequence with their 90 per cent absolute location uncertainties.Epicentres are shown on a LANDSAT ETM image (band 5) and are numbered as in Table 1.The two main shocks are shown in yellow to distinguish them from other seismicity (in red).
the observed arrival times of Pn phases closely and the entire model is similar to those derived from many other such analyses in the Iran region (e.g.Walker et al. 2005;Parsons et al. 2006;Tatar et al. 2007;Roustaei et al. 2010;Nissen et al. 2010;Walker et al. 2011;Ghods et al. 2012).The fit of observed arrivals to the local velocity model is shown in Fig. 5.
For the relocations presented in Table 1 we applied constraints on focal depth, where possible, from phase readings at short epicentral distance and from teleseismic depth phases.For algorithmic reasons, it is not feasible to let focal depth be a free parameter for all events in the HDC analysis, so we use the available constraints to fix the depth of the corresponding events prior to relocation.A 'cluster default' depth, based on the distribution of constrained depths, is set for the remaining events.
The strongest constraint on focal depth is provided by readings at epicentral distances that are comparable to, or less than, the focal depth (i.e. a few tens of kilometres).For the South Rigan sequence there were no permanent seismic stations near enough to provide such a constraint and the temporary stations deployed by the GSI were too far away for this purpose.It may yet turn out that there will be some data of this type from other temporary deployments of seismographs.However, several accelerometer stations of the Iran Strong Motion Network (ISMN), operated by the Building and Housing Research Centre (BHRC), are located near enough to provide constraint on focal depth.Unfortunately the ISMN stations have uncalibrated timing systems such that the reported arrival times cannot be used directly.We picked P and S arrival times from the original waveform records and then used the S-P delay timesforming a set of points on the surface of a hemisphere of given radius, and centred on the station-to constrain distance from the station.
As the cluster location is calibrated by other data, the S-P times provide useful constraint on the depth of nearby events.Theoretical S-P times are calculated from the local velocity model (Fig. 5).Reports of teleseismic depth phases from standard catalogues can provide useful constraints on focal depth also, but they require careful processing and interpretation.For the HDC analysis we consider only relative depth phases (pP-P and sP-P) in the teleseismic range (28-98 • ), which we have found to be more stable than raw depth phases.The relative times are calculated from the ak135 global traveltime model, which incorporates a crustal velocity model with somewhat faster velocities.Therefore, a slight bias towards greater depth is expected from the teleseismic depth phase data, as compared to the constraints on depth provided by near-source readings, which are based on theoretical traveltimes derived from the local velocity model (Fig. 5).In practice, there is enough scatter in both data sets that we do not see a clear expression of this potential discrepancy for events at relatively shallow depths, as found for the South Rigan cluster.Twenty-seven events out of the 93 listed in Table 1 have depths that have been constrained by either near-source phase readings or by teleseismic relative depth phases.Of these 27 events, the depths range between 4 and 18 km with a median of 9 km (average 9.3 km).For the subset of 10 events in the 2010-2011 South Rigan earthquake sequence for which depth could be constrained, the range is 6-13 km, with a median of 9 km (average 9.4 km).The cluster default depth was set at 10 km.Fig. 6 shows the distribution of depths for those events with independent constraint.Although we constrain the centroid depths of the two South Rigan main shocks at 5 and 9 km (see Section 3.4) we do not have traveltime data with which to constrain the initiation depths of the two  1 with depth constraint provided by near-source readings or teleseismic relative depth phases.Black line is for the entire cluster.Red line is for the 2010-2011 South Rigan earthquake sequence (events 39-93 on Table 1).main shocks and both were set at 10 km depth for relocation.As larger earthquakes often nucleate close to the base of their ultimate rupture area, fixing the main shock nucleation depths deeper than the centroids is a reasonable approximation.

Body-waveform modelling
We used long-period body waveforms to constrain the source parameters of the 2010 and 2011 South Rigan earthquakes.Broad-band digital records from the Global Digital Seismograph Network were obtained from the Incorporated Research Institutions for Seismology data centre, and convolved with a filter that reproduces the bandwidth of the WWSSN 15-100 s long-period instruments.We then used the MT5 version (Zwick et al. 1995) of the algorithm of McCaffrey & Abers (1988) and McCaffrey et al. (1991) to invert the P and SH waveform data to obtain the strike, dip, rake, centroid depth, seismic moment and source-time function.We always constrained the source to be a double-couple and assumed an elastic half-space with V P = 6 km s −1 .
The result of our best-fitting solution for the 2010 December 20 and 2011 January 27 main shocks are shown in Figs 7 and  8, and are summarized on Table 2.The nodal-planes in the bestfitting solutions for both events are all tightly constrained by the available distribution of stations and both show almost pure strikeslip, with either left-lateral faulting on a northwest-southeast fault, or right-lateral slip on a northeast-southwest fault.The centroid depth of the 2010 event is at 5 km.The 2011 event is deeper, with a centroid depth of 9 km.These depths are consistent with the histogram of aftershock depths in Fig. 6.Apart from the centroid depth estimates, for which the CMT method is relatively insensitive, there are no significant differences between our solutions and the Global CMT solutions for these events.We note that the sources of the 2010 and 2011 earthquakes could be modelled as pure doublecouples so there is no evidence of volcanic influence.Two other notable events in the Rigan region (Nos.66 and 83 on Table 1) at Bodleian Library on April 30, 2013 http://gji.oxfordjournals.org/Downloaded from   were too small for us to model their source parameters using bodywaveforms and we instead use the Global CMT solutions for these events (http:/ww.globalcmt.org;Ekstrom et al. 2012).

InSAR analysis
Additional constraint on the sources of the 2010 December 20 and 2011 January 27 South Rigan earthquakes is provided from maps of the associated ground deformation based on InSAR analysis.Elastic dislocation modelling of the ground deformation allows us to distinguish between the fault and auxiliary plane for both events and to obtain a location for both faults that can be compared to the observed surface rupturing and distribution of epicentres.We describe the InSAR analysis for the two earthquakes separately below.InSAR data used are listed in Table 3.

2010 December 20 main shock
Radar acquisitions covering the December 20 main shock are available for one ascending and one descending Japanese Aerospace Exploration Agency (JAXA) ALOS track.The ascending PALSAR (Phased-array L-band SAR) track data was acquired in fine-beam or strip-map mode, and the descending track data was acquired in wide-beam or ScanSAR mode.The SAR data were processed from raw data products using the JPL/Caltech ROI_PAC software (Rosen et al. 2004).Some of the interferograms were unwrapped with the Stanford SNAPHU program (Chen & Zebker 2002).The ScanSAR data was converted to faux strip-map data using the zero-pulsepadding method (Tong et al. 2010) and then processed with the conventional strip-map processing.The interferograms were corrected for orbital effects using precise ALOS satellite orbits from JAXA, and topographic effects were removed using a 3-arcsecond SRTM DEM (Farr et al. 2007).
The ascending interferogram (Fig. 9a) shows a predominantly two-lobe pattern, with fringes in the eastern lobe elongated in a NE-SW direction and a sharp decrease in fringe spacing close to the boundary between the two lobes.The descending interferogram (Fig. 9d) shows an asymmetric four-lobe pattern that is also elongated in a NE-SW direction.Both of these interferograms are qualitatively consistent with motion on a fault striking NE-SW.Estimation of the fault model geometry shown in Figs 9(b) and (e) from the InSAR data was performed after subsampling of the unwrapped interferograms using a quadtree algorithm (e.g.Jonsson et al. 2002).The best-fitting uniform-slip rectangular dislocation in an elastic half-space (Okada 1985) was inverted for using Powell's conjugate gradient descent method with multiple Monte Carlo restarts (e.g.Powell 1964;Wright et al. 2003).The inversion routine converges easily on the best-fitting solution shown in Figs 9(b) and (e).This solution has an RMS misfit to the data of 1.5 cm, and replicates all major features of the data interferograms.Corresponding residuals between data and model are shown in Figs 9(c) and (f).Details of the best-fit solution are presented in Table 4.
The data were also inverted for variable slip on an array of rectangular 1×1 km fault patches, after fixing the fault geometry from the uniform-slip solution and extending the fault plane along-strike and down-dip (e.g.Funning et al. 2007).This variable-slip solution has a reduced RMS misfit of 1.2 cm.The slip distribution and associated errors, the resulting synthetic interferograms and the residuals are shown in Figs 9(g)-(l).The variable slip model includes a region of minor slip extending northwards from the end of the main rupture.

27 January 2011 aftershock
For the 27 January aftershock, SAR data are available for one ascending JAXA ALOS track, one descending German Space Agency (DLR) TerraSAR-X track and two Italian Space Agency (ASI) COSMO-SkyMed tracks, one ascending and one descending.All data for this event were acquired in strip-map mode.The precise orbits from JAXA, DLR and ASI were used for the orbital corrections.Processing of the ALOS data was as described above for the main shock and processing of the TerraSAR-X and COSMO-SkyMed data was similar, except that a 1-arcsecond version of the SRTM data was used for topographic calculations and the TerraSAR-X data was processed from SLC (single-look complex) images supplied by DLR.ALOS and COSMO-SkyMed interferograms were processed from raw data.Each dataset was sub-sampled to ∼570 data-points as described for the main shock.Interferograms, model interferograms and residuals for the ALOS ascending dataset, and the COSMO-Skymed descending dataset are shown in Fig. 10.The equivalent images for the COSMO-SkyMed ascending and TerraSAR-X descending datasets have similar viewing geometries, and so are presented only in the supplementary material (Fig. S3).Both the ascending and descending interferograms show four-lobe patterns, and because the aftershock is deeper and smaller than the main shock, there is some ambiguity about which nodal plane ruptured during the earthquake.However, subtle differences in fringe spacing in Figs 10(a) and (d) appear to favour a NW-SE orientation for the fault plane.
A joint inversion of all four datasets was performed to optimize the geometry of a uniform-slip rectangular fault plane, again as described for the main shock.The results from this confirm that the data can distinguish between the two nodal planes.An inversion with all parameters free converges on a NW-striking left-lateral solution.This is almost identical to the solution that is found when the strike and rake are fixed to the NW left-lateral fault plane from the Global-CMT solution.When the strike and rake are instead fixed to the NE right-lateral fault plane from the CMT solution, the inversion routine finds a solution with higher residuals and an rms misfit that is larger by 25 per cent.Inversion for variable slip was also performed with fault geometry fixed from the uniform-slip model.Although this model did not reduce RMS misfit significantly we present the results here as it most likely represents a more realistic slip distribution than the uniform-slip model.
We note that in the TerraSAR-X interferogram covering the 2011 January 27 event, a NE trending discontinuity in phase is observed that corresponds to ∼2 cm of slip (Fig. S3d).This lineament is coincident with the modelled fault extent from the December earthquake and possibly relates to minor near-surface afterslip resulting from the main shock.The first SAR pass used to form this interferogram was just 5 d after the main event, whereas in later SAR acquisitions used to form other interferograms this discontinuity in phase is much fainter, indicating that postseismic deformation from the December earthquake was limited and was mostly restricted to the first couple of weeks after the main shock.

Surface ruptures
We made brief field visits to the epicentral region on 2010 December 25-27th and on the 2011 March 6th.A range of cracks and fissures were observed following both the 2010 December 20 and 2011 January 27 main shocks.The winter of 2010-2011 in this part of Iran was rainy, and the surface ruptures were quickly eroded because of the action of water.By the time of our second visit in March 2011 the surface ruptures were badly degraded but still visible.The ruptures occurred in two distinct groups; a zone of en-echelon leftstepping ruptures at the northern edge of the Saif Al-Dini plain (Fig. 11), and a more discontinuous zone of right-stepping fractures at the southern margin of the plain (Fig. 12).
The left-stepping tension cracks that were mapped near the northern margin of the Saif Al-Dini plain trend individually 50-60 o within a zone with an overall trend of ∼40 o (Fig. 11).The zone of surface cracking was ∼30 m in east-west extent and was traced along strike for a distance of several kilometres (Fig. 11; locations listed in Table 5).These ruptures are very likely associated with the 2010 December earthquake; they lie along the surface projection of the fault modelled for that earthquake (e.g.Fig. 3) and their orientation and left-stepping arrangement are consistent with right-lateral slip on a fault striking ∼40 o .b).These cracks are typically right-stepping, and trend ∼80 o within a zone that strikes ∼90 o .A number of additional zones of surface cracking were observed in neighbouring parts of the plain during the brief field visit in 2011 March.As fissuring along the southern margin of the Saif Al-Dini plain apparently occurred during both the 2010 and 2011 earthquakes (as described in the next paragraph), we tentatively assigned the cracks to either the first, or second, earthquake based on their level of preservation (compare, for instance, Figs 12c and d).We interpreted two sets of relatively fresh cracks trending ∼60 o (at 28 • 10.712 N 58 • 04.542 E and at 28 • 10.466 N 58 • 04.741 E) to the second earthquake.We assigned other, more eroded, sets of cracks, which typically had an ∼80 o trend, to the 2010 earthquake (e.g.Figs 12a-c).
In the second field visit in 2011 March, local people led us to a group of discontinuous cracks (at 28 • 09.845 N 59 • 04.031 E) that developed during the second earthquake (location shown on Fig. 3).The cracks are discontinuous, and are also right-stepping, with a trend of ∼80 o within a zone that strikes ∼90 o (Fig. 12d).We note that these cracks, which are associated with the 2011 earthquake according to the information that we received from local people, lie along the best-fitting fault modelled for the 2011 earthquake.However, we also note that the east-west trend of the zone is inconsistent with the northwest-southeast trend of the fault plane as determined from InSAR and body-wave seismology.

Rupture process of the 2010-2011 South Rigan (Baluchestan) earthquakes
Our improved epicentre for the 2010 December 20 main shock (M w 6.5) lies close to the northern end of the NNE-SSW fault modelled assuming uniform slip from InSAR data, and indicates southward rupture propagation.The epicentre is located ∼4 km north of the modelled fault, which may result from minor amounts of slip occurring north of the end of the uniform-slip fault model (as shown in the dotted lines in Fig. 9h).The centroid depth of  5).en-echelon cracks observed near the northern margin of the Saif Al-Dini plain lie directly along the surface projection of the modelled fault and are likely to be tectonic in origin, though the small (∼5 cm) displacements across the surface cracks suggest that most of the slip failed to reach the surface; an interpretation that is consistent with the best-fitting InSAR model (Section 3.5).The cracks may have been enhanced during a period of shallow postseismic slip in the days following the main shock (as described at the end of Section 3.5.2).The earthquake was followed by diffuse aftershock activity around the main fault plane.
The combined use of ALOS, TerraSAR-X and COSMO-SkyMed RADAR imagery constrains the fault plane of the 2011 January 27 earthquake (M w 6.2) to the northwest-southeast left-lateral nodal plane.The InSAR analysis indicates slip from the near surface to depths of around 20 km.A relatively large depth-range of rupture is supported by the centroid depth of 9 km obtained from P and SH body-wave modelling.The epicentre of the 2011 event is consistent with a unilateral rupture from the western end of the fault, though the 90 per cent confidence ellipse for location does not rule out bilateral rupture propagation.Some of the cracks found along the southern margin of the Saif Al-Dini plain are consistent with the modelled fault location, but might also be explained by settling of basin sediments against the bedrock-bounded basin margin.The 2011 January 27 event was followed by a number of prominent aftershocks, with a slight elongation in the distribution parallel to the northwest-southeast trend of the fault.The most significant of these aftershocks was an M w 5.2 event at the western end of the InSAR modelled fault.The 2011 January 27 main shock was preceded by 1.5 hours by an event of M w 4.9 with an epicentre 5 km west of the main shock location.The depth distribution of ∼6-13 km for events in the South Rigan sequence is roughly consistent with the 1.4-12.1 km depth extent of the 2010 earthquake obtained from InSAR.The 2011 event appears to have ruptured over a larger range of depth, with some slip apparently occurring up to depths of around 20 km.Seismic slip at depths of up to 20 km are plausible in this part of Iran.For instance, aftershocks of the 2003 Bam earthquake extended to ∼20 km in depth (e.g.Jackson et al. 2006), as also indicated by one event included in our cluster from Bam with a hypocentral depth of 18 km (event 23 in Table 1).
We examine the potential influence of the first main earthquake on the extent of rupture in the second by calculating the Coulomb stress change using the Coulomb 3.1 code developed by the USGS (Lin & Stein 2004).We assume an effective coefficient of friction of 0.4 and a shear modulus of 3.2×10 10 Pa to match that used in the InSAR modelling.We take the uniform slip solution for the 2010 December 20 event and determine the change in Coulomb stress on receiver faults with the same orientation as the 2011 January 27 event (Fig. 13) at a depth of 10 km, given that we know the fault orientation of the 2011 January 27 aftershock event well from the InSAR data.The epicentre of the 2011 January 27 event lies in a region of large Coulomb stress increase (0.5 MPa).Moreover, the along-strike rupture extent of this earthquake to the southeast ends at the transition to a decrease in Coulomb stress arising from the first event, suggesting that the limit of the fault slip distribution for this aftershock was controlled by stress changes from the main shock.

G E O M O R P H O L O G Y O F T H E E P I C E N T R A L R E G I O N
A remarkable feature of both earthquakes is the lack of evidence in the geomorphology for prior fault movements.The right-lateral fault responsible for the 2010 December 20 event is situated within a region of extensive volcanism, mainly of Quaternary and Plio-Quaternary age, which may have disguised any cumulative displacement (e.g Fig. 3).The fact that the 2010 earthquake fault passes directly through the middle of a large volcanic centre is perhaps not coincidental, as the location of the vents may have been influenced by the presence of the fault.The 2010 earthquake fault is one of several right-lateral faults trending ∼40 o in the eastern Narmashir desert (Fig. 2; examined in more detail in Section 5).These faults are often spatially associated with volcanism, either aligned along their lengths, or focussed at extensional step-overs between fault segments (Fig. 3).
Surprisingly, we see no geomorphic evidence of faulting south of the end of the 2010 rupture in the alluvial surfaces of the Saif Al-Dini plain (Fig. 14a).We do not have any direct constraints on the age of the alluvial fan deposits in Saif Al-Dini, but various studies around eastern Iran have found the most recent period of regional alluvial fan deposition ended at ∼8-10 ka (e.g.Walker & Fattahi 2011, and references therein) and it is likely that the deposits date from at least this time.The lack of evidence for faulting in Saif Al-Dini may, therefore, be indicative of a very slow slip-rate and long recurrence interval between earthquakes on a southern continuation of the 2010 earthquake fault.However, it might also indicate that the southern end of the 2010 rupture marks the end of the fault.The Saif Al-Dini plain is separated at its eastern margin from another small plain by a narrow linear outcrop, with a NNE-SSW alignment, of Plio-Quaternary and Miocene volcanics (Fig. 3).Given the similarity in orientation of this outcrop and the modelled fault of the 2010 earthquake it is possible that the volcanism follows a southern segment of the fault that is slightly displaced to the east of the 2010 fault trace.
A possible northward continuation of the 2010 earthquake fault is visible in the geology and geomorphology (Fig. 3).This fault, which runs along a boundary between Eocene and Miocene volcanic rocks, is marked on the geological map of the region (Aghanabati 1992).It is slightly to the east of the 2010 rupture, but is directly in line with the NNE-SSW trending volcanic ridge that forms the eastern margin of the the Saif Al-Dini plain.
The northwest-southeast left-lateral fault responsible for the 2011 January 27 earthquake fault is not clearly expressed in the geomorphology.It is, however, close to the southern margin of the Saif Al-Dini plain suggesting that the margin of the plain may be fault controlled.Immediately east of the modelled fault of the 2011 January 27 main shock there is a clear east-west scarp along the southern margin of the Saif Al-Dini plain (Fig. 15a).This northfacing scarp is expressed only where remnants of older alluvial fans are preserved.No indications of left-lateral displacement are visible.
Beyond the western end of the 2011 fault, scarps trending eastwest and northwest-southeast are visible in the geomorphology (Figs 15b-d).A foreshock of the 2011 January 27 event (event No. 30; M w 4.9), located ∼6 km west of the 2011 main shock epicentre, is adjacent to an east-west north-facing scarp (Fig. 15b).The presence of vegetation and a small settlement immediately north of the scarp in Fig. 15(b) may be sustained by spring waters emanating from the scarp.A large aftershock (event No. 47; M w 5.2) occurred at the western tip of the modelled fault of the 2011 January 27 main shock, and adjacent to scarps that continue northwest, and with the same trend, as the fault responsible for the 2011 main shock (Fig. 3).This northwesterly fault zone is composed of two parallel scarps that form the margins of a basin that is ∼7.5 km long in the direction parallel to the 2011 rupture and ∼1.5 km wide (Fig. 15c).The southern of the two scarps does not show any definitive indications of late Quaternary faulting.The northern scarp, however, is visible as a south-dipping step in the surface of an alluvial fan (Fig. 15d).There is a broad region of fissuring aligned east-west where the northern fault crosses volcanic outcrops (in the lower-right corner of Fig. 15c).
The most obvious active faults in the epicentral region are a system of NE-SW trending faults ∼10-15 km southeast of the Saif Al-Dini plain (Fig. 3).These faults appear to be predominantly dip-slip with downdropping on their northwestern sides; adjacent to the Saif Al-Dini plain.The faults show a close spatial correlation with locations of volcanism.Numerous volcanic cones are aligned along the fault trends, either in the immediate hangingwall, footwall or on the surface fault trace (Fig. 3).Several of the faults show displacement within alluvium.In Fig. 16   and Shahsavaran mountains.The detailed investigation of their locations and source mechanisms (Section 3) yields constraints on the types and orientations of faults that are present.The examination of the geomorphology of the faults in Section 4, along with studies of the geomorphology of the Bam region (e.g.Jackson et al. 2006), are valuable in informing the search for other zones of active faulting within the region.Using the constraints provided from the earthquake studies we now provide a survey of active faulting within this part of southeast Iran.We then use these observations to address their roles in the tectonics (described in Section 6).A detailed description of all active faults within the region is beyond the scope of this paper.Instead, we concentrate on delineating the main styles of faulting though the description of a few type examples.We also concentrate on those faults that appear to be close to population centres.We separate our discussion into left-lateral and right-lateral faults.

Left-lateral faulting in the Shahsavaran and Barez mountains
The 2011 January 27 earthquake involved left-lateral rupture on a fault trending roughly NW-SE.Although geomorphic evidence for cumulative left-lateral faulting along this orientation is not easy to find in the epicentral region, subtle indications of this type of faulting do exist elsewhere within the Shahsavaran mountains.East of the immediate epicentral zone left-lateral faulting is visible along the northern margin of the Shahsavaran mountains on the southern slopes of Bazman volcano (Fig. 17a).The main fault zone trends E-W across the centre of the image.A number of shorter scarps trending E-W to NE-SW in mixed volcanic colluvium and alluvium are visible in the left-hand part of the image.The faults possess a down-to-the-north component of slip that has formed north-facing scarps.A possible left-lateral component of motion is also visible in the apparent displacement of volcanic units of Pliocene age (Nogole-Sadat 1996).In Fig. 17(a) we mark displacements of 1-1.5 km of two highly oxidised units (which show up as red in the band combination used and are labelled 'x' and 'z') and of an intervening unit that is black in the image (and labelled 'y').Southward-flowing rivers between these units also show sharp left-lateral deflections at the fault.
Active fault scarps oriented roughly WNW-ESE are visible in the geomorphology at the western end of the Barez mountains (∼50 km southeast of Jiroft, Fig. 2).We show an example of one of these faults in Fig. 17(b).The fault is visible as discontinuous scarps in alluvium.Vegetation and small settlements are sited along the downthrown northern sides of the faults.An anticline axis in basin sediments is exposed in the southern, upthrown, side of the fault suggesting that the dip-slip component is caused by shortening (Fig. 17b).Note that NW-SE trending faults, apparently with a considerable component of shortening across them, are also mapped along the northern margin of the Barez mountains (e.g.Jackson et al. 2006;Fig. 2).No obvious indications of left-lateral displacement are visible along the faults in the satellite imagery.
Given the occurrence of E-W to NW-SE faulting at widely spaced localities along the Shahsavaran and Barez range-fronts, we speculate that the left-lateral faulting along the range fronts may be more continuous than suggested from its expression in the   14a) and it is thus probable that additional active faults exist that we cannot identify in the geomorphology.
The right-lateral faults extend northwards in the gravel plains of the Narmashir desert (Fig. 2).The faults are segmented on a scale of 10-15 km, with the segments separated by extensional jogs that are often occupied by volcanic centres.NNW-SSE trending folds in Neogene and Quaternary deposits often link separate fault segments (Fig. 2).The oases of Rigan and Farad, and the surrounding desert regions, are shown in Fig. 18(a).Two NNE-SSW scarps are marked on the image.The scarps are visible from the marl deposits that are exposed on their western, uplifted, sides (the marls show up as light colours in the band combination used in the image) and by the concentration of vegetation close to the scarps.The Farad scarp is east-facing and is marked by a line of vegetation along it (Fig. 18b).
A detail of the fault scarp near the town of Rigan is shown in Fig. 18(c).The east-facing scarp trends NNE-SSW between the white arrows.Marl deposits (light-coloured in the satellite image) are exposed on the western (upthrown) side of the fault.Incision of eastward-flowing rivers, with the incision ending abruptly at the scarp, has occurred as a response to uplift west of the fault.Several of the river courses appear to be deflected southwards as they approach the scarp, which may indicate lateral fault growth.
The geomorphology of the Rigan scarp is reminiscent of the geomorphology of the Baravat fault escarpment close to Bam (e.g.Fig. 4a; Jackson et al. 2006).The Baravat escarpment, which had been identified as an active fault several decades before the 2003 earthquake (Berberian 1976), is ∼30-m-high and also shows deflection of stream channels around its southern end (e.g.Fig. 8 in Jackson et al. 2006).The Baravat fault accommodates only a small component of shortening across the overall Bam fault (e.g.Talebian et al. 2004;Berberian 2005;Jackson et al. 2006) but it is visible in the geomorphology because it has formed an east-facing topographic scarp and because it has exposed fine-grained marl deposits on its uplifted western side.The main Bam earthquake fault, which is almost pure right-lateral strike-slip, is situated ∼5 km west of the Baravat fault and has no geomorphic expression apart from the ruptures formed in 2003 (e.g.Jackson et al. 2006).
There are indications that spatial separation (partitioning) of strike-slip and dip-slip components of slip, similar to those described at Bam, also occur at Rigan.South of the prominent eastfacing scarp we identify a second possible fault scarp trending ∼350 • across alluvial gravels (running between the white arrows in Fig. 18d).The feature is notably different in appearance from nearby unpaved roads.There is also an apparent height change across it, with uplift on the western side suggested by a slight increase in the amount of shadowing (and hence incision) of stream channels.This southern scarp continues north of the view shown in Fig. 18(d) and overlaps with the main Rigan scarp in Fig. 18(c).Although there are similarities between the geomorphology at Bam and Rigan, we cannot say whether two parallel faults exist at Rigan, as they do at Bam.The similarities do suggest, however, that a significant strikeslip component is likely to exist on the Rigan fault; either as oblique slip on a single fault, or as spatial separation of dip-and strike-slip components onto parallel structures.
The Rigan scarp projects northward towards the Kahurak fault (Fig. 2).Particularly clear fault scarps are developed in playa lake deposits at the southern end of the Kahurak fault (Fig. 19).The two parallel scarps have a component of dip-slip, such that the region between them has subsided.This vertical component of slip, combined with a southwest-directed slope in the region shown in Fig. 19(b), has led to an apparent left-lateral displacement of the lake margin across the northern fault strand.It is likely that the faults possess a large component of right-lateral strike-slip, though direct evidence for this is lacking because of the rather subdued geomorphology.In Fig. 19

D I S C U S S I O N : T H E K I N E M AT I C S O F FAU LT I N G I N S E I R A N
The South Rigan earthquake faults, and the wider population of faults that we have described, are related to the accommodation of N-S right-lateral shearing across eastern Iran (e.g.Jackson & McKenzie 1984;Vernant et al. 2004).GPS measurements show that ∼15 mm/yr of N-S right-lateral shearing is accommodated across eastern Iran (Vernant et al. 2004;Fig. 1b).North of the epicentral region, this right-lateral shearing is accommodated on the N-S right-lateral faults of the Sabzevaran-Gowk-Nayband system at the western margin of the Dasht-e-Lut, Narmashir desert and Jaz Murian depression; and of the Sistan Suture Zone at the eastern margin of the Dasht-e-Lut, Narmashir desert, and Jaz Murian depression.
The set of right-lateral faults trending ∼40 o through the eastern Narmashir desert (as described in Section 5.2) form a horsetail  causative fault of the 2003 Bam earthquake and also includes the N-S scarps that we identify at the oases of Farad (Figs 2 and 18) and Narmashir (this latter scarp is not described in detail, but is shown on Fig. 2).This second group of right-lateral faults also appears to end southwards at the northern margin of the Shahsavaran mountains.
Faulting at the Shahsavaran mountain range-front, and within the interior of the range, appears to be predominantly on faults parallel to the strike of the range (Section 5.1).The left-lateral nature of these faults is highlighted in the 2011 earthquakes and also by cumulative left-lateral displacements visible in a few locations.It is important to note that further west, the orientation of the faults within the Barez range is ∼NW-SE, which, along with their association with folding in Tertiary rocks, indicates a component of reverse faulting.Adjacent to the Shahsavaran mountains, however, the orientation of the faults is E-W to NE-SW; indicating a component of extension.The band of discontinuous left-lateral faults running through the margins and interior of the Shahsavaran mountains is reminiscent of the morphology of the Dasht-e-Bayaz and other E-W left-lateral faults encountered in NE Iran (Ambraseys & Tchalenko 1969;Walker et al. 2004;Walker et al. 2011) and it is possible that they aid the accommodation of regional N-S right-lateral shearing across eastern Iran through clockwise vertical axis rotation.
Although the slip on the Kahurak fault ends north of the Shahsavaran mountains, the activity on the Nosratabad fault of the Sistan Suture Zone continues at least as far south as the city of Iranshahr, at latitude 27 • 15 N (Fig. 2).Roughly 25-km east of Iranshahr three short scarps are visible in apparently late Quaternary alluvium (these scarps are marked on Fig. 2 In the west of the region shown in Fig. 2, faulting continues south of the Barez mountains in the form of the Sabzevaran fault (e.g.Regard et al. 2005), but along the western margin the slip-rate on the major faults appeared actually to increase southwards (e.g.Fig. 2).Regard et al. (2005) estimate a right-lateral slip-rate of 4.0-7.4mm yr −1 on the Sabzevaran fault from the displacement of alluvial fans of late Quaternary and Holocene age, whereas Walker et al. (2010), from displacements in lake sediments of early Holocene age, estimate a minimum slip-rate 3.8 ± 0.7 mm yr −1 for the Gowk fault.

C O N C L U S I O N S
The 2010 and 2011 South Rigan earthquakes are instructive in demonstrating how tectonic strain is accommodated in SE Iran through a combination of distributed NE-SW right-lateral strike-slip and SE-NW to E-W left-lateral faulting.They are also instructive in showing that the low-lying desert regions between the major fault Gowk-Nayband and Sistan fault systems, display behaviour that is far from the aseismic rigid behaviour that is often attributed to them.The 2010 and 2011 earthquake faults are similar to the causative fault of the destructive 2003 Bam earthquake in that they do not have any clearly identifiable effect on the geomorphology, and it is likely that faults other than those identified in our reconnaissance exist.Bam is not the only settlement in the Narmashir desert; our investigation of the geomorphology suggests that other oasis settlements are located close to active fault scarps.The close relationship between sites of habitation and active faults is one that is seen across desert parts of Iran, and is caused by the shallow water-tables and springs along faults in otherwise arid basins (Jackson 2006).The slip-rate on any individual active fault within the Narmashir desert is unlikely to be high, which goes some way towards explaining why the presence of many of the faults have not previously been recognized.

Figure 1 .
Figure 1.(a) Map of Iran showing epicentres of shallow earthquakes from the catalogue of Engdahl et al. (1998).Z, Zagros; A, Alborz; K, Kopeh Dagh.(b) Map of Iran with GPS velocities of points relative to Eurasia from Vernant et al. (2004).Both maps are in a Mercator projection.

atFigure 2 .
Figure 2. SRTM digital topography (e.g.Farr et al. 2007) of southeast Iran showing faults, major geographic features, and population centres (white squares).The Shahsavaran mountains are predominantly composed of Tertiary and Quaternary volcanic rocks.The main active volcanic centre is Bazman, though smaller volcanic centres are scattered both along the northern slopes of the Shahsavaran mountains and in the alluvial plain to the north of it.The limits of outcrops of Palaeozoic and Mesozoic rocks in the southern Shahsavaran mountains are highlighted.The causative faults of the 2010 December 20 and 2011 January 27 South Rigan earthquakes are shown in red.Fault-plane solutions of major shallow (centroid depth < 35 km) earthquakes are from waveform modelling (black; listed in Jackson 2001;Jackson et al. 2006; and this paper), the Global CMT catalogue (dark grey; http://www.globalcmt.org/), and from first-motions (light grey; listed inMcKenzie 1972).Earthquakes of the Bam and South Rigan sequences are plotted at the locations listed in Table1.All other focal mechanisms are plotted at their Global CMT locations(Ekstrom et al. 2012).Fault slip-rates are fromRegard et al. (2005) andWalker et al. (2010).This map is in a Mercator projection.

Figure 3 .
Figure 3. LANDSAT ETM image (RGB 541) of the epicentral region of the 2010 December 20 and 2011 January 27 South Rigan earthquakes.The best-fitting faults from InSAR analysis and elastic dislocation modelling are shown as red lines.Focal mechanisms are as displayed in Table2.Epicentres of smaller earthquakes are shown as red circles.All epicentral locations are those derived from HDC analysis and are displayed with their 90 per cent confidence ellipses for relative location (see Section 3.2; listed in Table1).The locations of observed surface breaks from the first earthquake are shown as blue circles and those from the second earthquake as green circles.Note that the central section of ruptures originated from both the 2010 and 2011 earthquakes, and the blue circles are partly obscured by the overlying green symbols.Faults suspected as active in the region are shown as dashed black lines.The region is dominated by Quaternary and late Tertiary volcanism.Small basins filled with alluvial fans and small salt pans are sited between volcanic outcrops.Volcanic vents are distinguished by oxidised scoria, which appears red in the band combination used in the figure.This map is in a local zone UTM projection.

Figure 4 .
Figure 4. (a) The cluster of locations listed in Table1was calibrated by fixing the three main shock epicentres (in yellow, with their 90 per cent confidence ellipses) to lie on the faults-shown as green lines-modelled from uniform-slip inversion of InSAR data for the Bam (fromJackson et al. 2006)  and South Rigan earthquakes (Section 3.5).The Bam fault model ofJackson et al. (2006)  includes a subevent on a west-dipping reverse fault projecting to the surface at the Baravat escarpment.The epicentre of the 2010 South Rigan main shock lies ∼4 km north of the modelled fault for that earthquake.This apparent mislocation is likely to result from rupture extending north of the limit that is not present in the uniform-slip model (see variable-slip model in Fig.9).(b) Relocated events in the South Rigan sequence with their 90 per cent absolute location uncertainties.Epicentres are shown on a LANDSAT ETM image (band 5) and are numbered as in Table1.The two main shocks are shown in yellow to distinguish them from other seismicity (in red).

Figure 5 .
Figure 5. Observed phase arrivals and traveltimes calculated from the local velocity model used for origin time calibration (Section 3.3).Traveltime curves are shown for both P (Pg in red, Pn in green) and S (Sg in red, Sn in green) phases.P phases are indicated by an 'x', S phases by circles.The larger circles are S-P phases derived from accelerometer stations that do not have calibrated timing systems; they are plotted by adding the S-P time to the theoretical P arrival time at that distance.For all phases, theoretical traveltimes are shown (assuming they exist) for three source depths, the average depth of events in the cluster (solid line), and the minimum and maximum depths (dashed lines).

Figure 6 .
Figure 6.Histogram of depths for events in Table1with depth constraint provided by near-source readings or teleseismic relative depth phases.Black line is for the entire cluster.Red line is for the 2010-2011 South Rigan earthquake sequence (events 39-93 on Table1).

Figure 7 .
Figure 7. Minimum misfit solution for the 2010 December 20 South Rigan main shock (M w 6.5) determined from inversion of P and SH body-waveforms.The top hemisphere represents the P-wave radiation and the bottom hemisphere represents the SH waveforms.Station positions on the focal spheres are identified by capital letters and arranged clockwise from north.Waveforms for each station (identified by a three-or four-letter code, and a capital letter corresponding to those within the focal spheres) are positioned azimuthally around the focal hemispheres.Observed waveforms are shown as solid lines and synthetic waveforms as dashed lines.Numbers beneath the header line are strike, dip, rake, centroid depth (in kilometres) and moment (in Nm).The source-time function and waveform timescale are plotted between the two focal spheres.The synthetics are calculated for a half-space of V P 6.0.The source parameters are listed in Table2.

Figure 8 .
Figure 8. Focal mechanism for the 2011 February 27 event (M w 6.2) determined from inversion of P and SH body-waveforms.Notation is the same as in Fig. 7.The source parameters are listed in Table 2.

Figure 9 .
Figure 9. (a) Interferogram constructed from ascending ALOS scenes from 2010 September 30 and 2010 December 31 and imaging ground deformation from the 2010 December 20 South Rigan earthquake.(b) Best-fitting fault model assuming uniform slip on a rectangular dislocation in an elastic half-space.The fault model is from a joint inversion of two interferograms (ALOS ascending and ALOS descending) and is here shown in the viewing geometry for the ascending ALOS interferogram.(c) Residuals between the data and the uniform-slip model.(d) Descending ALOS interferogram constructed from scenes from 2010 July 13 and 2011 January 13.(e) Best-fitting fault model from joint inversion of four interferograms shown in the viewing geometry for the descending ALOS interferogram.(f) Residuals between the data and the uniform-slip model.(g) Best-fitting slip distribution on fault plane from variable-slip fault model.The region of rupture from the uniform-slip model is outlined by the black dashed rectangle.(j) 2-sigma errors for the slip distribution from Monte Carlo analysis, shown on the same colour scale.(h), (i), (k), (l) Same as (b), (c), (e), (f) but for the variable-slip model.

Figure 10 .
Figure 10.(a) Interferogram constructed from ascending ALOS scenes from 2010 December 31 and 2011 February 15 and imaging ground deformation from the 2011 January 27 South Rigan earthquake.(b) Best-fitting fault model assuming uniform slip on a rectangular dislocation in an elastic half-space.The fault model is from a joint inversion of four interferograms (ALOS ascending, COSMO-SkyMed ascending and descending and TerraSAR-X descending) and is here shown in the viewing geometry for the ascending ALOS interferogram.(c) Residuals between the data and the uniform-slip model.(d) Descending interferogram constructed from descending COSMO-SkyMed scenes from 2011 January 08 and 2011 February 09.(e) Best-fitting fault model from joint inversion of four interferograms shown in the viewing geometry for the descending COSMO-SkyMed interferogram.(f) Residuals between the data and the uniform-slip model.(g) Best-fitting slip distribution on fault plane from variable-slip fault model.The region of rupture from the uniform-slip model is outlined by the black dashed rectangle.(j) 2-sigma errors for the slip distribution from Monte Carlo analysis, shown on the same colour scale.(h), (i), (k), (l) Same as (b), (c), (e), (f) but for distributed fault model.

Figure 11 .
Figure 11.Field photographs of ruptures in the northern part of Saif Al-Dini arising from the first earthquake.All photographs were taken early 2011 March.The locations are in degrees and decimal minutes.See Fig. 3 for approximate locations.(a) Left-stepping en-echelon fissures (pointed out by white arrows).(b) View south from the same location as in 'a'.Although degraded, the en-echelon fissures can still be picked out by the accumulations of light-coloured silt along them.(c) Distributed fissuring (white arrows point towards six parallel fissures.Figure is circled for scale (d) Close-up of one of the zones of fissuring slightly north of those shown in 'c'.During the 2010 December field survey we also observed discontinuous cracks close to the southern margin of the Saif Al-Dini plain near 28 • 10.601 N 58 • 04.676 E (Figs 12a-b).These cracks are typically right-stepping, and trend ∼80 o within a zone that strikes ∼90 o .A number of additional zones of surface cracking were observed in neighbouring parts of the plain during the brief field visit in 2011 March.As fissuring along the southern margin of the Saif Al-Dini plain apparently occurred during both the 2010 and 2011 earthquakes (as described in the next paragraph), we tentatively assigned the cracks to either the first, or second, earthquake based on their level of preservation (compare, for instance, Figs12c and d).We interpreted two sets of relatively fresh cracks trending ∼60 o (at 28 • 10.712 N 58 • 04.542 E and at 28 • 10.466 N 58 • 04.741 E) to the second earthquake.We assigned other, more eroded, sets of cracks, which typically had an ∼80 o trend, to the 2010 earthquake (e.g.Figs12a-c).In the second field visit in 2011 March, local people led us to a group of discontinuous cracks (at 28 • 09.845 N 59 • 04.031 E) that developed during the second earthquake (location shown on Fig.3).The cracks are discontinuous, and are also right-stepping, with a trend of ∼80 o within a zone that strikes ∼90 o (Fig.12d).We note

Figure 12 .
Figure 12.Field photographs of ruptures in the southern part of Saif Al-Dini arising from both the 1 st and 2 nd earthquakes.See Fig. 3 for approximate locations.(a) Photos taken in late 2010 December of fresh left-stepping cracks from the 1st earthquake.Compass, circled, for scale.(b) Detail of one of the cracks with slight vertical component.(c) Fissures close to those shown in 'a' and 'b' but photographed in early 2011 March.The cracks have been heavily damaged by rain but are still identifiable from the concentration of light-coloured fine deposits along them.(d) Fresh cracks in the southern part of the plain that we were informed were caused by the second earthquake (photograph taken in 2011 March).The differences between more and less degraded cracks as shown in 'c' and 'd' were used to assign additional zones of cracking mapped in the March field visit to either the 2010 or 2011 earthquakes (listed in Table5).

Figure 13 .
Figure 13.(a) Coulomb stress change (MPa) at 10 km depth because of the 2010 December 20 event (white line) resolved onto receiver faults with the orientation of the 2011 January 27 event (black line).Relocated aftershocks (Table 1) from the 2010 December 20 event (black star) up to the 2011 January 27 event (grey star) are shown as black circles (error ellipses are omitted for clarity but are shown in Fig. 3 at the same scale).(b) Coulomb stress change at 10 km depth because of both the 2010 December 20 and 2011 January 27 events resolved onto receiver faults with the orientation of the 2011 January 27 event.Relocated aftershocks after the 2012 January 27 event are show as grey circles.
we show the exceptionally well-preserved northwest-facing scarp of one of these faults.A very clear scarp is visible in an abandoned alluvial fan surface in the lower-left part of the image.There is a second very clear scarp in the upper-right part of the image where the fault is crossed by an eastward-flowing ephemeral river.The presence of three river terraces on the eastern, upthrown, side of the fault is indicative of multiple events.The NE-SW orientation of the faults, along with their proximity to N-S right-lateral and E-W left-lateral faults, suggests that they are normal faults.

Figure 14 .
Figure 14.Colour SPOT 5 imagery (2.5 m pixels) showing right-lateral faults at the northern margin of the Shahsavaran mountains.(a) Alluvial fan in the Saif Al-Dini plain at the southern end of the 2010 fault modelled by InSAR.The southward projection of the 2010 South Rigan earthquake fault runs approximately between the two white arrows.There is no indication of active faulting in the geomorphology.See Fig. 3 for location.(b) a N-S scarp cutting through alluvial fans in the desert east of the 2010-2011 epicentral zone.Location shown on Fig. 2. (c) Detailed view of the scarp in 'b'.Possible right-lateral stream deflections are circled.

Figure 15 .
Figure 15.Colour SPOT5 imagery showing evidence of late Quaternary faulting along the trend of the 2011 South Rigan earthquake.Faults are marked at their ends by white arrows.All locations are shown on Fig. 3. (a) North-facing scarps along the southern margin of the Saif Al-Dini plain.Scarps (marked by black arrows) are only visible in remnants of older fan deposits.(b) North-facing scarps in bed-rock.Late Quaternary movement is visible near the centre of the image, where there is a right-step between two overlapping segments.There is a scarp in light-coloured fan material at the eastern end of the northern fault segment (marked by black arrow).(c) Possible faulting along strike of the western end of the 2011 rupture.Volcanic rocks in the lower-right corner of the image are fractured on an east-west trend.There is an overall WNW-ESE trend to the topography in this region, with a narrow basin bounded both to north and south by steep scarps in volcanic rocks.(d) Detail of the northern WNW-ESE scarp in 'c' showing apparent uplift and incision of fan deposits north of the scarp.

Figure 16 .
Figure 16.SPOT5 image with black arrows pointing to a scarp crossing alluvial fans southeast of the 2010-2011 South Rigan epicentral zone (see Fig. 3 for location).A series of higher river terraces on the southern, upthrown, side of the fault may indicate earlier periods of fault slip.Northeast of this image the fault is expressed as a series of open fissures within the surface of a lava flow.
(b) we show a close-up view of the southern fault scarp.Incision of streams southeast of the scarp is indicative of uplift of the southeastern side of the fault.A possible, though not definitive, right-lateral stream deflection is visible (circled region on Fig. 19b).

Figure 17 .
Figure 17.(a) LANDSAT ETM image (bands 5,4,1 displayed as R,G,B) of east-west active faults along the southern slope of Bazman volcano (see Fig. 2 for location).White arrows mark the ends of the faults.All the faults have a dip-slip component shown by north-facing scarps.A prominent sequence of Pliocene volcanic units (labelled 'x', 'y', and 'z') appear to have been displaced left-laterally by 1-1.5 km across the main fault strand (displacements shown by black arrows).(b) SPOT5 image (from GoogleEarth) of a WNW-ESE fault scarp southeast of Jiroft (see Fig. 2 for location).Small villages and vegetation are aligned along the fault.Folding on the southern, upthrown, fault block indicates a component of shortening.
Figure 18.(a) LANDSAT ETM image (RGB 541) of the region around Rigan and Farad villages north of the 2010-2011 epicentral zone (see Fig. 2 for location).Two zones of active faulting are visible in the image.Both are associated with clusters of vegetation (green in this band combination), and both expose light-coloured marl deposits on their uplifted western sides.(b) SPOT5 image of the east-facing fault scarp at Farad village.Patches of vegetation (red in this band combination) along the scarp are marked by black ovals.(c) Southern end of the east-facing scarp at Rigan village.Eastward flowing streams are deflected southwards around the end of the scarp.Vegetation (in red) is concentrated along the scarp (within the oval).This scarp is reminiscent of the Baravat escarpment near Bam (e.g.Jackson et al. 2006).(d) Apparent north-south scarp in the desert south of Rigan.A slight increase in shadowing on the western side of the scarp suggests that it faces to the east.It has a different appearance to nearby unpaved roads and is visible only in older (darker) fan material.

Figure 19 .
Figure 19.(a) SPOT5 image of fault scarps cutting through playa deposits northeast of Rigan town (see Fig. 2 for location).The northern fault is uplifted on its northwest side (generating an apparent left-lateral displacement of the playa margin).The southern fault is uplifted to the south.(a) Close-up of the southern scarp.Vegetation (in red) is concentrated in stream-beds and along the base of the west-facing scarp.One stream (circled) shows a possible right-lateral deflection of 10-15 m.These scarps are close to the estimated epicentre of an earthquake in 1923.
).A series of roughly N-S trending folds in Neogene and Quaternary rocks directly north of Iranshahr also suggest recent activity in this region (Nogol-e-Sadat 1996; Fig. 2).The east-west fold axes within the Makran accretionary prism are not displaced right-laterally by large amounts south of the Sistan Suture Zone indicating that no significant N-S right-lateral shearing occurs south of latitude ∼27 • N.Although there are no firm constraints on the slip-rates of any individual faults within the southern part of the Sistan Suture Zone, the termination of the Karurak fault at the latitude of the Shahsavaran mountains and the southward reduction in the geomorphic expression of the Nosratabad fault, point towards a southward reduction in the overall rate of right-lateral strike-slip across the Sistan Suture Zone.A latitudinal variation in slip-rate might explain the normal component at Bodleian Library on April 30, 2013 http://gji.oxfordjournals.org/Downloaded from of faulting observed in the eastern Shahsavaran mountains, with extension required to compensate for any difference in right-lateral shearing across those parts of the Suture Zone south of, and north of, the Shahsavaran mountains.

Table 2 .
Source parameters of instrumentally recorded earthquakes in the Rigan region that have been modelled using body waves (Section 3.4) or for which centroid moment tensor solutions exist (http://www.globalcmt.org).The two body-waveform solutions are shown in bold.

Table 3 .
SAR data analysed for this this study.Date1 and Date2 are the dates of the master and slave images in the interferograms.Letter A before track number is for ascending tracks, D for descending tracks.FB is fine beam and WB is wide beam (ScanSAR), subswath is noted in parentheses.TSX stripmap beam number is shown, and CSK central look angle is shown in the mode column.Bperp is the perpendicular component of the baseline between the two orbits at the centre of the swath at the top and bottom of the interferogram.

Table 4 .
Wright et al. 2003)f the two South Rigan earthquakes obtained from elastic dislocation modelling (see Section 3.5).Parameters are given for a one-fault model.The errors are derived from Monte Carlo error estimation (e.g.Wright et al. 2003)and are reported at the 1-sigma level.Full details of uncertainties and tradeoffs are shown in Figs S1-S3.

Table 5 .
Zones of surface cracking and fissuring observed after the 2010 and 2011 South Rigan earthquakes.All coordinates were recorded during the 2011 March 6 field survey.Latitudes and longitudes are given in degrees and decimal minutes.