Kinematics of the Pamir and Hindu Kush regions from GPS geodesy


Corresponding author: R. Bendick, Department of Geosciences, The University of Montana, 32 Campus Dr., Missoula, MT 59812, USA. (


[1] GPS velocities measured in the Pamir and surrounding regions show a total of ~30 mm/yr of northward relative motion between stable Pakistan and Eurasia. The convergence budget is partitioned into 10–15 mm/yr of localized shortening across the Trans-Alai Thrust, which bounds the Pamir on the north, consistent with southward subduction of intact lithosphere. Another 10–15 mm/yr of shortening is distributed across the Chitral Himalaya and Hindu Kush, suggesting that Hindu Kush seismicity might be related to northward subduction of Indian lithosphere. Modest shortening at <5 mm/yr occurs north of the Trans-Alai Thrust, across the South Tien Shan and between the Ferghana Valley and Eurasia. Negligible north-south shortening occurs within the high Pamir, but as much as 5 mm/yr, and perhaps 10 mm/yr, of east-west extension occurs within this region. This extension is matched by a comparable amount of east-west shortening in the Tajik Depression. The localization of shortening to the margins of the Pamir combined with observations of distributed internal extension implies that the east-west vertically averaged, horizontal compressive normal stress is smaller than the north-south compressive stress.

1 Introduction

[2] The Pamir-Hindu Kush region of Central Asia (Figure 1), like the Tibetan Plateau to its east, has accommodated the collision of the Indian subcontinent with Asia by crustal shortening and thickening. Moreover, the same belts of ophiolites and related magmatic zones that reflect suturing of crustal fragments identified within Tibet also extend through the Pamir [e.g., Burtman and Molnar, 1993; Schwab et al., 2004]. These observations combined with comparable mean elevations of ~5000 m for the Tibetan Plateau and ~4000 m for the Pamir have been used to argue for similar evolution and dynamics of the two areas [e.g., Burtman and Molnar, 1993; Ducea et al., 2003; Hacker, 2005; Mechie et al., 2011; Yin and Harrison, 2000].

Figure 1.

(a) Regional velocities relative to Eurasia. Error ellipses show 95% confidence intervals. Vectors are color coded by geographic region to correspond to Figure 2; open arrows are from Zubovich et al. [2010]. Red circles are earthquake epicenters from Engdahl and Villasenor [2002]. Faults discussed in the text are shown in black and labeled as follows: MKT: Main Karakorum Thrust; TM: Tirich Mir Fault; SPT: South Pamir Thrust; DK: Darvaz-Karakul Fault; MPT: Main Pamir Thrust. The transparent grey lines with labels correspond to the velocity transects in Figure 3. (b) As in Figure 1a, for the Pamir and Tajik Depression. Some site names are omitted for clarity.

[3] Differences between the two regions, however, are as important as their similarities. Northward underthrusting is well documented beneath the Himalaya and southern Tibet, but intermediate depth seismicity is sparse and cannot be related to subduction of Indian lithosphere. Intermediate depth seismicity is abundant beneath the Pamir however, and this seismicity, which occurs north and east of the adjacent Hindu Kush region, has been interpreted as occurring either within lithosphere attached to the India plate and continuous with Hindu Kush intermediate depth seismicity [Billington et al., 1977; Pegler and Das, 1998; Pavlis and Das, 2000], or within lithosphere subducted southward beneath the Pamir and separate from the lithosphere containing Hindu Kush seismicity [Burtman and Molnar, 1993; Chatelain et al., 1980; Hamburger et al., 1992; Koulakov and Sobolev, 2006; Kumar et al., 2005; Mechie et al., 2011; Negredo et al., 2007; Roecker et al., 1980]. Furthermore, although subduction of Indian lithosphere may include oceanic lithosphere that formed beneath the Tethyan Ocean and north of the Indian subcontinent, the origin of the seismically active materials at depth in the Pamir and Hindu Kush, as well as their mechanical and chemical properties, is disputed [Khalturin et al., 1977; Mechie et al., 2011; Mellors et al., 1995; Roecker, 1982]. In fact, if the Pamir intermediate depth seismicity occurs in lithosphere subducted from the north, that lithosphere could have been continental in origin, with subduction initiating in thinned continental crust [Burtman and Molnar, 1993; Chatelain et al., 1980].

[4] With these similarities and differences between the Pamir and Tibet in mind, we use GPS geodesy from the Pamir and Hindu Kush regions to calculate a surface velocity field. In the following descriptions and discussion, we consider the Hindu Kush range to extend eastward from northern Afghanistan near Faizabad through northernmost Pakistan and the Wakhan Corridor to a junction with the Karakorum Range northwest of K2 (Figure 1). These surface velocities, plus the distribution of seismic moment release from a global catalog [Engdahl and Villaseñor, 2002], topographic and gravitational potential energy gradients, and maps and estimated rates of active faulting from surface geology [e.g., Arrowsmith and Strecker, 1999; Chevalier et al., 2011; Cowgill, 2010; Kuchai and Trifonov, 1977; Nikonov, 1974, 1977; Nikonov et al., 1983; Robinson et al., 2004, 2007; Ruzhentsev, 1963; Strecker et al., 1995; Trifonov, 1978] are then used to place bounds on the present-day kinematics of crustal deformation in Central Asia.

2 Methods

[5] We installed and collected data at 39 new geodetic sites in Tajikistan from 2005 to 2011 (Table 1). Four of these have recorded continuously since 2007, and the rest were measured in campaign-style occupations of permanent monuments. Campaign sites used in this work have a minimum of two 2 day occupations, and most sites have three 2 day observation epochs. Data were collected by Trimble netRS instruments at continuous sites and Trimble R7 survey instruments at campaign sites.

Table 1. Geodetic Velocities for Sites Used in This Analysis, in the ITRF08 Reference Frame and a Eurasia-Fixed Framea
SiteLongitudeLatitudeITRF08 EastITRF08 NorthEurasia-Fixed EastEurasia-Fixed NorthEast SigmaNorth SigmaRho
  1. aUncertainties are 1-σ.

[6] These raw data were combined with observations from nine other sites in India, Pakistan, and Afghanistan plus 19 sites in Kyrgyzstan with data available from the UNAVCO archive from historical geodetic surveys [Abdrakhmatov et al., 1996], and 16 IGS (International GNSS Service) sites for loosely constrained daily position estimates calculated in GAMIT. These loosely constrained solutions were then combined with daily global solutions from the Massachusetts Institute of Technology processing center to generate time series using GLRED and GLOBK. We edited time series for each station to remove outliers and to calculate appropriate site weights to estimate a consistent velocity solution. Adjustments from the ITRF08 were minimized to define the reference frame. Finally, we averaged position estimates over 15 day intervals and used these averages to estimate velocities in the ITRF08 frame using GLOBK. The time series of position averages are provided in the supporting information. In order to account for temporal correlations in the displacement data, which usually dominate the velocity uncertainties, we added to each station a random-walk component determined by fitting the daily time series to a Gauss-Markov process, as described by Herring [2003] and Reilinger et al. [2006]. For the campaign stations, whose observations are too few to allow a reliable estimate of the random walk, we used the median noise value from the continuously recording stations. We also calculated transformations from the ITRF08 reference frame to a Eurasia-fixed frame by choosing sites located on the stable Eurasian interior and by minimizing their velocities (Table 1).

[7] We supplemented our regional solution with 21 additional velocities reported by Zubovich et al. [2010], primarily to add additional information about the velocity of the Tarim Basin relative both to stable Eurasia and to the study sites. In order to include these sites, we calculated a six-parameter (three rotations and three translations) transformation from the Eurasia-fixed frame used in Zubovich et al. [2010] to our Eurasia-fixed frame using 57 common stations, mainly global IGS sites.

[8] To estimate relative velocities between geographic regions, we differenced the weighted mean velocities for several sites in a region, rather than simply differencing the velocity vectors of selected representative individual sites. Differences in velocities among nearby sites, which are comparable to uncertainties for each, make more detailed analysis premature. For these weighted estimates, the reported uncertainties are the square roots of the sums of the variances of the weighted mean velocities. For example, to estimate the east-west extension within the Pamir, we calculated the difference between the weighted average eastward component of velocity of all sites in the western Pamir and that of all sites in the eastern Pamir. Similarly, the estimate of relative velocity between the eastern Pamir and Tarim Basin is the difference between the average velocity of all sites in the eastern Pamir and all sites in the Tarim. The estimate of relative velocity across the western Pamir margin is complicated by internal deformation within the Tajik Depression. So, for that region, we use only the three extreme easternmost Tajik Depression sites and compare their average westward components of velocity with that of the four westernmost sites in the Pamir. Geographic regions were defined using elevation, surface geology, and velocity, so that defined regions reflect previous regional geological interpretations and do not contain significant (greater than 2σ) internal relative velocities. Figure 2 shows the northward and eastward velocities for all reported sites and the regional averages color coded by these geographic regions.

Figure 2.

(a) Northward and (b) eastward components of velocity versus longitude of sites. Symbols are color coded by geographic region, with the weighted mean velocities used for relative velocity estimates shown as bold symbols of the same colors. Error bars show 1σ.

[9] We prefer weighted averaging of site velocities in this way to a block-fitting algorithm because we are interested both in regions whose internal strain rates are very low (such as within the High Pamir) and in regions with systematic patterns of deformation (such as within the Tajik Depression). We also consider the main structures in the region and their kinematics to be too poorly known to allow block-like domains to be defined sensibly.

3 Results

[10] Maps of surface velocities relative to Eurasia (Figure 1) with surface velocities classified by region (Figure 2) and components plotted along three profiles (Figure 3) reveal several aspects of the regional kinematics. First, four regions have northward components of velocity greater than 10 mm/yr relative to stable Eurasia: northern Pakistan, including both the Chitral Himalaya and the Hindu Kush; the Pamir; the Tarim Basin; and the westernmost Tibetan Plateau in Ladakh (Figure 2a). Sites in the Tajik Depression, the South Tien Shan, and near the western end of the Tien Shan in Kyrgyzstan move northward more slowly with respect to Eurasia, from 3 mm/yr at the western end of the Tien Shan to as much as 6 mm/yr in the South Tien Shan and northernmost Tajik Depression. This faster movement of the South Tien Shan than regions farther north reflects both a counter-clockwise rotation of the Ferghana Valley about an axis near its southwest end [Reigber et al., 2001; Zubovich et al., 2010] and convergence between the South Tien Shan and the Ferghana Valley. Northward components of velocity in the Pamir also vary systematically with longitude, faster in the east and center than in the west (Figure 2a).

Figure 3.

Profiles of northward components of velocity with 1σ error bars for (a) the Chitral Himalaya, Hindu Kush, western Pamir, and western South Tien Shan, and (b) the westernmost Tibetan Plateau, eastern Pamir, and Tien Shan. The mean elevation of the profile is in black. (c) Profile of eastward components of velocity with 1σ error bars for the Tajik Depression, Pamir, and westernmost Tarim Basin. The solid black line is the mean elevation of the profile.

[11] North-south profiles across the western and central Pamir (Figures 3a and 3b) suggest different distributions of strain and accommodation of India's convergence with Eurasia. The western transect (Figure 3a) from Peshawar (NCEG) across the westernmost Pamir (ISHA and MANM) to the South Tien Shan (GARM, KHAA, and KUM4) shows that 10 mm/yr of convergence is distributed across the Chitral Himalaya and Hindu Kush (Figure 2a) over a zone approximately 200 km wide, that 11 ± 1 mm/yr occurs across the northern margin of the Pamir and the Vakhsh Valley, and that the remaining small amount of relative motion is accommodated north of the South Tien Shan. The eastern transect, from Leh, Ladakh (RSCL) across the eastern Pamir and Alai Valley, shows no convergence between westernmost Tibet and the eastern Pamir, but 15 ± 1 mm/yr across the eastern Alai Valley, with an additional 5 ± 2 mm/yr of relative motion accommodated in the South Tien Shan and in the Chatkal Ranges west of the Tien Shan. In both transects across the Pamir, no significant shortening occurs internally within the Pamir (Figures 2 and 3). A comparison of the two transects indicates that although Leh and Peshawar are at approximately the same latitude, their positions relative to stable India are different. Peshawar has nearly the Indian plate's velocity, but Leh moves northward as much as 10 mm/yr more slowly, suggesting a N-S component of shortening of at least 10 mm/yr in the Kashmir Himalaya and Pir Pinjal [Bilham et al., 2011]. Thus, the main zone of shortening across the Himalaya bends sharply southward between Peshawar and Ladakh as do the primary faults through the western Himalayan syntaxis.

[12] Although north-south shortening occurs mainly in localized zones on the margins of the Pamir, east-west extension occurs both on the margins and within the high terrain. The difference between weighted average eastward components of velocity in the eastern Pamir and the westernmost Tarim Basin suggests that they diverge at 5 ± 1 mm/yr and that between the western Pamir and easternmost Tajik Depression suggests divergence between them at 2 ± 1 mm/yr (Figures 2b and 3c). Thus, the easternmost edge of the Tajik Depression moves west at 7 ± 2 mm/yr with respect to the western edge of the Tarim Basin. The difference in eastward weighted average velocities between the eastern and western sides of the Pamir is 3 ± 1 mm/yr (Figures 2b and 3c), much of which is then recovered by shortening in the Tajik Depression. Averaging velocities of Pamir sites in two geographic bins provide a minimum bound on the internal extension; the difference in easterly components of velocity between Khorog (MANM) and our easternmost site (SHMA) gives internal extension of 6 ± 1 mm/yr (Figure 3c). This estimate omits some additional extension between the easternmost Pamir and Tarim Basin suggested by observations of normal faulting there [e.g., Arnaud et al., 1993; Brunel et al., 1994; Chevalier et al., 2011; Tapponnier and Molnar, 1979; Robinson et al., 2010].

[13] Significant shear does not appear to occur between the Tarim Basin and the Pamir [Zubovich et al., 2010] despite the presence of mapped strike-slip faults with Quaternary offsets [Chevalier et al., 2011; Cowgill, 2010; Robinson et al., 2007; Ruzhentsev, 1963; Sobel and Dumitru, 1997]. We measure only 1 ± 1 mm/yr of sinistral relative motion between the average northward velocity for all sites in the eastern Pamir and all sites in the western Tarim, or less than 2 ± 2 mm/yr along faults striking NW-SE. In contrast, the difference between the average northward components of velocities of sites in the western Pamir and in the Tajik Depression shows that the Tajik Depression margin accommodates 10 ± 1 mm/yr of sinistral shear, which may be localized entirely on the Darvaz-Karakul Fault, given geological evidence suggesting a similar rate [Kuchai and Trifonov, 1977; Trifonov, 1978]. Some of this shear may also be accommodated over a wider zone by counter-clockwise rotations, as implied by paleomagnetic declinations from samples along the eastern margin of the Tajik Depression [Bazhenov and Burtman, 1981, 1982, 1986, 1990; Burtman and Molnar, 1993; Thomas et al., 1994]. Rates of rotation for the Tarim Basin and Ferghana regions from geodetic observations have been presented previously [e.g., Reigber et al., 2001; Zubovich et al., 2010] and these observations do not provide additional constraints.

4 Discussion

[14] The kinematic budget for Indian-Asian collision across Central Asia differs from that of the Tibetan region to the east in two ways. First, in Central Asia, most of the relative motion between India and Eurasia is accommodated in relatively localized deforming zones, with large convergence rates over short baselines implying high strain rates; in Tibet, relative motion is accommodated in both localized and broadly distributed deforming zones. From the Pakistani lowlands across Central Asia to the Kazakh platform of Eurasia, this total velocity is separated into ~10 mm/yr across the combined Chitral Himalaya and Hindu Kush, ~15 mm/yr about the Trans-Alai Thrust, and <5 mm/yr across the South Tien Shan farther north. In contrast, across the central Himalaya and Tibet, ~20 mm/yr is accommodated across the Himalaya, ~10 mm/yr by conjugate strike-slip faulting throughout the Tibetan Plateau, and ~5 mm/yr across the Qilian Shan on the northwestern edge of Tibet [Zhang et al., 2004].

[15] The Trans-Alai Thrust, a system of active Quaternary faults, accommodates shortening on the north side of the Pamir [Arrowsmith and Strecker, 1999; Nikonov, 1974, 1977; Nikonov et al., 1983; Strecker et al., 1995]. The southward dipping zone of intermediate depth seismicity projects to the surface near the trace of Trans-Alai Thrust and its westward continuation [e.g., Burtman and Molnar, 1993; Hamburger et al., 1992]. Lower bounds on Quaternary slip rates are consistent with the present-day geodetic observations [Mohadjer et al., 2010; Reigber et al., 2001; Zubovich et al., 2010]. These observations combine to make a strong case for subduction, or at least underthrusting, of lithosphere beneath the Alay Valley under the northern edge of the Pamir. That lithosphere presumably continues west and underlies the Tajik Depression.

[16] The Tajik Depression is underlain by relatively thin crust, which is overlain by ~10–15 km of Mesozoic and Cenozoic sedimentary rock [Kulagina et al., 1974]. Near the base of that sedimentary rock is a thin layer of Jurassic salt, along which the overlying sedimentary rock detaches to form folds throughout the Depression [e.g., Bekker et al., 1974a, 1974b, 1983; Nikolaev, 2002]. Bourgeois et al. [1997] estimated as much as 150 km of horizontal shortening across the Depression. Moreover, Leith [1985] showed that the subsidence history of the basin resembled that associated with cooling of the plate beginning in Jurassic time. He inferred that crust had been stretched and thinned at that time. If the crust beneath the Tajik Depression continued eastward beneath the territory now occupied by the Pamir, then the lithosphere there would have been thick, comparable to that of oceanic lithosphere of the same age (~100 km), but capped by thin crust (≤20–25 km). For these reasons, Hamburger et al. [1992] and Burtman and Molnar [1993] inferred that the intermediate depth earthquakes occur in an eastward continuation of the lithosphere that underlies the Tajik Depression.

[17] The other main contributor to the total displacement budget in Central Asia is not, however, distributed shortening throughout the high plateau, as it is in Tibet, but a second relatively localized zone of thrusting and shortening across the Chitral Himalaya and Hindu Kush, intersecting the surface between Malakand Pass (MALA) and Dir (QLND) (Figure 1a). Making the simplest possible interpretation that this shortening is accommodated on a planar thrust dislocation separating downgoing Indian lithosphere from overriding Hindu Kush material, we may draw a plane from this location to meet the top of the intermediate depth seismicity (at 100 km) in the Hindu Kush. Doing so constrains the dip of the top of Indian lithosphere to be 25°–30°, with a minimum total offset of ~500 km, more consistent with previous inferences of northward subduction of Indian lithosphere [Koulakov and Sobolev, 2006; Negredo et al., 2007; Priestley et al., 2008], than interpretations of Hindu Kush seismicity isolated in a relict piece of lithospheric material [Pavlis and Das, 2000; Pegler and Das, 1998] unconnected to surface tectonics. Furthermore, attributing shortening in the Chitral region to convergence between India and the Pamir requires a sharply bent geometry for the India-Eurasia subduction interface in map view for it to be located far to the south of Ladakh in its eastward extension. Whether the origin of either the north dipping slab beneath the Hindu Kush or the south dipping slab beneath the Pamir is entirely continental, however, requires additional data from seismology, xenolith studies, and other sources. Existing tomography [e.g., Koulakov and Sobolev, 2006; Mechie et al., 2011; Negredo et al., 2007; Priestley et al., 2008] does not have sufficient resolution to demonstrate continuity of a subducting slab.

[18] Although the shortening in Central Asia is localized to two zones of relatively rapid convergence, this is not the case with the east-west extension. The total rate of east-west extension is at most 10 mm/yr from the easternmost edge of the Tajik Depression (thus ignoring shortening therein) to the Tarim Basin. As this is less than half of the total shortening rate across the whole region, it precludes overall planar pure shear deformation. Moreover, the internal east-west extension is much greater than the internal shortening within the Pamir region itself, since most, if not all, of the total shortening occurs across the Pamir margins. As appears to be the case in Tibet, the surface area of the Pamir must be increasing. Insofar as these rates apply to the entire lithosphere, the crust must be thinning. Thus, to remain in isostatic equilibrium, the surface should be subsiding [England and Houseman, 1989; Fleitout and Froidevaux, 1982]. Given that the east-west extension rate within the Pamir is close to the shortening rate across the Tajik Depression, this thinning may be accommodated by the advection of material from the Pamir into the Tajik Depression and, to a considerably lesser extent, into the Tarim Basin. We note too, however, that shortening across the Tajik Depression seems to be thin skinned and limited to the upper 10–15 km of sedimentary rock [Bekker et al., 1974a, 1974b, 1983; Burtman and Molnar, 1993; Kulagina et al., 1974], but extension within the Pamir must involve much the crust, if not all of it, given the deep exhumation by tectonic, not erosional processes [e.g., Ducea et al., 2003; Hacker, 2005; Schmidt et al., 2011].

[19] Also, GPS velocities give no indication of shortening on the eastern edge of the Pamir where it abuts against the stable Tarim Basin. Yet, that margin reveals abundant evidence of folding and thrust faulting in early Cenozoic time [e.g., Bershaw et al., 2012; Yin et al., 2002]. Thus, it appears that convergence at that margin has ceased. As the northward movement of Tarim toward stable Eurasia at near 20 mm/yr cannot have been sustained for more than ~10 Ma [Abdrakhmatov et al., 1996; Reigber et al., 2001; Zubovich et al., 2010], we presume that the thrust faulting, folding, and convergence along the eastern margin of the Pamir ceased at approximately the same time.

[20] We may use these kinematic observations to place some bounds on the magnitudes of the principal stresses within the Pamir, assuming that shear stresses on horizontal planes are negligible. Because there is negligible internal north-south strain, the north-south compressive stress must approximately equal the vertical compressive stress. The east-west compressive stress, however, must be smaller than either of these. This is consistent with the topographic form of the region, where large basins east and west of the Pamir allow accommodation space for lateral extrusion of material. This is also consistent with limited focal mechanisms from the Pamir (, most of which show normal faulting on north-south trending planes and therefore east-west extension [e.g., Burtman and Molnar, 1993; Strecker et al., 1995]. In fact, the absence of east-west components of crustal shortening, except in the Tajik Depression where salt allows lateral transport of overlying sedimentary rock, might imply that the vertically averaged east-west compressive stress is much smaller than the vertically averaged north-south compressive stress. At present, it appears that the Pamir is both moving northward and collapsing laterally, suggesting that it is unlikely to reach the size of the neighboring Tibetan Plateau and that the properties of the lithosphere or the boundary conditions also differ from those of Tibet.


[21] This research was supported by the National Science Foundation under grants EAR-0636080 and 0636092. The authors thank two anonymous reviewers for their constructive efforts. We also thank the Civil Defense Research Fund for logistic assistance. Some of the GPS data from the Tien Shan were obtained with support from NSF grants EAR-8915334, EAR-9117889, EAR-9614302, EAR-9708618, and EAR-0636092 and NASA through grants NAG5-1941 and NAG5-1947.