Coronavirus (Covid-19): Latest updates and information
Skip to main content Skip to navigation

Evaluating the Style of Continental Deformation in Eastern Turkey using InSAR and GPS

Ryan Lloyd[1], School of Earth and Environment, University of Leeds


The way in which continents deform has implications for the potential magnitude and distribution of seismic hazards (earthquakes). I used Interferometric Synthetic Aperture Radar (InSAR) and GPS data to model the motion of East Anatolia. This data was then used to calculate the distribution of strain accumulation across the region. I found the strain accumulation was up to 20 nanostrain per year greater within ~30km of the major fault structures than across the Anatolian plateau. In addition, elevated strain accumulation is predicted up to ~85km west of the Karliova Triple Junction, which is where two major faults meet.

Deviations from the velocity predicted by modelling Anatolia as a rigid block coincide with regions of elevated strain accumulation. Block modelling predicts ~10 mm/yr of convergence across the North Anatolian Fault that is not seen in the data. This is interpreted as evidence for non-rigid behaviour within the Karliova triple junction.

I concluded that the Anatolian plate acts, overall, as a rigid block, except in the triple junction. This finding suggests that the greatest seismic hazard is localised along the North and East Anatolian Faults.

Keywords: ENVISAT, InSAR, GPS, earthquakes, Anatolia, Karliova Triple Junction.


It is expected that in the twenty-first century 2.57±0.64 million people will be killed by earthquakes worldwide (Holzer and Savage, 2013: 155). Earthquakes can occur wherever regions of the earth move relative to each other. This condition is most obviously met at plate boundaries, where large-scale narrow discontinuities separate the earth's tectonic plates. Earthquakes are also observed within continental interiors, however, and the region over which they occur here can be relatively broad. The relative motions within continents are also often smaller compared to plate boundaries and so the conditions required for an earthquake will be met less frequently (Liu et al., 2011: 128-132). The infrequency of large earthquakes within continents has meant many hazardous continental regions are unknown, leaving the population in these areas unaware of the potential seismic risk (England and Jackson, 2011: 348).

Figure 1: Tectonic setting of the Anatolian region

Figure 1: Tectonic setting of the Anatolian region.

The red box highlights the Karliova triple junction, also shown are the right-lateral North (NAF) and left-lateral East (EAF) Anatolian Faults. GPS vectors were from several studies, with respect to a fixed Eurasia (Aktug et al., 2012; Tatar et al., 2012; Yavaşoğlu et al., 2011; Alchalbi et al., 2010; Ozener et al., 2010; Aktug et al., 2009; Reilinger et al., 2006). Error ellipses indicate 95% confidence bounds. GPS vectors show an east-west velocity gradient and counter-clockwise rotation corresponding to extension across Western Anatolia.

In this article the strain accumulation directly across the triple junction region in Eastern Turkey, where the North and East Anatolian Faults meet (Figure 1) is quantified to investigate how and where the ground is deforming. Estimations of the distribution of strain accumulation across the North Anatolian Fault (NAF), East Anatolian Faults (EAF) and the Anatolian plateau have already been made by Walters (2012), who found that it is concentrated around the faults, as expected for rigid deformation. This article investigates whether the predictions made by Walters (2012) of strain rates and the style of deformation within the triple junction are correct given the inclusion of InSAR data from across it. It was expected that strain rates would be low (Walters, 2012), so InSAR was used in conjunction with GPS data to measure ground deformation (Figure 2). This is the first time the strain accumulation directly within triple junction has been investigated using InSAR and GPS data. Predictions of the velocity field, the rate and distribution of ground motion, by a block model by Walters (2012), however, found that where the major faults meet, rigid deformation produced residuals of up to 10mm/yr, hinting at non-rigidity. This article compliments Walters (2012) by including InSAR data from across the triple junction region to test whether reduced uncertainty influences the fit of block modelling. The extent and magnitude of any misfit gives an indication of regional departures from rigidity. The consequence of this is to determine whether the seismic hazard in Eastern Anatolia is confined to the major faults (as predicted by block modelling) or distributed diffusely.

Figure 2: Geographic distribution of the data used in this article, part 1

Figure 2: Geographic distribution of the data used in this article, part 2

Figure 2: Geographic distribution of the data used in this article.
Top: The blue box indicates the descending InSAR track (493) coverage processed for this article, red box the complimentary ascending track (357)
Bottom: Eastern Anatolian region with GPS vectors (ellipses indicate 95% confidence bounds) from several studies (Aktug et al., 2012; Yavaşoğlu et al., 2011; Alchalbi et al., 2010; Ozener et al., 2010; Reilinger et al., 2006) that were used in this report. Black lines show major faults in the region (Şaroğlu, 1992).

The Earthquake Cycle

An earthquake is the sudden release of elastic energy. The accumulation of this elastic energy within the crust comes from stresses exerted by distant forces over a period of time. The amount of accumulating energy, and the time required for it to build up, will be different for every earthquake. The earthquake occurs as friction along the fault is overcome and results in a rupture of the earth (Wright, 2002: 2873-88). This rupture may occur at any depth within the seismogenic crust and sends seismic waves outward from the epicentre. The repetition of earthquakes at a given fault is termed the earthquake cycle (Reid, 1910).

The earthquake cycle contains three separate periods. The first, interseismic, describes the slow accumulation of energy and deformation prior to an earthquake. The release of this energy is the coseismic period. After an earthquake, the ground will deform again as it relaxes: this is the postseismic period (Wright, 2002: 2873-88).

From the accepted assumption that earthquakes occur cyclically following the accumulation of strain throughout the interseismic period, one may suppose that if the build-up of strain can be quantified and given an expected critical value, then the time, magnitude and location of an earthquake can be estimated. To make this estimation one must also take into consideration that factors including the initial strain, the rate of strain accumulation and release, fluids present, fault geometry, the frictional force dynamics and also the critical value must be known. These parameters may also vary within and between earthquake cycles. Cumulatively, these unknowns currently prevent the prediction of earthquake occurrence and recurrence.

However, by measuring ground motion over long periods of time it is possible to determine where strain is accumulating and hence where an earthquake may occur. Spatially continuous measurements of this interseismic strain accumulation can be precisely made using InSAR (Walters et al., 2011; Wright et al., 2001), which together with GPS, was how I quantified the ground deformation across Eastern Turkey.

Modelling continental deformation

There are two end member descriptions of continental deformation: block and continuum. Block-like deformation holds for many continental regions (Floyd et al., 2010; Avouac and Tapponnier, 1993) where strain is focused along narrow zones, but an alternative approach has also been suggested where this is not the case: continuum, or fluid, deformation (England and McKenzie, 1982: 295-321). This approach chooses to diverge from the constraints of rigidity, and allows plates to flow like a viscous sheet under external boundary conditions and internal buoyancy forces. This has the effect of spreading out the strain accumulation which would otherwise build up at block boundaries.

Tectonic regime of Eastern Anatolia

The 1200km NAF separates the Eurasian and Anatolian plates. Studies using GPS data generally agree that slip rates of the NAF are 24±1mm/yr (McClusky et al., 2000). Its conjugate, the EAF, is 600km long and separates Anatolia from the Arabian plate. Several studies found the geodetic slip rate of the EAF is ~10mm/yr (Walters, 2012; McClusky et al., 2000). These two faults intersect to form the Karliova Triple Junction (KTJ) (Figure 1). The high slip rates on the faults, as well as recent seismic activity, indicate that Karliova is an important region for investigation into seismic hazard.

Regionally, studies such as Reilinger et al. (2006) have used large blocks to describe successfully the kinematics over the greater Middle Eastern region. Focusing on Anatolia, Ozener et al. (2010: 1167) have hypothesised using GPS that strain is accumulating along secondary faults between the NAF and EAF. This supports Chorowicz et al. (1999: 531), who state that Eastern Anatolia is comprised of blocks less than 50km wide, and Aktug et al. (2013) who invoke four large blocks to model the region. The approach of using rigid blocks however breaks down at the vertex of the triple junction, else the existence of 'gaps' would occur, which is forbidden.

Chorowicz et al. (1999) suggest that the blocks are undergoing some internal deformation, primarily extension, to accommodate the necessary changes in surface area to remove the gap. From seismic data Aktug et al. (2013) also conclude that some deformation must be occurring around ~5mm/yr across Anatolia. These conclusions suggest that continuum models of deformation, characterised by more diffuse strain accumulations, may better describe the deformation of Anatolia. It is specifically this distinction between whether Eastern Anatolia undergoes block or continuum deformation that I investigated using high-resolution geodetic techniques (Figure 2).

InSAR Methodology


In this article Synthetic Aperture Radar (SAR) images acquired by the European Space Agency's ENVISAT satellite between 2002 and 2010 were used. The Advanced Synthetic Aperture Radar equipment on board used C-band microwaves (wavelength 5.6cm) as its active illumination source, and had a repeat time of 35 days. The SAR images were acquired at an incidence angle of 23˚ (to the centre of the image), with swath width of 105km.

Interferometric Synthetic Aperture Radar (InSAR)

SAR is a satellite-based remote sensing technique. It is a radar system, using the two-way travel time of electromagnetic radiation to determine the distance between the satellite and a reflector (Wright, 2002) (Figure 3). Signals which reflect off the same point, but are acquired successively as the satellite moves, are combined to produce a large synthetic aperture and much greater spatial resolution (Massonnet and Feigl, 1998: 441-500).

Figure 3: Schematic diagram of satellite-ground geometry used for SAR imagery

Figure 3: Schematic diagram of satellite-ground geometry used for SAR imagery. At each time, t, the satellite is in a different location and so its swath of radio waves illuminates a different footprint on the ground. These footprints partially overlap; the purple regions show an example of how the satellite sees one location in this way to construct a SAR image, eventually the whole track will be covered by an overlap and a SAR image can be constructed.

Each SAR image contains a record of the phase and magnitude of the returning signal. Using the phase differences between a pair of SAR images, transmitted and reflected from the same locations, the one-dimensional change in the radar line-of-sight (LOS) can be determined to the precision of a fraction of the signal wavelength. This record of the phase change is an InSAR image. The displacement can then be related to the ground velocity using the temporal separation of each SAR image pair. For further detail about the technique of producing and using InSAR images see Bürgmann et al. (2000: 169).

Contributions to phase delay

To use InSAR to measure interseismic deformation I had to ensure the observed phase changes were caused by ground deformation alone. However, there are several other factors that can introduce a phase shift, \Phi, in addition to ground motion. For this article they are considered as noise, and are summarised in Equation 1 and Table 1. Each contribution had to be taken into consideration and removed when processing the InSAR data.


\Delta\Phi_{\textnormal{observed}} Total phase change recorded in an interferogram.
\Delta\Phi_{\textnormal{deformation}} The signal: caused by ground movement towards or away from the satellite along its line of sight between successive SAR image acquisitions.
\Delta\Phi_{\textnormal{atmospheric}} Phase contribution caused by interaction with the atmosphere. Water vapour usually provides the largest phase delay, causing incoherence near the coast and during times of high humidity.
\Delta\Phi_{\textnormal{orbital}} The relative location of the satellite is different for each orbit, and to cm precision (Zandbergen et al., 2003) the change between orbits can be known and accounted for. However, the deformation signal within an individual interferogram is smaller than the precision in the satellite location. This introduces an unknown phase contribution, an orbital error, to each SAR image.
\Delta\Phi_{\textnormal{topographic}} A topographic error is an addition to the path length as a result of varying incidence angles with the ground. This error is larger for steeper topography and greater perpendicular baselines.
\Delta\Phi_{\textnormal{error}} The physical ground surface between acquisition dates may change due to vegetation or land use, which will result in the phase at such pixels to change and, when considered with their neighbours, to be incoherent.

Table 1: Contributions to the phase delay within an InSAR image.


GPS is an accurate method for measuring ground motions (Reilinger et al., 2006) at one discrete point of acquisition. The GPS data for this article is from 162 ground-based station locations throughout Eastern Anatolia (Figure 2). This data is from several separate studies and technical details on collection and processing can be found in the respective papers.

Data Processing


SAR images used to produce interferograms are available from the European Space Agency (ESA). The images selected for processing had the largest coverage of the triple junction and the region where the greatest misfit between the block model and velocity field from Walters (2012) is observed.

The ESA provide the acquisition date and the orbital elevation at which each SAR image was acquired. To determine which dates could be used to produce interferograms, a network perpendicular-baseline plot was produced for each track (Figure 4). The perpendicular-baseline distance is the separation perpendicular to the look direction of the orbital trajectories between two orbit pass dates. By plotting the date of acquisition against the relative baseline distance, possible interferograms with short perpendicular distances, but long time spans, can be identified. If the perpendicular distance is too large, increased orbital errors are introduced, which leads to a loss of coherence across the interferogram. The closer the satellite orbits to a previous position, the more coherent an interferogram is likely to be.

Coherency is a measure of how the phase contributions across one pixel sum: sufficient changes in the scattering properties along the signal path may lead to a destructive summation of the returning signal and the loss of information. If this persists for many pixels the interferogram will be incoherent, having no relationship between adjacent pixels, and cannot be used.

The temporal span of each interferogram must be also considered. The longer the time over which an interferogram is produced the greater the recorded signal associated with the deformation is. However, over long periods the surface may change too much, leading to incoherency.

The network perpendicular baseline plot for track 493 (Figure 4) highlighted 129 possible interferograms with perpendicular baselines <200m, however, due to incoherency only 94 of these were possible.

Figure 4: Temporal and perpendicular baseline plot for track 493 (94 interferograms)

Figure 4: Temporal and perpendicular baseline plot for track 493 (94 interferograms).
Green lines connect dates (table A1) between which interferograms (table A2) with perpendicular baseline less than 200m were successfully created.

ROI_PAC Processing

SAR images that cover the study area were processed using open-source ROI_PAC (Repeat Orbit Interferometry Package) software (Rosen et al., 2004) provided by JLP/Caltech. The topographic contribution to the phase was corrected using a 3-arc-second STRM DEM (Farr et al., 2007).

This software has the capability to produce raw interferograms from the SAR images, and to unwrap them. The process of unwrapping takes the phase of an unwrapped interferogram (between -π and π) and translates it into the line of sight displacement. The output unwrapped interferograms are then individually manually inspected. Their suitability is evaluated by visual inspection and interferograms with low coherency are reprocessed manually.

The reprocessing involved resampling the raw interferograms to a larger pixel sizes (Massonnet and Feigl, 1998), on the assumption that by finding the phase over a larger area it, on average, will be more coherent, and by manually searching for and correcting unwrapping errors. Unwrapping error corrections were then checked using loop closure (Biggs et al., 2007: 1169-79).


PiRATE (Poly-interferogram Rate And Time series Estimator) (Wang et al., 2009) is a software package that produces a map of LOS velocity from multiple temporally overlapping interferograms. Using a multi-interferogram approach, as opposed to simply stacking the interferograms, allowed empirical corrections made throughout the processing to be solved more reliably on an epoch-by-epoch basis. These corrections include removing the orbital errors and topographically correlated atmospheric errors (Elliot et al., 2008). Secondly, the approach removes the requirement that, for a pixel to be present in the final LOS velocity map, it must be present in all interferograms. This is because the multi-interferogram approach calculates LOS velocity pixel-by-pixel, which can be done from different combinations of interferograms. The processing steps performed by PiRATE are summarised in Figure 5.

Figure 5: PiRATE processing steps

Figure 5: PiRATE processing steps.
Flow map describing the processing steps performed by PiRATE to take a series of interferograms and produce a map of LOS velocity. TVCM stands for temporal variance-covariance matrix. See Walters (2012) and Biggs et al. (2007: 1165-79) for full description of TCVM, minimum spanning tree network and the ratemap estimation.

Initial Deformation Model

Orbital errors can appear similar to long wavelength interseismic deformation signals, and in the InSAR data alone the contribution from either is not known separately. An attempt to remove the orbital errors by an empirical correction may therefore remove some of the signal. To resolve this, a-priori information on the velocity field (from the GPS data) was used to construct a deformation model which was initially removed from the data (Wang and Wright, 2012). The same methodology as Walters (2012) was used to calculate the deformation model in this article, and it should be noted that this approach does not lead to a loss of independence of the InSAR data (Wei et al., 2010: 236-49). A comparison between the use of an initial model and not is seen in Figure 6.

Figure 6: Comparison of different PiRATE processing parameters, part 1

Figure 6: Comparison of different PiRATE processing parameters, part 2

Figure 6: Comparison of different PiRATE processing parameters.
TL: initial model and atmospheric effects considered, quadratic orbital correction.
TM: Initial model considered, quadratic orbital corrections, no atmospheric corrections.
TR: Initial model and atmospheric corrections, linear orbital corrections.
BL: No initial model considered, quadratic orbital corrections, atmospheric corrections applied.
BM: No initial model, linear orbital corrections, atmospheric corrections applied.
BR: No initial model or atmospheric corrections considered, linear orbital corrections.
In all cases the tectonic signal across the NAF and EAF is present, indicating it is not an artefact of the processing.

Orbital Corrections

I used the epoch-by-epoch approach in this article to remove orbital errors since it considers the covariance of dependant interferograms, which has been shown to have a ~9% improvement of estimated orbital errors (Biggs et al., 2007).

In addition, there is a choice between approximating the orbital errors as a plane or a quadratic. I tested both approximations on the data in this article, and chose to use quadratic orbital corrections since they are more appropriate for long InSAR tracks such as these (Biggs et al., 2007). A comparison between them is shown in Figure 6.

Atmospheric corrections

A delay to the radar signal is caused by passing through the atmosphere. The European Centre for Medium-range Weather Forecasting (ECMWF) produces a global atmospheric model (ERA-I) from satellite and surface data. This model includes estimates of the temperature and relative humidity four times a day at 75km horizontal spacing (Dee et al., 2011: 553-597). Hydrostatic and wet delay maps were produced from this model via the method shown in Jolivet et al. (2011) to correct for atmospheric influences. Figure 6 shows a comparison between considering the atmospheric influence derived in this way, and without.

Following an empirical evaluation of each of the possible parameter configurations, the processing and modelling hereafter is performed on the ratemap produced using an initial deformation model, the removal of quadratic orbital errors and with atmospheric effects considered.


I used the methodology of Wang and Wright (2012) to derive a strain rate map using the velocity field inverted from InSAR and GPS data using the VELMAP software package (Wang and Wright., 2012). For this I divided Eastern Anatolia, and the respective geodetic data, into an arbitrary triangular mesh (England and Molnar, 2005) (Figure 7). The velocity data within each triangle was related to the triangle nodes by an interpolation function which were then interpolated across the region to create a continuous velocity field. Wang and Wright (2012) justify the use of a Laplacian smoothing operator to remove short-wavelength noise from the interferograms, which I chose to reproduce. The chosen degree of smoothing was based on the trade-off curve produced by Walters (2012), and the assumption that including my data would not relative change the trade-off curve. Using the same smoothing as Walters (2012) also allowed a comparison between the velocity model of Walters (2012) and this article.

Figure 7: Inputs for the velocity field inversion

Figure 7: Inputs for the velocity field inversion.
Extent and style of the triangular mesh used in the velocity field inversion, as well at the input data. Tracks 78, 307, 35, 171 and 400 are those processed by Walters (2012).

I inverted the geodetic data for East and North velocity components only. This is justified given that negligible vertical motion was been previously found (Walters, 2012) using five of the six InSAR tracks and the same GPS data as this article, and that it is also accepted in the literature (Cavalié and Jónsson, 2014; Reilinger et al., 2006). The velocity values at each node were then related to the strain rate via the equations given by Savage et al. (2001: 995-1002). For a more detailed mathematic description of the VELMAP procedure see Walters (2012).

I followed this procedure twice: once using the same data as Walters (2012) and then again with the inclusion of the InSAR data processed in this article to test whether, and in what way, the extra InSAR data improved the certainty in the velocity model.

Block modelling methodology

To test whether the East Anatolian velocity field can be described by the motions of a rigid block I used the modelling technique introduced by McCaffrey (1995; 2002) using the DEFNODE software. I performed this block modelling using the data provided by Walters (2012) as well as that processed here to see whether more data and reduced uncertainty affects the fit of the block.

The blocks used in this modelling are those proposed by Reilinger et al. (2006). They are defined as rigid regions bounded by zones of elastic strain accumulation associated with the faults. The best fitting Euler pole is then inverted for, and used to calculate the velocity field predicted by the rotation of the block.


Velocity field for East Anatolia

Figure 8 shows the velocity field modelled from the joint solution of all six InSAR tracks and GPS data, and shows the expected overall westwards motion. The modelled field also matches the GPS and InSAR data well (Figure 9). The GPS residual locations are consistent with those found by Walters (2012).

Figure 8: Best estimate of the velocity field

Figure 8: Best estimate of the velocity field.
Left: Best solution velocity field produced from inversion of GPS and six tracks of InSAR data.
Right: Velocity field with uncertainty ellipses (95% confidence bounds) and GPS data vectors for comparison of prediction.

Figure 9: Prediction of input GPS data by the velocity field model and input InSAR LOS rate by modelled velocity field, part 1

Figure 9: Prediction of input GPS data by the velocity field model and input InSAR LOS rate by modelled velocity field, part 2

Figure 9: Prediction of input GPS data by the velocity field model and input InSAR LOS rate by modelled velocity field.
Top: Prediction of input GPS data by the velocity field model.
- Left: Velocity field model (red) presented in this article at the locations of GPS data (black).
- Right: Residuals (blue) between the velocity field and GPS data. Error ellipses are given to 95% confidence.
Bottom: Prediction of input InSAR LOS rate by modelled velocity field.
- Left: Input InSAR data.
- Middle: LOS velocity predicted by the velocity model.
- Right: Residual between input InSAR data and modelled velocity field. Ratemaps are resampled to ~10km resolution.

A comparison between the uncertainty in this velocity field model and that from Walters (2012) is presented in Figure 10. The inclusion of the InSAR data processed in this article reduces the east-west uncertainties in the KTJ region. Almost all of the Anatolian plate in this model has east-west velocity field uncertainties of less than 1mm/yr. When considering the whole region, the extra InSAR data has caused a 36% increase in the number of velocity field points with east-west uncertainties less than 1mm/yr, and a 65% decrease between 1 - 4mm/yr. The inclusion of this InSAR data however provided no appreciable reduction in the north-south uncertainties, although this was expected.

Figure 10: Uncertainty in velocity field model, part 1

Figure 10: Uncertainty in velocity field model, part 2

Figure 10: Uncertainty in velocity field model, part 3

Figure 10: Uncertainty in velocity field model.
Top: Error in the velocity field produced in this article in E-W (left) and N-S (right) directions. Histogram shows the frequency of locations with given uncertainties.
Middle: Same as top, using data provided by Walters (2012).
Bottom: The change (decrease) in uncertainty of the velocity field given the inclusion of track 493. Negligible difference is made to the N-S velocities, and so are not included in the histogram. Dashed lines show extent of InSAR coverage, white circles show GPS station locations.

Distribution of strain accumulation across East Anatolia

The velocity field (Figure 8) was used to calculate the strain rate map presented here. It is given that the magnitude and distribution of the uncertainty in the velocity field will reflect that in the strain rate.

The strain accumulation in the Eastern Anatolian region is localised close to the NAF and north-east section of the EAF. The inclusion of the extra InSAR data has shown strain accumulation is confined closer to the faults outside of the KTJ, compared to Walters (2012), and highlighted a consistent region of higher strain within the KTJ which was not so pronounced in Walters (2012) (Figure 11). The true values of strain however cannot be obtained given the level of subjectivity in selecting, for example the smoothing parameters, although the distribution and relative changes in strain rate across the region seen here are valid.

Figure 11: Comparison of strain rate to published models, part 1

Figure 11: Comparison of strain rate to published models, part 2

Figure 11: Comparison of strain rate to published models.
Top left: Strain map calculated from the velocity field produced using six InSAR tracks and 162 separate GPS station locations.
Top right: Strain rate field produced using the data from Walters (2012).
Bottom: Magnitude of the difference in the strain field between this article and Walters (2012).

Block model of Eastern Anatolia

The extent to which the velocity field of Eastern Anatolia can be described as the rotations of a rigid block is shown in Figure 12. The block model predicts velocities across central Anatolia well, but increasingly worse towards the KTJ. Most notably block modelling fails to predict the orientation of the velocity field in the KTJ, estimating 10mm/yr of convergence across the NAF that is not present in the data and extends almost 100km from the KTJ. The block model predicts ~20mm/yr of slip along the NAF, and ~8mm/yr along the EAF.

Figure 12: Block model predictions of the velocity field

Figure 12: Block model predictions of the velocity field.
Left: Thick black lines represent block boundaries used in the inversion. Red and black arrows represent the block modelled and interpolated velocity fields respectively. Error ellipses are given at 95% confidence.
Middle: Residuals between block and interpolated velocity field. Colour gradient across the NAF and EAF represent slip predicted by block model.
Right: Colour gradient along faults give degree of convergence as predicted by the block model (i.e. blue boundaries represent convergence, red show divergence).


The ratemap and velocity field show velocity contrasts across the NAF and EAF that support the hypothesis that the East Anatolian region acts as a rigid body. This is also supported by the observed distribution of strain accumulation, and thus likely highest hazard, which is mostly focussed within ~30km of the faults.

The addition of my InSAR data to that of Walters (2012) has reduced the east-west uncertainties in the velocity field across the whole Eastern Anatolian plate to less than 1mm/yr. Despite this improvement, I find no geodetic evidence to support the dynamically separate blocks proposed by Chorowicz et al. (1999). Should such blocks exist, this article therefore provides an upper constraint (~1mm/yr) on the magnitude of the deformation they exhibit. In addition, this article did not find any strain accumulation along secondary faults between the NAF and EAF as suggested by Ozener et al. (2010: 1184). However, I did find elevated strain accumulation up to ~85km west of the KTJ.

In a previous article (Walters, 2012) a block model was used to test whether the motions of rigid blocks could describe the modelled East Anatolian velocity field. The article found residuals in the TJ between the velocity predicted by a rotating rigid block and the velocity field that were greater than the associated uncertainty. This suggested that eastern Anatolia cannot be described by a rigid block alone. After including InSAR data from across the TJ the block model continues to overestimate the velocity across the NAF (~10mm/yr) and suggests ~10mm/yr of convergence in the KTJ. However, there is no evidence of this convergence at the faults in the data, therefore this motion must be accommodated via elastic deformation inside the KTJ instead.

Convergence greater than 5mm/yr is predicted extending 100km from the KTJ along the NAF, after which its magnitude decays. The extent of this predicted convergence corresponds to that of elevated strain accumulation, supporting the hypothesis that convergence predicted by block modelling but not seen in the data is accommodated by deformation. This agrees with the results of Walters (2012) in that the KTJ region does not act as a rigid block. This was expected, given that that the KTJ must undergo some deformation to remove the requirement of 'gaps' occurring.

However, despite the evidence that the KTJ is undergoing some non-block-like deformation I do not believe that this contradicts the description of Anatolia as a rigid block. Overall, East Anatolia does behave block-like: little deformation is observed, except that which is confined to the major faults. The definition of block-like allows deformation at the faults, and so deformation where faults meet, as observed in this article, should also be expected. I suggest that since this TJ deformation occurs over a small region compared to the size of the plate, the description of Anatolia as a rigid block is valid.

Limitations and suggestions for further research

The InSAR data used in the velocity inversion primarily covers the Anatolian plate, so the nature of the deformation east of the TJ is poorly constrained. It is possible that assuming it is rigid is an oversimplification and that the diffuse deformation required at the TJ is in part accommodated elsewhere. This article could be expanded to include InSAR data which covers the east of the KJT to investigate strain rates there. The inclusion of track 357 would also reduce the uncertainty, and allow an estimation of the vertical deformation which is independent of Walters (2012).

This article also raises further questions on deformation around triple junctions, such as whether the spatial extent of ductile deformation, for example, for a given rheology could be a function of the angle at which the faults meet and the slip along the faults. Research into other TJ regions with different slip and intersection angles would be required to test this. How the transition between different styles of deformation is reconciled along the faults at the surface and at depth would also be an interesting topic, given its impact on the possible rupture area, and hence earthquake size.


Strain accumulation maps allow investigation into continental deformation, and highlight areas of likely seismic risk. In agreement with the literature, I found that strain in Eastern Anatolia is accumulating primarily within ~30km of the major fault structures.

The geodetic evidence suggests that Eastern Anatolian acts, overall, as a rotating rigid body, with some ductile deformation towards the KJT where systematic misfits (~10mm/yr) with block model predictions occur. The behaviour here is interpreted as a necessity of rigid block deformation with such a geometry. This result highlights the oversimplification of defining continental regions as deforming as either blocks or a viscous fluid, and that it is likely all regions deform somewhere on the continuum between.


I would like to thank Dr Richard Walters for his continued patience and guidance, and my supervisor, Professor Tim Wright, for his help and advice. I really appreciate the time you have both put in for me. I would also like to thank ESA for the raw InSAR data, Walters (2012) for his processed data, ECMWF for the atmospheric models and Dr Walters for the framework of some of the GMT figures presented here.

All figures were prepared using GMT (Wessell and Smith, 1998) unless otherwise stated.

List of Figures

Figure 1: Tectonic setting of the Anatolian region. Author' s own figure.

Figure 2: Geographic distribution of the data used in this article. Author's own figure.

Figure 3: Schematic diagram of satellite-ground geometry used for SAR imagery. Author's own figure.

Figure 4: Temporal and perpendicular baseline plot for track 493 (94 interferograms). Author's own figure.

Figure 5: PiRATE processing steps. Author's own figure.

Figure 6: Comparison of different PiRATE processing parameters. Author's own figure.

Figure 7: Inputs for the velocity field inversion. Authors' own figure.

Figure 8: Best estimate of the velocity field. Author's own figure.

Figure 9: Prediction of input GPS data by the velocity field model and input InSAR LOS rate by modelled velocity field. Author's own figure.

Figure 10: Uncertainty in velocity field model. Author's own figure.

Figure 11: Comparison of strain rate to published models. Author's own figure.

Figure 12: Block model predictions of the velocity field. Author's own figure.

List of Tables

Table 1: Contributions to the phase delay within an InSAR image.

Table A1: SAR image Acquisition dates and Perpendicular baseline for track 493.

Table A2: Track 493 interferograms.


In this appendix I present the acquisition dates (Table A1) and the interferograms (Table A2) formed using them for track 493.

ID number Acquisition date Perpendicular baseline (m)
1 2003-08-11 0
2 2003-09-15 -1097
3 2003-11-24 -269
4 2003-12-29 -1450
5 2003-12-29 -1451
6 2004-02-02 -1480
7 2004-03-08 -910
8 2004-04-12 -1643
9 2004-05-17 -596
10 2004-06-21 -627
11 2004-08-30 -971
12 2004-10-04 -668
13 2004-11-08 -428
14 2004-12-13 -594
15 2005-01-17 -433
16 2005-03-28 45
17 2005-03-28 45
18 2005-05-02 -1357
19 2005-06-06 -377
20 2005-09-19 -700
21 2005-10-24 -1121
22 2005-11-28 -570
23 2006-03-13 -824
24 2006-03-13 -807
25 2006-07-31 -1652
26 2006-10-09 -199
27 2007-02-26 -784
28 2007-05-07 -744
29 2007-09-24 -1179
30 2007-12-03 -1104
31 2008-03-17 -650
32 2008-04-21 -1177
33 2008-05-26 -689
34 2009-08-24 -893
35 2009-09-28 -1355
36 2010-01-11 -582
37 2010-03-22 -952
38 2010-03-22 -951
39 2010-04-26 -1133
40 2010-04-26 -1130

Table A1: SAR image Acquisition dates and Perpendicular baseline for track 493. Perpendicular baseline relative to first date, which is arbitrarily chosen. ID numbers correspond to those shown in Figure 4.

Index Interferogram start and end date (YY/MM/DD)
1 030915-040308
2 030915-040830
3 030915-051024
4 030915-070924
5 030915-071203
6 030915-080421
7 030915-090824
8 030915-100426
9 040202-060731
10 040202-090928
11 040308-040830
12 040308-060313
13 040308-070226
14 040308-070507
15 040308-071203
16 040308-090824
17 040412-060731
18 040517-040621
19 040517-041004
20 040517-041213
21 040517-050117
22 040517-050919
23 040517-051128
24 040517-070226
25 040517-070507
26 040517-080317
27 040517-080526
28 040517-100111
29 040621-041004
30 040621-041213
31 040621-050919
32 040621-051128
33 040621-070226
34 040621-070507
35 040621-080317
36 040621-080526
37 040621-100111
38 040830-051024
39 040830-060313
40 040830-070226
41 040830-071203
42 040830-090824
43 040830-100426
44 041004-041213
45 041004-050919
46 041004-051128
47 041004-070226
48 041004-070507
49 041004-080317
50 041004-080526
51 041004-100111
52 041108-050117
53 041108-050606
54 041108-051128
55 041213-051128
56 041213-070226
57 041213-070507
58 041213-080317
59 041213-080526
60 050117-050606
61 050117-100111
62 050502-070924
63 050502-080421
64 050502-090928
65 050606-051128
66 050606-061009
67 050606-100111
68 050919-051128
69 050919-060313
70 050919-070226
71 050919-070507
72 050919-080317
73 050919-080526
74 050919-100111
75 051024-070924
76 051024-071203
77 051024-080421
78 051128-070507
79 051128-080317
80 051128-080526
81 051128-100111
82 060313-070226
83 060313-070507
84 060313-080317
85 060313-090824
86 070226-070507
87 070226-080317
88 070226-080526
89 070226-090824
90 070507-080317
91 070507-080526
92 070924-071203
93 070924-080421
94 070924-090928
95 071203-080421
96 071203-100426
97 080317-080526
98 080421-100426

Table A2: Track 493 interferograms. Dates used in name of interferogram are those of the first and second SAR acquisition dates respectively.


[1] Ryan Lloyd completed a BSc Geophysics Sciences at the University of Leeds in 2014. He presented this work at the 2014 British Conference of Undergraduate Research (BCUR) hosted by the University of Nottingham, where he won the Best Spoken Presentation award for the first day. He is now working towards a PhD at the University of Bristol, using InSAR to investigate volcanic hazards and the processes associated with continental rifting at the East African Rift.


Aktug, B., A. Nocquet, A. Cingoz, B. Parsons, Y. Erkan, P. England, O. Lenk, M. A. Gurdal, A. KIlicoglu, H. Akdeniz and A. Tekgul (2009), 'Deformation of western Turkey from a combination of permanent and campaign GPS data: Limits to block like behaviour', Journal of Geophysical Research,114, B10404

Aktug, B., E. Parmaksiz, M. Kurt, O. Lenk, A. Kilicoglu, M. Gurdal and S. Ozdemir (2012), 'Deformation of Central Anatolia: GPS Implications', Journal of Geodynamics, 67, 78–96

Aktug, B., U. Dikmen, A. Dogru and H. Ozener (2013), 'Seismicity and strain accumulation around Karloiva Triple Junction (Turkey)', Journal of Geodynamics, 67, 21–29

Alchalbi, A., M. Daoud, F. Gomez, S. McClusky, R. Reilinger, M. A. Romeyeh, A. Alsouod, R. Yassminh, B. Ballani, R. Darawcheh, R. Sbeinati, Y. Radwan, R.A. Masri, M. Bayerly, R. A. Ghazzi and M. Barazangi (2010), 'Crustal deformation in northwestern Arabia from GPS measurements in Syria: Slow slip rate along the northern Dead Sea Fault', Geophysical Journal International, 180, 125–35

Avouac, J. and P. Tapponnier (1993),'Kinematic model of active deformation in central Asia', Geophysical Research Letters, 20, 895–98

Biggs, J., T. Wright, Z. Lu and B. Parsons (2007), 'Multi-interferogram method for measuring interseismic deformation: Denali Fault, Alaska', Geophysical Journal International, 170, 1165–79

Bürgmann, R (2000), 'Synthetic Aperture Radar Interferometry to Measure Earth's Surface Topography and Its Deformation', Annual Review of Earth and Planetary Science, 28, 169

Cavalié, O. and S. Jónsson (2014), 'Block-like plate movements in eastern Anatolia observed by InSAR', Geophysical Research Letters, 41, 26–3 1

Chorowicz, J., D. Dhont and N. Gundogdu (1999), 'Neotectonics in the eastern North Anatolian fault region (Turkey) advocates crustal extension: mapping from SAR ERS imagery and Digital Elevation Model', Journal of Structural Geology, 21, 511–32

Dee, D. P., S. M. Uppala, A. J. Simmons, P. Berrisford, P. Poli, S. Kobayashi, U. Andrae, M.A. Balmaseda, G. Balsamo, P. Bauer, P. Bechtold, A. C. M. Beljaars, L. van de Berg, J. Bidlot, N. Bormann, C. Delsol, R. Dragani, M. Fuentes, A. J. Geer, L. Haimberger, S. B. Healy, H. Hersbach, E. V. Hólm, L. Isaksen, P. zållberg, M. Köhler, M. Matricardi, A. P. McNally, B. M. Monge-Sanz, J. J. Morcrette, B. K. Park, C. Peubey, P. de Rosnay, C. Tavolato, J. N. Thépaut and F. Vitart (2011) 'The ERA-Interim reanalysis: configuration and performance of the data assimilation system', Quarterly Journal of the Royal Meteorological Society, 137, 553–9 7

Elliott, J. R., J. Biggs, B. Parsonsand T. J. Wright(2008), 'InSAR slip rate determination on the Altyn Tagh Fault, northern Tibet, in the presence of topographically correlated atmospheric delay', Geophysical Research Letters, 35, L12309

England, P. and J. Jackson (2011), 'Uncharted seismic risk', Nature Geoscience, 4, 348– 49

England, P. and D. McKenzie (1982), 'A thin viscous sheet model for continental deformation', Geophysical Journal International, 70, 295–321

England, P. and P. Molnar (2005), 'Late Quaternary to decadal velocity field in Asia'. Journal of Geophysical Research: Solid Earth, 110, B12401

Farr, T. G., P. A. Rosen, E. Caro, R. Crippen, R. Duren, S. Hensley, M. Kobrick, M. Paller, E. Rodriguez, L. Roth, D. Seal, D. Shaffer, J. Shimada, J. Umland, M. Werner, M. Oskin, D. Burbank and D. Alsdorf (2007), 'The Shuttle Radar Topography Mission', Reviews of Geophysics, 45, RG2004

Floyd, M. A., H. Billiris, D. Paradissis, G. Veis, A. Avallone, P. Briole, S. McClusky, J. Nocquet, K. Palamartchouk, B. Parsons and P. England (2010),'A new velocity field for Greece: Implications for the kinematics and dynamics of the Aegean', Journal Geophysical Research,115, B10403

Holzer, T. L. and J. C. Savage (2013), 'Global Earthquake Fatalities and Population'. Earthquake Spectra, 29, 155–175

Jolivet, R., R. Grandin, C. Lasserre, M. Doin and G. Peltzer (2011), 'Systematic InSAR tropospheric phase delay corrections from global meteorological reanalysis data', Geophysical Research Letters, 38, L17311

Liu, M., S. Stein and H. Wang (2011), '2000 years of migrating earthquakes in North China: How earthquakes in midcontinents differ from those at plate boundaries', Lithosphere,3, 128–32

Massonnet, D. and K. L. Feigl (1998),'Radar interferometry and its application to changes in the Earth's surface', Reviews of Geophysics, 36, 441–500

McCaffrey, R. (1995), 'DEFNODE users' guide', Rensselaer Polytechnic Institute, Troy, 2004

McCaffrey, R. (2002), 'Crustal block rotations and plate coupling', Geodynamic Series, 30, 101–22

McClusky, S., S. Balassanian, A. Barka, C. Demir, S. Ergintav, I. Georgiev, O. Gurkan, M. Hamburger, K. Hurst, H. Kahle, K. Kastens, G. Kekelidze, R. King, V. Kotzev, O. Lenk, S. Mahmoud, A. Mishin, M. Madariya, A. Ouzounis, D. Paradissis, Y. Peter, M. Prilepin, R. Reilinger, I. Sanli, H. Seeger, A. Tealeb, M. N. Toksoz and G. Veis (2000), 'Global Positioning System constraints on plate kinematics and dynamics in the eastern Mediterranean and Caucasus', Journal of Geophysical Research, 105, 5695–720

Ozener, M. S. and W. E. Holt (2010), 'The dynamics of the eastern Mediterranean and eastern Turkey', Geophysical Journal International, 183, 1165–84

Reid, H. (1910), 'The Mechanics of the earthquake', Carnegie institution of Washington, 2

Reilinger, R., S. McClusky, P. Vernant, S. Lawrence, S. Ergintav, R. Cakmak, H. Ozener, F. Kadirov, I. Guliev, R. Stepanyan, M. Nadariya, G. Hahubia, S. Mahmoud, K. Sakr, A. ArRajehi, D. Paradissis, A. Al-Aydrus, M. Prilepin, T. Guseva, E. Evren, A. Dmitrotsa, S. V. Filikov, F. Gomez, R. AlGhazzi and G. Karam (2006), 'GPS constraints on continental deformation in the Africa-Arabia-Eurasia continental collision zone and implications for the dynamics of plate interactions', Journal of Geophysical Research, 111, L05411

Rosen, P., A. Hensley, G. Peltzer and M. Simons (2004), 'Update Repeat Orbit Interferometry Package Released', EOS Transactions AGU, 85, 47

Şaroğlu, F., Ö. Emre and I. Kuşçu (1992), 'The East Anatolian Fault of Turkey', Annalles Tectonicae, 206, 99–125

Savage, J. C., W. Gan and J. L. Svarc (2001), 'Strain accumulation and rotation in the Eastern California Shear Zone', Journal of Geophysical Research,106, 995– 1002

Tatar, O., H. Poyraz, H, Gursoy, Z. Cakir, S. Ergintav, Z. Akpinar, F. Kocbulut, F. Sezen, T. Turk, O. Hastaoglu, A. Polat, B. L. Mesci, O. Gursoy, E. Ayazli, R. Cakmak, A. Belgen and H. Yavasoglu (2012), 'Crustal deformation and kinematics of the Eastern Part of the North Anatolian Fault Zone (Turkey) from GPS measurements', Tectonophysics, 518, 55–62

Thatcher, W. (2009), 'How the continents deform: the evidence from tectonic geodesy', Annual reviews of Earth and Planetary Sciences, 37, 237–6 2

Walters, R., R. Holley, B. Parsons and T. Wright (2011), 'Interseismic strain accumulations across the North Anatolian Fault from Envisat InSAR measurements', Geophysical Research Letters, 38, L05303

Walters, R. J. (2012), 'Geodetic observation and modelling of continental deformation in Iran and Turkey', D.Phil. thesis, University of Oxford, Department of Earth Sciences, Oxford, UK

Wang, H. and T. J. Wright (2012),'Satellite geodetic imaging reveals internal deformation of western Tibet', Geophysical Research Letters,39, L07303

Wang, H., T. J. Wright and J. Biggs (2009), 'Interseismic slip rate of the northwestern Xianshuihe fault from InSAR data', Geophysical Research Letters, 36, L03302

Wei, M., D. Sandwell and B. Smith-Konter (2010), 'Optimal combination of InSAR and GPS for measuring interseismic crustal deformation', Advances in Space Research, 42,236–4 9

Wessel, P. and W. H. F. Smith (1998), 'New, improved version of Generic Mapping Tools released', EOS Transactions of American Geophysical Union, 79, 579

Wright, T., E. Fielding and B. Parsons (2001), 'Triggered slip: Observations of the 17 August 1999 Izmit (Turkey) Earthquake using radar interferometry', Geophysical Research Letters, 28, 1079–82

Wright, T. (2002), 'Remote monitoring of the earthquake cycle using satellite radar interferometry', Philosophical Transactions of the Royal Society of London, Series A: Mathematical, Physical and Engineering Sciences, 360, 2873–88

Wright, T. J., B. E. Parsons and Z. Lu(2004), 'Toward mapping surface deformation in three dimensions using InSAR', Geophysical Research Letters,31, L01607

Yavaşoğlu, H., E. Tarı, O. Tüysüz, Z. Çakır and Z. S. Ergintav (2011), 'Determining and modelling tectonic movements along the central part of the North Anatolian Fault (Turkey) using geodetic measurements', Journal of Geodynamics, 51, 339–43

Zandbergen, R., M. Otten, P. Righetti, D. Kuijper and J. Dow(2003),'Routine operational and high-precision orbit determination of ENVISAT', Advances in Space Research, Integrated Space Geodetic Systems and Satellite Dynamics, 31, 1953–5 8

To cite this paper please use the following details: Lloyd, R. (2014), 'Evaluating the Style of Continental Deformation in Eastern Turkey using InSAR and GPS', Reinvention: an International Journal of Undergraduate Research, BCUR 2014 Special Issue Date accessed [insert date]. If you cite this article or use it in any teaching or other related activities please let us know by e-mailing us at Reinventionjournal at warwick dot ac dot uk.