Variable structural style along the Karakoram fault explained using triple-junction analysis of intersecting faults

Understanding the pattern of deformation that occurs where two or more fault systems intersect is a basic problem in determining how fault systems help accommodate continental deformation during orogenesis. Several studies have addressed this problem. Bohannon and Howell (1982) examined the intersection between the San Andreas and Garlock faults and argued that slip on the Garlock fault may have influenced the formation of the Big Bend and the Big Pine fault. Work by Wang et al. (1998) demonstrated that slip along the Xianshuihe–Xiaojiang fault resulted in an ∼60-km-wide deflection of the Red River fault. Additional studies of the San Andreas system and Eastern California shear zone have used geodetic data to show that fault intersections are often associated with increased strain rates (King and Cocco, 2001; Snay et al., 1996) and may result in transient strain accumulation, secular variation in fault slip rates, and earthquake clustering (e.g., Peltzer et al., 2001). Still other studies have shown that fault intersections play fundamental roles in how fault systems geometrically evolve and transfer strain throughout an orogen (Ando et al., 2004; Spotila and Anderson, 2004; Van der Woerd et al., 1999). Most of these previous studies have focused on small (<100-km-long) faults and have been geographically restricted to the San Andreas and Red River fault systems. In addition, the origin, and thus general kinematic significance of fault junctions such as that defined by the intersection of the San Andreas and Garlock faults, remains an open question (e.g., Bohannon and Howell 1982; Spotila and Anderson, 2004). This deficiency in the understanding of the geometric and kinematic evolution of intersecting major faults warrants further investigation of natural examples of such systems.

A number of very large active strike-slip fault systems cut the Indo-Asian collision zone (Fig. 1, inset), the largest active continental orogen on Earth. As a result, this collision serves as an excellent natural laboratory to investigate the geometric and kinematic evolution of intersecting major faults (Armijo et al., 1989; Molnar and Tapponnier, 1978; Peltzer et al., 1989; Tapponnier and Molnar, 1977). These structures have strike lengths in excess of 1000 km and cumulative displacements of several hundred kilometers or more (Armijo et al., 1989; Molnar and Tapponnier, 1975; Molnar and Tapponnier, 1978; Peltzer et al., 1989; Tapponnier and Molnar, 1977). One of the clearest zones of intersection between two major strike-slip faults within this collision zone occurs in western Tibet, where the 325°-striking, right-slip Karakoram fault lies near the southwest end of the Gozha–Longmu Co extension of the 070°-striking, left-slip Altyn Tagh fault (Fig. 1) (e.g., Peltzer et al., 1989). The active, left-slip Gozha–Longmu Co fault system separates the Tianshuihai terrane to the northwest from the Tibetan Plateau to the southeast (Fig. 1) (Liu, 1993). Evaluation of the extent to which the Altyn Tagh and Karakoram faults may influence one another requires investigation of the geometric and kinematic evolution of the zone of intersection between them (Avouac and Tapponnier, 1993; Molnar and Tapponnier, 1975, 1977; Peltzer and Tapponnier, 1988; Tapponnier and Molnar, 1976; Tapponnier et al., 1982).

The Altyn Tagh fault (Fig. 1) extends for over 1200 km and separates the Tibetan Plateau to the south from the Tarim Basin to the north (Molnar et al., 1987; Peltzer et al., 1989; Tapponnier and Molnar, 1977). Total left-slip along the fault exceeds 450 km (Cowgill et al., 2003; Peltzer et al., 1989; Ritts and Biffi, 2000; Ritts et al., 2004; Yue et al., 2001, 2004a, 2004b, 2005). From 87°E to 90°E, the fault is thought to extend to the base of the lithosphere (Wang et al., 2003; Wittlinger et al., 1998; Zhao et al., 2006). The Karakoram fault extends over 1000 km and separates the Pamirs and NW Himalayas to the southwest from the Tibetan Plateau to the northeast. Total right-slip along the Karakoram fault is disputed, but it appears to be between 150 and 500 km along the central portion of the fault, which lies between 34° to 36°N latitude (Lacassin et al., 2004a; Phillips et al., 2004; Searle, 1996). Its vertical extent remains to be determined.

Previous work along the Karakoram fault has established that the late Miocene to recent slip direction along the fault north of its intersection with the Gozha–Longmu Co fault (34.5°N latitude) differs from that observed to the south. Along the northern Karakoram fault, transpressional deformation is suggested by Neogene to Quaternary fault strands that have both thrust and strike-slip kinematics (Searle et al., 1998) and apatite fission-track ages that indicate 5 Ma to recent rapid exhumation of the ranges flanking the fault (Foster et al., 1994). In contrast, transtensional deformation is suggested along the southern portion of the Karakoram fault by active normal faults and tension gash data (Murphy and Burgess, 2006; Murphy et al., 2000, 2002; Ratschbacher et al., 1994), as well as reconstructed offset terraces and moraines (Brown et al., 2002; Chevalier et al., 2005). Although this variation in structural style is a first-order characteristic of late Cenozoic slip along the Karakoram fault, the cause of this variation remains poorly understood.

Here we investigate the extent to which this pronounced variation in structural style along the Karakoram fault may be due to the relative motions between the NW Himalaya, the Tianshuihai terrane, and the Tibetan Plateau (Fig. 1) and associated slip along the Gozha–Longmu Co extension of the Altyn Tagh fault. We present new (Fig. 2, red faults) and compiled (Fig. 2, orange, yellow, and blue faults) neotectonic mapping of the region encompassing the intersection of the western Altyn Tagh and Karakoram faults. We then review previously published work on the slip rates of these faults and combine them with the newly mapped fault geometries to construct a simple kinematic model of the intersection of the Altyn Tagh and Karakoram faults using the principles for plate triple junctions (McKenzie and Morgan, 1969). Although our analysis is broadly similar to that of Liu (1993), here we are focused on understanding along-strike variation in structural style along the Karakoram fault and the role of the Longmu Co fault, if any, in producing this variation. Our work suggests that rigid-block kinematics provides an accurate first-order depiction of continental deformation in this region.

Several important studies have reported neotectonic mapping from western Tibet (Armijo et al., 1986, 1989; Peltzer et al., 1989), and the most extensive work has been reported in a Ph.D. thesis by Liu (1993). To build on this work and better determine the geometry and kinematics of active faulting in western Tibet, we compiled a new neotectonic map of the area using the published maps, shown as orange, yellow, and blue faults in Figure 2, and new results from our own analysis of remotely sensed imagery locally augmented with field mapping, shown as red lines in Figure 2. Our work had three aims: (1) to reproduce the earlier results, (2) to determine if the Gozha–Longmu Co fault system is geometrically linked with the Altyn Tagh fault, and (3) to identify and map in detail the intersection between the Karakoram and Gozha–Longmu Co faults. Although Peltzer et al. (1989) hypothesized that the Gozha–Longmu Co fault system links with the Altyn Tagh fault, detailed maps of this area have not been reported; thus, this linkage remained to be definitively established until this study. Likewise, detailed mapping of the intersection between the Karakoram and Gozha–Longmu Co faults also remained to be reported. Confirmation of these linkages suggests that the major faults of western Tibet form a fault circuit, and this justifies the use of the triple junction approach to construct an initial kinematic model of western Tibet. The new compilation is provided as an interactive, Web-based map (located at http://dx.doi.org/10.1130/GES00067.S1 and henceforth referred to the as the Web Map), and it is shown in simplified form in Figure 2.

The previously published neotectonic maps from western Tibet fall into two broad categories: (1) regional compilation maps (e.g., Armijo et al., 1986, 1989), which generally include minimal observational data in support of the reported mapping, and (2) detailed maps of 10–50-km-long reaches of the faults, in which the geomorphic criteria used to deduce the fault geometry and kinematics are clear but must be extrapolated over large distances to infer the kinematics along the entire length of the fault (e.g., Liu, 1993; Peltzer et al., 1989). The neotectonic mapping presented in this study (Web Map and Fig. 2) builds on this previous work in three ways. First, it offers the opportunity to simultaneously view both the mapping and the primary data from which the mapping was derived over the whole area, thereby permitting users to independently evaluate the geomorphic criteria we used to generate the map. This capacity is particularly important for regions that we interpret to show minimal geomorphic evidence of active faulting, such as in the center of the Tianshuihai terrane (Fig. 1). Second, the level of mapping detail we report is comparable to that in the detailed maps presented by Liu (1993) and Peltzer et al. (1989), but it spans an area comparable to that covered in summary form by the regional work of Armijo et al. (1986).

To identify geomorphic evidence of active faulting between 33° and 37°N latitude and 76° and 83°E longitude, we initially used the Real-Time Interactive Mapping System (RIMS), a new interactive three-dimensional (3-D) visualization and analysis tool that allows extraction of surface information through real-time, georeferenced vector mapping on a virtual terrain model from any perspective (see Fig. 3). In RIMS, bands 7, 4, and 2 from a mosaic of Landsat 7 Thematic Mapper (TM) (http://glcf.umiacs.umd.edu/index.shtml) scenes with a resolution of 28 m/pixel were draped over a Shuttle Radar Topographic Mission digital elevation model (DEM), which has a resolution of ∼90 m/pixel. File-size limitations in RIMS necessitated a second iteration of mapping at a higher resolution in Google Earth Plus! (http://earth.google.com) using bands 1, 2, and 3 from a mosaic of Landsat 7 Enhanced Thematic Mapper Plus (ETM+) scenes (http://onearth.jpl.nasa.gov/), pan-sharpened using panchromatic band 8 to increase the spatial resolution of these bands to 14.25 m/pixel. Where possible, the kinematics of previously unmapped faults were determined using standard geomorphic criteria (e.g., Molnar and Tapponnier, 1978).

The dynamic Web-based map (Web Map) and Figure 2 support previous work indicating strike-slip motion along four major fault zones. In particular, the Altyn Tagh, Karakax, Gozha–Longmu Co, and Karakoram faults dominate active faulting in the region. Geomorphic evidence of large dip-slip structures occurs in only two areas. The first is along the northern margin of the Western Kunlun Shan (Figs. 1 and 2), where we infer the location of the blind Hotan thrust from the northernmost extent of the region of recent incision, as described by Avouac and Peltzer (1993). The second case occurs in the southeastern portion of the mapped area (Fig. 2), where diffuse N-S–striking normal faults are the predominant active structures (e.g., Tapponnier et al., 1981). Most of the active deformation occurs along the margins of the Tianshuihai terrane, within which we found no clear geomorphic evidence of active internal deformation (Web Map and Fig. 2). This lack of geomorphically expressed active faulting contrasts with recent Interferometric Synthetic Aperture Radar (InSAR) measurements (Wright et al., 2004), suggesting that the observed InSAR signal may result from atmospheric interferences, as suggested by Wright et al. (2004), or from recent initiation of rapid slip along a fault system that is poorly expressed geomorphically. The geomorphic setting of this locality on the high and arid Tibetan Plateau makes it unlikely that the InSAR data reflect a long-lived, fast-slipping fault that has no geomorphic expression because the pace of geomorphic resurfacing outpaces the rate of surface deformation.

Along the Altyn Tagh fault, well-defined fault scarps, shutter ridges, and both beheaded and deflected drainages that show left-separations characterize the deformation in the vicinity of 83°E (Web Map and Fig. 3A). To the west, the active trace of the Altyn Tagh fault splays and becomes less discrete as it approaches the intersection between the Karakax and Gozha–Longmu Co faults between 81°E and 83°E (Web Map) in a geometry similar to that mapped previously (Chinese State Bureau of Seismology, 1992). The Karakax fault, commonly referred to as the western extension of the Altyn Tagh fault (e.g., Peltzer et al., 1989), strikes 110° and forms the southern boundary of the Western Kunlun Shan (Web Map and Fig. 2). We differentiate the Karakax and Altyn Tagh faults here because they have significantly different regional orientations. Along the Karakax fault, left-laterally offset fluvial terraces and glacial moraines clearly mark the trace of the fault within the Karakax valley (Ding et al., 2004; Peltzer et al., 1989; Ryerson et al., 1999) (Web Map and Fig. 3B). However, the active trace of the Karakax fault becomes poorly expressed for ∼150 km between ∼79°E, where it leaves the Karakax valley, and ∼81°E, where it joins one of the main splays of the Altyn Tagh fault (Web Map).

The 070–090°–striking Gozha–Longmu Co fault system (Liu, 1993) extends from the Altyn Tagh fault in the northeast to the Karakoram fault in the southwest (Web Map). The northeastern end of the fault system lies between 81.5°E and 83°E, where NE-SW–striking faults associated with predominantly extensional earthquake focal mechanisms (Fig. 1) mark its trace. This set of faults merges with the diffuse splays of the southwestern Altyn Tagh fault and appears to transfer left-slip from the Altyn Tagh fault to the Gozha Co fault. As the Web-based map (Web Map) and Figure 3C indicate, drainages crossing the Gozha Co fault show sinistral separations. To the southwest, a geometrically complex set of left-stepping structures in the vicinity of Longmu and Sumxi lakes transfers displacement along the Gozha Co fault to the left-slip Longmu Co fault, as was also shown by Liu (1993).

Our mapping along the Gozha–Longmu Co fault system also appears to constrain the total slip along the system to 25–32 km. Along three widely separated reaches of the system, bedrock markers appear to show total separations of ∼25 km (Fig. 4A), ∼32 km (Fig. 4B), and ∼28 km (Fig. 4C). We identified these separations by matching contacts between regions of differing spectral response (and thus color) in the Landsat image on opposite sides of the fault. Although we have not completed bedrock mapping of these areas to confirm the proposed correlations, our experience in the area comparing our field mapping with color changes in equivalent Landsat images suggests that the separations shown in Figure 4 likely reflect true bedrock displacements. Therefore, 25–32 km constitutes the first rough estimate of total left slip along the Gozha–Longmu Co fault system.

The Karakoram fault strikes 325° and extends from near the Kongur Shan in the north to the Gurla Mandhata detachment in the south (Fig. 1). The Karakoram fault is characterized by two geomorphically distinct regions that are separated by a restraining double bend in the fault near its intersection with the Longmu Co fault (Web Map; Figs. 2 and 3D). This bend is located at 34.5°N latitude and is ∼27 km wide, as measured perpendicular to the trace of the Karakoram fault outside of the bend. To the north of this intersection, the occasional occurrence of a distinct fault trace showing right-laterally offset terraces weakly defines the trace of the Karakoram fault. Conversely, to the south, fault scarps that simultaneously expose steep triangular facets and right-laterally offset glacial moraines attest to oblique, right-normal motion (Fig. 3E) and clearly define the trace of the Karakoram fault. As Figure 2 indicates, fault-perpendicular topographic profiles across the northern and southern reaches of the Karakoram fault also highlight the differences in the geomorphic expression of the fault north and south of the bend. In particular, the northern Karakoram fault lies in the bottom of a deep axial valley that is flanked on both sides by steep topography, whereas the southern Karakoram fault lies within an axial valley that is both shallower and wider than that to the north. Along the northern Karakoram fault, earthquake focal mechanisms (Fig. 1) and observations of faults with thrust and strike-slip motions (Searle et al., 1998) indicate transpressional kinematics, further supporting this morphologic distinction. As the following kinematic analysis demonstrates, the geomorphic and kinematic differences between the northern and southern Karakoram fault near its intersection with the Longmu Co fault have significant implications.

In addition to fault geometries and slip directions, we must also know the slip rates before we can construct a kinematic model of western Tibet. The slip rates on the Karakoram, Altyn Tagh, and Karakax faults as well as the rate of thrusting in the Western Kunlun Shan are all particularly critical. Because geodetic and Quaternary rates along these structures show systematic differences, we use both in our kinematic analysis. ⁰¹ presents a compilation of known slip rates for these structures.

Reported slip rates along the Altyn Tagh and Karakoram faults appear to vary significantly, depending on the method used to obtain them. For instance, reconstructions of cosmogenically dated (¹⁰Be) fluvial terraces and moraines yield slip rates of 26.9 mm/yr along the central Altyn Tagh fault (Meriaux et al., 2004) and 10.7 mm/yr (Chevalier et al., 2005) or 4 mm/yr (Brown et al., 2002) along the southern Karakoram fault. In contrast, slip rates determined using global positioning system (GPS) measurements for the Altyn Tagh and Karakoram faults are 9 mm/yr (Bendick et al., 2000; Shen et al., 2001) and 4 mm/yr (Jade et al., 2004), respectively. We favor the geodetic rates because they better explain the timing of exhumation of mid-crustal rocks along the southern Karakoram fault, more closely approximate the Cenozoic-averaged slip rates for these faults (Cowgill et al., 2003; Lacassin et al., 2004a; Phillips et al., 2004; Searle, 1996; Yin et al., 2002; Yue et al., 2001, 2004a), and do not involve disputed age determinations for offset features as discussed by Brown et al. (2005), Cowgill (2007), and England and Molnar (2005).

Published literature contains a range of kinematic models of continental deformation, including two-dimensional (2-D) rigid-block models (e.g., Jezek et al., 2002), microplate models that use Euler poles to describe relative motions between plates on a sphere (e.g., Avouac and Tapponnier, 1993; Liu, 1993; Peltzer and Saucier, 1996; Replumaz and Tapponnier, 2003), and continuum models that employ a wide range of initial and boundary conditions, as well as constitutive relationships (e.g., Bird and Piper, 1980; England and McKenzie, 1982; Vilotte et al., 1982). The simplest of such kinematic models are 2-D, “flat-earth” models, in which discrete block boundaries are used to approximate complex fault systems. Here we start with this most basic approach to explore the first-order kinematics of western Tibet. In particular, we aim to evaluate if relative motions between the NW Himalaya, the Tianshuihai terrane, and Tibet can produce both the along-strike variation in structural style along the Karakoram fault and the bend at its intersection with the Longmu Co fault.

A critical component of such a block model is the intersections of major block-bounding faults, which should define triple junctions, and the kinematic evolution of these junctions, which should produce diagnostic variations in structural style along these faults (e.g., Liu, 1993; Peltzer and Tapponnier, 1988). The new neotectonic mapping described herein and shown as red lines on the Web-based map and Figure 2 confirms the linkage between the Altyn Tagh, Karakax, Gozha–Longmu Co, and Karakoram faults and justifies the triple junction approach of our initial kinematic model. Kinematic analysis of isolated triple junctions has proven successful in explaining spatial and temporal variations in fault kinematics in both oceanic and continental deformation zones (e.g., Dickinson and Snyder, 1979; Kleinrock and Morgan, 1988).

Using the fault geometries and structural style from the new neotectonic mapping reported here, we simplified the mapped region into four internally rigid blocks: Tarim, Tianshuihai, Tibet, and the NW Himalaya. By using published slip rates for several of the block-bounding faults, we then solved for all of the relative motions between the blocks using the principles of plate triple junctions (i.e., velocity triangles) (McKenzie and Morgan, 1969). The discrepancy in geodetic and Quaternary slip rates has significant implications for this analysis, and for this reason, we calculated the relative motions between the four plates using both geodetic and Quaternary slip rates. Our analysis (1) predicts the direction and rate of slip along the Gozha–Longmu Co fault system, (2) explains along-strike variation in the kinematics of the Karakoram fault, and (3) produces a testable kinematic model for the evolution of the junction between the Karakoram and Longmu Co faults.

The analysis employs the following five assumptions: (1) The blocks are perfectly rigid. (2) The relative velocity along fault segments does not change, or the blocks do not rotate. (3) Motion along the Altyn Tagh and Karakax faults is parallel to the regional strike of these faults (i.e., pure strike-slip). (4) Multiple strands of the Gozha–Longmu Co fault system are simplified into a single fault to define a boundary between the Tibet and Tianshuihai blocks. (5) No estimate exists for the magnitude of transpression across the northern Karakoram fault; thus, only the azimuth of the fault trace and the magnitude of the slip rate are left to define relative motion between Tianshuihai and the NW Himalaya. Motion along the northern Karakoram fault is assumed to be pure strike-slip, but the effect of transpression along this segment of the fault is also considered (Foster et al., 1994; Searle et al., 1998).

In order to use the slip rates and fault geometries to calculate relative block motions, we calculated the set of velocity triangles shown in Figure 5, which proceeded along the lines of a similar analysis conducted by Liu (1993). Velocities derived from the published data shown in ⁰¹ are shown in black and represent the arithmetic averages of the fault slip rates published for the Hotan, Karakax, Altyn Tagh, and Karakoram faults. The colored values were derived according to the methods outlined in the remainder of this section and give average relative motions between Tarim and Tianshuihai (green), Tianshuihai and Tibet (blue), and Tibet and the NW Himalaya (red and purple). We used the following approach to account for uncertainty in the choice of vectors when constructing a velocity circuit. At each step, we first used all possible combinations of individual values for the two known legs to compute all possible individual values for the unknown third leg by assuming that each individual circuit closes. We then computed the arithmetic mean of these individual solutions to determine the derived vectors that are shown for each step in Figure 5. Thus, while the circuit for each individual solution was closed, the combinations of these average velocities shown in Figure 5 do not always close because these velocity triangles are not actual solutions. They are a comparison of the average known slip rates with the average of all unknown vectors that could be determined from the slip rates in ⁰¹. The extent to which the triangles shown in Figure 5 close thus reflects the extent to which the data from which they are derived are homogeneous (closed) or heterogeneous (open). Thus, the unclosed triangles derived for Tianshuihai-Tibet and Tibet–NW Himalaya motion using Quaternary slip rates in steps 2 and 3 of Figure 5 result from the large discrepancies in reported Quaternary slip rates along the Altyn Tagh and Karakoram faults. These large discrepancies result in an average derived motion that, when combined with the average of the known slip rates, does not result in a closed velocity triangle.

In the first step, we determined average Tarim-Tianshuihai relative motion (green vector in Fig. 5) by combining range-perpendicular shortening values across the Western Kunlun Shan with left-slip values along the Karakax fault, and the average values are shown as black vectors in Figure 5. Velocity triangles derived using geodetic and Quaternary slip rates yielded Tarim-Tianshuihai rates of 6.7 ± 2.9 mm/yr and 18.93 ± 5.1 mm/yr, respectively (Fig. 5). Relative to Tianshuihai, Tarim moves toward 265° ± 14° and 268° ± 11° in the geodetic and Quaternary velocity triangles, respectively.

The second step yielded the average for possible Tianshuihai-Tibet relative motions, shown as blue vectors in Figure 5. To do this, we combined the individual Tarim-Tianshuihai vectors computed in step 1 (averages of which are shown as green vectors in Fig. 5) with individual slip-rate values along the Altyn Tagh fault and an assumption of pure left slip along the Altyn Tagh fault to constrain the azimuth of the Tarim-Tibet vector (average Tarim-Tibet vectors are shown in black in Fig. 5 and are listed in ⁰¹). Because this step yielded average Tianshuihai-Tibet relative motion, it also predicted the rate and direction of slip along the Gozha–Longmu Co fault system. In particular, the blue vectors in the velocity triangle shown in step 2 of Figure 5 predict that Tibet moves at either 2.8 ± 1.1 mm/yr toward 030.9° ± 32.9° or 8.3 ± 2.7 mm/yr toward 302.7° ± 16.1° relative to Tianshuihai, depending on the use of either geodetic or Quaternary rates, respectively. Because the orientations of the relative motion vectors differ between these two solutions, they make significantly different predictions regarding the kinematics of the 070°–090°–striking Gozha–Longmu Co fault system. In particular, the motion of Tibet toward 031° relative to Tianshuihai in the geodetic solution predicts that the Gozha–Longmu Co fault system should have left-reverse transpressional kinematics. In contrast, the motion of Tibet toward 303° relative to Tianshuihai in the Quaternary solution predicts that the fault should have right-reverse transpressional kinematics, which conflicts with the observed kinematics reported here.

The third step yields average NW Himalaya–Tibet relative motions (red vectors in Fig. 5) using the individual solutions computed for the Tianshuihai-Tibet vector determined as part of step 2 (the average of which is shown as the blue vectors in Fig. 5), the individual slip-rate values from along the Karakoram fault, and an assumption of pure right slip along the northern Karakoram fault to constrain the azimuth of the Tianshuihai–NW Himalaya vector (the average Tianshuihai–NW Himalaya vectors are shown in black on Fig. 5 and are listed in ⁰¹). The resulting red vector in step 3a of Figure 5 shows the average predicted motion of Tibet relative to the NW Himalaya across the Karakoram fault south of its junction with the Longmu Co fault. Relative to the NW Himalaya, the geodetic and Quaternary models predict that Tibet moves at 6.3 ± 0.1 mm/yr toward 096.7° ± 18.4° or 13.3 ± 1.9 mm/yr directed toward 118.6° ± 11.6°, respectively.

Although both previous work and our observations attest to transpressional deformation along the northern Karakoram fault, the amount of fault-perpendicular shortening is unconstrained. In an attempt to evaluate the effect of this transpression on NW Himalaya–Tibet relative motion, we also considered an alternative solution for the third step in which we arbitrarily assumed 10° of oblique convergence across the northern Karakoram fault. This solution is indicated by step 3b in Figure 5. Incorporation of 10° of oblique convergence across the northern Karakoram fault has no significant effect on the NW Himalaya–Tianshuihai relative velocity vector in the Quaternary velocity triangle. In contrast, addition of a component of oblique convergence along the northern Karakoram fault to the geodetic velocity triangle predicts a more easterly motion of Tibet relative to the NW Himalaya at 6.3 ± 0.1 mm/yr toward 088.8° ± 23.7°, thereby increasing the fault-normal component of motion along the southern Karakoram fault. Extrapolation of this calculation indicates that motion along the southern Karakoram fault would be entirely normal at ∼55° of oblique convergence across the northern Karakoram fault; however, the velocity triangle would no longer close.

The new neotectonic mapping presented in the Web-based map and Figure 2 outlines three previously unrecognized observations about the active deformation of western Tibet. First, at its southwestern end, the Altyn Tagh fault bifurcates into two geometrically and kinematically complex fault networks, each of which is ∼100 km along strike. These fault zones strike roughly parallel to the Karakax and Gozha–Longmu Co faults (Fig. 2 and Web Map). Second, the southwestern end of the Longmu Co fault coincides with a 27-km-wide bend in the Karakoram fault. This bend is associated with a distinct change in topography along the strike of the Karakoram fault (Fig. 2B). Third, bedrock markers evident in Landsat imagery are displaced 25–32 km along the Gozha–Longmu Co fault system.

Taken together, these new observations, previously published work, and the kinematic analysis presented herein support four conclusions: (1) The Altyn Tagh fault is geometrically and kinematically linked with both the Karakax and Gozha–Longmu Co faults, although such linkages are both complex and poorly expressed geomorphically. Peltzer et al. (1989) and the Chinese State Bureau of Seismology (1992) both reached a similar conclusion on the basis of their own work and the absence of other suitable structures in the region. (2) The lack of geomorphic evidence for active deformation within the interior of the Tianshuihai terrane, and the concentration of active faulting into narrow zones along its margins with Tarim, the NW Himalaya, and Tibet, suggest that the Tianshuihai region has behaved essentially as a rigid block during the Quaternary. (3) The position of the 27-km-wide bend in the Karakoram fault at its intersection with the Longmu Co fault suggests that these two features are genetically related. (4) The Karakoram fault is transpressional to the north of this bend but transtensional to the south.

As mentioned already, the discrepancy between geodetic and Quaternary slip-rate determinations for the Karakax, Altyn Tagh, and Karakoram faults resulted in a significant difference in the calculated velocity triangles. The combination of average geodetic slip rates for the Western Kunlun Shan, Karakax, Altyn Tagh, and Karakoram faults and the averages of all possible solutions for the relative motions of Tarim, Tianshuihai, and Tibet calculated from them (Fig. 5) yields closed velocity triangles within their 2σ uncertainties. The triangles derived from the Quaternary slip rates for the same faults, however, yielded velocity triangles that failed to close, suggesting that either the assumption of perfectly rigid blocks is flawed or the slip-rate determinations are inaccurate. Nevertheless, Brown et al. (2005) suggested that the Quaternary slip rate determined by Chevalier et al. (2005) for the Karakoram fault could be revised to 4–5 mm/yr, and England and Molnar (2005) and Cowgill (2007) suggested that the slip rate along the central Altyn Tagh fault determined by Meriaux et al. (2004) could be revised to 9–10 mm/yr. These modifications bring the Quaternary slip rates into accordance with the geodetic ones. These data suggest that our assumption of perfectly rigid blocks is a valid approximation for western Tibet, and we proceed with the assumption that the results obtained from the geodetic slip rates present the most accurate description of late Cenozoic deformation of this region.

To explore the relationship between relative block motions in western Tibet and the geometry and kinematics of the Karakoram fault, we developed the reconstruction shown in Figure 6, which is in a NW Himalaya–fixed reference frame. To generate this reconstruction, we started with the present fault geometry and then retrodeformed the bend in the Karakoram fault by sliding the Tibet and Tianshuihai blocks back along their respective velocity vectors as derived using the geodetic triangle shown in step 3 of Figure 5. Implicit in this approach is an assumption that the plate kinematics remained constant during formation of the bend.

This reconstruction matches four first-order characteristics of the neotectonic map (Fig. 2), earthquake focal mechanisms, and previous work (Fig. 1): (1) The Karakoram fault is transtensional south of its intersection with the Gozha–Longmu Co fault system; however, the magnitude of transtension could vary depending upon the amount of transpression across the northern Karakoram fault. (2) The intersection of the Karakoram and Longmu Co faults coincides with a restraining double bend. (3) The Gozha–Longmu Co fault system is transpressional. (4) Faults linking the Gozha Co and Altyn Tagh faults are slightly extensional.

Previous geologic work along the Karakoram fault is consistent with our kinematic analysis of the Karakoram–Gozha–Longmu Co triple junction (Fig. 5). In comparison to the predicted east-west extension between NW Himalaya and Tibet, measurements of tension gashes and normal faults along the southern Karakoram fault and along extensional jogs of the Shiquanhe fault (Fig. 1) indicate a mean extension direction of 85° ± 28° (Ratschbacher et al., 1994). Similarly, Murphy et al. (2000) and Murphy and Burgess (2006) found a mean extension direction of 089° along the southernmost portion of the Karakoram fault. We also suggest that the units in the Pangong Range, which lies along the Karakoram fault between 34.5°N and 33°N (Fig. 1), may have been partially exhumed (Rolland and Pecher, 2001) due to transtension along the northernmost segment of the southern Karakoram fault. Strands of southern Karakoram fault that truncate glaciers have down-to-the-east kinematics, consistent with transtension along this mountain front (Searle et al., 1998).

Our kinematic analysis, and in particular the reconstruction shown in Figure 6, can also be used to predict the initiation age, total slip, and slip rate along the Longmu Co fault. To do this, we assume that the bend in the Karakoram fault and slip along the Gozha–Longmu Co fault are genetically related and that the present kinematics of the triple junction remained steady during bend formation. As Figures 2 and 6 indicate, the bend in the Karakoram fault at 34.5°N is ∼27 km wide when measured perpendicular to the strike of the fault outside the bend region. The size of this bend is in striking agreement with the estimates of 25–32 km for total left separation along the Gozha–Longmu Co fault system presented in Figure 4. Depending on whether we use the geodetic or Quaternary velocity triangles, the kinematic model in Figure 4 predicts that the bend has widened at either 2.8 ± 1.1 mm/yr or 8.3 ± 2.7 mm/yr, respectively. Extrapolation of these rates backward from the present-day to smooth the entire 27 km bend yields an age for the onset of bending of either 9.6 ± 2.8 Ma, if geodetic rates are used, or 3.2 ± 2.5 Ma, using Quaternary rates. Importantly, this analysis does not specify which fault deforms the other: left slip along the Gozha–Longmu Co fault system can be viewed as causing deformation of the Karakoram fault as a passive marker where the two faults intersect, or the formation of a bend in the Karakoram fault could have triggered development of the Gozha–Longmu Co fault system.

Based on these findings, Figure 7 outlines a new synthesis of the Miocene to Holocene evolution of the major structures of the western India-Asia collision zone and their relation to the deformational and thermal history of rocks exposed along the Karakoram fault. This synthesis extrapolates the results of the geodetic velocity triangles from the previous section over the past 20 m.y.; therefore, the proceeding conclusions are speculative since they assume that the geodetic slip rates are similar to the long-term geologic rates. In particular, if there are numerous, slowly slipping faults within the area that are not well expressed in the geomorphology or GPS data but that have moved faster in the past and/or integrate to regionally significant deformation rates, then our analysis could be in error. Nevertheless, the concurrence between the geodetic and geologic slip rates along the Altyn Tagh and Karakoram faults, as discussed in the previous section, supports our assumption that the geodetic slip rates are grossly similar to the long-term geologic rates in this area.

In accordance with the findings of Dunlap et al. (1998), we propose that the Karakoram fault has experienced three main phases of deformation, which we argue were genetically linked with the structural evolution of the southwestern termination of the Altyn Tagh fault and other major structures in the western India-Asia collision zone. Nevertheless, we must emphasize that the schematic reconstruction shown in Figure 7 is largely based on temporal associations from isolated regional studies along the major structures in the region; thus, the extent to which these structures are kinematically linked in a regional deformation zone, and therefore have a cause and effect relationship, remains undetermined.

Figure 7A shows the middle Miocene configuration of the western India-Asia collision zone. Initial slip along the Karakoram fault was likely transpressional on the basis of ⁴⁰Ar/³⁹Ar cooling ages from amphibole, muscovite, biotite, and potassium feldspar from the Baltoro granite and the Pangong gneisses, which indicate rapid cooling from 17 to 13 Ma related to vertical exhumation along the Karakoram fault (Fig. 7A) (Dunlap et al., 1998; Searle and Tirrul, 1991; Searle et al., 1998). Furthermore, initial slip along the Karakoram fault likely began in the vicinity of the Baltoro granite, the unit with a maximum undisputed displacement of ∼150 km along the Karakoram fault (Phillips et al., 2004; Searle, 1996; Searle et al., 1998). U-Pb zircon ages from leucocratic dikes in the Pangong Range, which are the youngest rocks known to be cut by the fault (Phillips et al., 2004), constrain the initiation age of the Karakoram fault to 15.68 ± 0.53 to 13.73 ± 0.28 Ma. It could be as old as 30 Ma, however, depending on whether U-Pb zircon ages from the same region reflect synkinematic (Lacassin et al., 2004a, 2004b) or prekinematic (Searle and Phillips, 2004) magmatism. Apatite fission-track cooling ages from the eastern Pamirs and Western Kunlun Shan indicate that exhumation began during this time (Sobel and Dumitru, 1997) and suggest that the Main Pamir and Kumtagh fault system initiated at this time. The Main Pamir–Kumtagh fault system may have linked with the Tikilik fault to the north of the Western Kunlun Shan (Cowgill, 2001; Sobel and Dumitru, 1997). Finally, we infer that Cenozoic slip along the Karakax fault began at ca. 13 Ma by assuming that the average geodetic slip rate of 6 mm/yr (Shen et al., 2001; Wright et al., 2004) has been constant throughout the development of the ∼80-km-long left-lateral deflection of the Karakax river where it crosses the Karakax fault (Ding et al., 2004). Arnaud et al. (2003) presented additional ⁴⁰Ar/³⁹Ar potassium feldspar thermochronologic data to support activity along the Karakax fault during this time.

Figure 7B shows the late Miocene configuration of the western India-Asia collision zone. Potassium feldspar ⁴⁰Ar/³⁹Ar thermochronologic data indicate very slow cooling during this time period (Dunlap et al., 1998), and ⁴⁰Ar/³⁹Ar muscovite cooling ages of ca. 11 Ma on either side of the Karakoram fault indicate that both sides of the fault were at equivalent structural levels at this time, suggesting pure strike-slip motion (Fig. 7B, plot) (Searle and Tirrul, 1991; Searle et al., 1998). Currently, no evidence exits to support a similar cessation of vertical exhumation along the other major structures in the western India-Asia collision zone, and the cause for this transition along the Karakoram fault remains to be determined.

Finally, Figure 7C shows the proposed late Miocene–Pliocene to Holocene configuration of the western India-Asia collision zone. Apatite fission-track cooling ages from the northern Karakoram fault indicate that vertical exhumation began at ca. 5 Ma along the northern Karakoram fault (Foster et al., 1994). The ⁴⁰Ar/³⁹Ar potassium feldspar thermochronology (Dunlap et al., 1998), ⁴⁰Ar/³⁹Ar cooling ages (Rolland and Pecher, 2001), and apatite fission-track cooling ages (Arnaud, 1992) from the southern Karakoram fault suggest that renewed vertical exhumation began somewhat earlier along this reach of the fault at ca. 8 Ma (Fig. 7C, plot). We propose that this renewed vertical exhumation is related to the ca. 9 Ma initiation of the Gozha–Longmu Co fault system as indicated by our kinematic analysis. Therefore, vertical exhumation along the northern Karakoram fault relates to transpressional motion, whereas vertical exhumation along the southern Karakoram fault relates to transtensional motion. This sequence differs from the scenario outlined by Searle et al. (1998) and Dunlap et al. (1998), who suggested that renewed vertical exhumation was related to transpressional deformation along the entire length of the Karakoram fault. This change in kinematics along the Karakoram fault and initiation of the Gozha–Longmu Co fault system was also contemporaneous with a change in the sedimentation rate in the Tarim Basin to the north of the Western Kunlun Shan at ca. 5 Ma (Zheng et al., 2000). We infer that this change was related to initiation of slip along the blind Hotan thrust and cessation of slip along the Tikilik fault, which may have resulted from reorganization of the southwestern Altyn Tagh fault. The initiation of normal faulting along the Kongur Shan extensional system at ca. 8 Ma (Robinson et al., 2004) also coincided with the inferred initiation age of the Gozha–Longmu Co fault system and the onset of renewed cooling along the Karakoram fault. Furthermore, the late Miocene reorganization of the western India-Asia collision zone outlined in Figure 7C occurred at approximately the same time as the onset or rejuvenation of normal faulting across the entire Tibetan Plateau (Pan et al., 1993; Yin, 2000; Yin et al., 1999; Dunlap et al. 1998).

In addition to the Karakoram–Longmu Co fault junction, a number of other intersecting strike-slip faults are known. Examples of simultaneously active, major (along-strike length >50 km), strike-slip faults that intersect include the San Andreas and Garlock faults (Bohannon and Howell, 1982; Spotila and Anderson, 2004), the Blackwater and Garlock faults in the Eastern California shear zone (Peltzer et al., 2001), the Red River and Xianshuihe–Xiaojiang faults in eastern Tibet (Wang et al., 1998), and the North and East Anatolian faults in eastern Turkey (Şengör et al., 1985). Although more work on these systems is needed to advance our understanding of the crustal-scale geometry of these intersections and their kinematic evolution, current work suggests that intersecting strike-slip faults fall into two broad classes: those with surface traces that intersect, in which case one fault truncates the other, and those for which the surface traces cross but in detail fail to intersect. Although the surface trace of the San Andreas fault clearly intersects and truncates the Garlock fault, and Bohannon and Howell (1982) proposed that slip along the Garlock fault has played a role in altering the geometry of the San Andreas fault, there seems to be minimal evidence in support of this model (Spotila and Anderson, 2004). In contrast, the surface traces of the Blackwater and Garlock faults do not appear to directly intersect one another, although the traces of both faults continue on either side of the inferred zone of intersection, so the faults do, at some level, appear to be mutually crosscutting. InSAR data indicate that slip on one fault directly impacts slip on the other (Peltzer et al., 2001). The surface traces of the Red River and Xianshuihe–Xiao jiang faults likewise do not directly intersect one another, although, again, these traces do cross and continue beyond the zone of fault intersection. In this instance, the surface trace of the Red River fault is deflected by an amount that is equivalent to the total slip on the Xianshuihe–Xiaojiang fault (Wang et al., 1998), suggesting that slip along the latter has deformed the former. The intersection of the Karakoram and Gozha–Longmu Co faults shares characteristics of intersecting strike-slip faults in both classes. Like the San Andreas–Garlock intersection, the surface trace of the through-going Karakoram fault clearly truncates the Gozha–Longmu Co fault system, similar to the intersection of the San Andreas and Garlock faults. Yet, in this case, slip on the Gozha–Longmu Co and bending of the Karakoram fault appear to be genetically related to one another in a way that is similar to the intersection of the Red River and Xianshuihe–Xiaojiang faults. The reasons for such behavior remain to be determined, but it seems likely that an important factor in explaining such differences in deformational behavior at fault intersections stems from variations in crustal structure related to variations in both crustal thickness and thermal gradient. In particular, intersecting faults in which slip along one fault deforms the surface trace of the other, such as the Blackwater-Garlock, Red River–Xianshuihe, and Karakoram–Longmu Co intersections, all occur in locations with thick crust and/or anomalously high thermal gradients and thus may be underlain by weak, subhorizontal detachment horizons within the middle crust (e.g., Burchfiel et al., 1989).

Our results have several broad implications for studies of large intercontinental strike-slip fault zones and how they interact on time scales less than 1 million years. First, our results highlight the importance of accounting for potential strain influences of intersecting faults when studying along-strike variability of deformation along major fault zones. The Longmu Co fault appears to exert a first-order control on the structural style along the Karakoram fault, which may have significant implications for along-strike differences in determinations of the slip rate and total magnitude of displacement along the Karakoram fault. Second, our results suggest that local changes in the geometry and kinematics of a fault system could indicate important changes occurring throughout an entire orogen. This appears to be true to the first order for the late Miocene of the western India-Asia collision zone. Finally, the results demonstrate the applicability of a rigid-block conceptualization of continental crust and kinematic analyses in obtaining a first-order understanding of continental deformation.

This work was supported by National Science Foundation grant EAR-0310415 and the University of California–Davis Durrell Fund. We would like to thank Ryan Gold, Mike Murphy, Alex Robinson, and Eldridge Moores for helpful comments that improved the initial drafts. Rebecca Bendick, Jim Spotila, and an anonymous reviewer provided helpful comments on an earlier version of this work. We would like to thank Jim Spotila and Clark Burchfiel for their comments on the present version of this work. Finally, we would also like to thank Nick Kent-Basham for help during the field work.