The exact location of the northern Karakorum fault (KF) in western Tibet is unclear and its current activity is debated. Here, we investigate the possible northern extension of the KF, the Muji fault, located in the Chinese Pamir, which belongs to the Kongur Shan extensional system, and provide the first quantitative estimate of its Holocene slip-rate. The fault cuts and offsets a series of 6 fluvial terraces, yielding a minimum slip-rate of 4.5 ± 0.2 mm/yr, by matching the largest terrace riser offset with its upper surface age (10Be, n = 24). Field evidences of right-lateral movement along the Kongur Shan fault, as well as geometry and kinematic similarities with the southern half of the KF attest that the Muji fault belongs to the KF system. Therefore, its fast slip-rate combined with the slow slip-rates along minor splays of the northern KF (maybe up to 4 mm/yr) southwest of the Tashkorgan basin agrees with the late Pleistocene southern KF slip-rate (>8 mm/yr).
 The Pamir Mountains, located in Central Asia, lie at the western end of the Indo-Asian collision zone and are bounded by the Tian Shan, Hindu Kush, Tarim basin and Tajik depression to the north, south, east and west, respectively (Figure 1). The north-south trending Chinese Pamir, with two peaks (Kongur Shan, 7719 m, and Mustagh Ata, 7546 m) much higher than the rest of the range, separate the low-elevation Tarim basin (<1200 m) to the east from the high-elevation Tashkorgan and Muji basins (>3000 m) to the west. Mountain building was induced by the late Cenozoic collision between the Pamir salient and the Tian Shan [Burtman and Molnar, 1993], with as much as 300 km of north-south shortening accommodated on the Main Pamir Thrust [e.g.,Arrowsmith and Strecker, 1999; Brunel et al., 1994; Burtman and Molnar, 1993; Strecker et al., 1995], coupled with strike-slip faulting and radial thrusting on the eastern and western flanks of the orogen, respectively [Cowgill, 2010]. The east-west trending Muji range belongs to the Chinese Pamir, and has been rather inaccessible due to its location at high elevation and in the vicinity of the politically unstable borders of Pakistan, Afghanistan, Tajikistan and Kyrgyzstan. To the south, the Kongur Shan extensional system bounds the eastern Tashkorgan basin. The system accommodates a minimum of 34 km of east-west extension in the Pamir along the Kongur Shan massif [Robinson et al., 2004], ∼20 km or less along the Mustagh Ata massif, and <3 km along the Tashkorgan fault to the south [Robinson et al., 2007].
 The Muji fault extends from the China-Tajikistan border and bounds the southern side of the Muji range (Figure 2). Its western section is mostly strike-slip with a slight normal component (presence of triangular facets all along the range) and becomes predominantly normal to the east. At 74°36′E, the Muji basin and fault suddenly veer to the SSE and become the Tashkorgan basin and Kongur Shan detachment fault, respectively (Figure 2).
 Fluvial terraces in glaciated regions provide quantifiable records of climatic and tectonic change, visible today as offset channels or risers. The ability of a stream to entrain sediments and incise its channel is sensitive to climatic and tectonic influences: a reduction of sediment load or an increase in stream power will cause the stream to incise its channel, leaving behind the former stream bed as a terrace [Whipple and Tucker, 1999]. Determining fault slip-rates from offset landforms requires identification of equivalent piercing points on opposite sides of the fault and determination of their age [Gold et al., 2011], in addition to datable material, such as radiocarbon or quartzite/granite cobbles (used in 10Be dating). Fluvial terrace risers, which are bounded by upper and lower terrace surfaces, provide a particularly useful tool for reconstructing lateral slip-rates along a strike-slip fault due to their linearity on each side of the fault. For most fluvial systems, it is likely that some lateral refreshment occurs synchronously with riser displacement, thus yielding the age of the offset riser to be intermediate between the upper and lower surface abandonment ages. Therefore, a conservative approach is to date the abandonment age of both surfaces to bracket the age of the riser [e.g.,Cowgill, 2007; Cowgill et al., 2009; Gold et al., 2009; Mériaux et al., 2005], with the upper terrace age providing a maximum age for the riser, which could not have formed prior to incision and abandonment of the upper surface, while the lower terrace age provides a minimum age for the riser because after incision and abandonment of the lower terrace, its riser cannot have been modified by lateral erosion.
 We first explored the Muji basin on Google Earth Quickbird imagery and targeted the Muji terraces site, which seemed ideal to constrain the right-lateral slip-rate along the Muji fault at this location, due to the presence of multiple terraces that have been offset by the fault. We mapped the fault and geomorphic features as well as measured the offsets, both on satellite images and in the field. Precise measurements of riser heights and offsets were determined using total station profiles both parallel and perpendicular to the fault. We sampled quartzite cobbles (3–6 cm in diameter, Figure S1 of theauxiliary material) for cosmogenic 10Be surface-exposure dating, and collected samples from a depth-profile (Figure S2 of theauxiliary material) to constrain inheritance. The samples were processed at Stanford University's cosmogenic facility and run at the Lawrence Livermore National Laboratory Center for Accelerator Mass Spectrometry. Model ages were calculated in Cronus Earth 2.2 (with constant file 2.2.1) [Balco et al., 2008] using the 5 production rate scaling models [Desilets and Zreda, 2003; Desilets et al., 2006; Dunai, 2001; Lal, 1991; Lifton et al., 2005; Stone, 2000]. Below, we present the results using the Lal /Stone time-dependent model for expediency (data and ages calculated with other models are presented inTable S1). No erosion rate was applied, following Kirby's observation of thick weathering patina on the cobble surfaces along the Kongur Shan detachment in addition to the well-preserved vertical scarps due to normal faulting activity in the basin.
3. Site Description and Offsets
 We targeted one fluvial site in the Muji basin (Figure 3), where a flight of 6 terraces (T1 to T6) are right-laterally offset by the Muji fault. The site is located about 30 km NNW of the town of Muji, at 39°14′N–74°15′E and ∼4300 m asl (Figure 2). It lies at the outlet of the largest and longest glacial valley (∼6 km-long with 5 merging glaciers) of the range, coming from the highest peak of the Muji range (Aksay Bax, 6102 m), which is still highly glaciated today. The valley is deeply (>1000 m) entrenched by glacial erosion, depositing boulders and sediments downstream. The glacier terminus is located just 700 m north of the terraces or ∼1 km north of the fault. Intense incision by fluvial/glacial meltwater yields sharp and high terrace risers (up to ∼35 m, measured by total station,Figure 4). Isolated patches of glacial deposits (possibly moraines from Marine Isotope Stage 3 or Last Glacial Maximum, ∼40 and ∼20 ka, respectively) are present as far as 4 km south of the ice margin, as well as on T4S and T2 (pink on Figure 3).
 West of the river-bed T0, the two largest terraces (T4 and T6) are cut and right-laterally offset by the fault. The western part of T6 is partly covered by post-alluvial fan deposits. Several mole tracks (about 50 × 10 m-large and 1–2 m-high above the terrace surface) are clearly visible on the satellite images as well as in the field (Figure 3 and Figure S3 of the auxiliary material). They were likely produced by the most recent earthquake, possibly the M = 7.0 and M = 6.0, 1944 Bulenke earthquake (39.1°N–75°E), located ∼65 km east of our site.
 Riser offsets were accurately determined from the Quickbird satellite images and in the field by total station profiles (Figure 4): T6/T4 offset is 41 ± 0.5 m and T4/T0 (north of the fault) or T4/T2 (south of the fault) is 14 ± 1 m, but is considered a minimum since the northern riser is constantly refreshed by the river, while the southern riser is protected from erosion by T2 (Figure S4 of the auxiliary material). In addition to the horizontal offsets, vertical offsets of 1.1 to 1.8 m (on T4) and 4.1 m (on T6) are deduced from the profiles (Figure 4).
4. Surface Ages and Slip-Rates
 We collected 19 quartzite samples for 10Be surface-exposure dating (Figure S1 of theauxiliary material): 9 on T4 north and south of the fault (MJ-10 to MJ-19) and 10 on T6 (MJ-20 to MJ-31) (only south of the fault, to avoid possible contamination by the nearby fan deposit to the west). In addition, to further constrain the age of T4 and constrain possible inheritance, 5 successive levels were sampled from 45 to 200 cm-depth in a depth profile located on the riser between T4S and T2 (n = 5, see location onFigure 3 and data in Table S1 and Figure S2 of the auxiliary material).
 The 6 samples on T4S range from 3.2 to 4.05 ka, with an average of 3.65 ± 0.3 ka while the 3 samples on T4N range from 3.3 to 4 ka, with an average of 3.75 ± 0.35 ka (Table S1). Together, T4's average age is tightly constrained at 3.68 ± 0.3 ka. The 10 samples on T6S range from 8.7 to 9.5 ka and tightly cluster at 9.08 ± 0.34 ka (Table S1). These ages most likely correspond to terrace emplacements during the warm Holocene Climatic Optimum (9–5 ka). While our depth profile data do not allow any interpretation (therefore discussed in the auxiliary material and on Figure S2), data from two depth profiles at the same location [Schoenbohm et al., 2011] constrain the ages of T4 and T6 surfaces at 2.9+1.1/−0.7 ka and 8.5+2.5/−1.3 ka, respectively, using the method of Hidy et al. . These ages are identical, within error, to what we obtain by dating surface samples, confirming negligible erosion and inheritance in the area.
 Bounds on the right-lateral slip-rate along the Muji fault at this site may be derived from the measured offsets and ages, and the errors associated with such rates are derived from the Root Mean Square errors on both offsets and ages. Matching the upper (T6) and lower (T4) surface ages with their riser offset (41 ± 0.5 m) yields a slip-rate of minimum 4.5 ± 0.2 to maximum 11.1 ± 0.9 mm/yr. However, this maximum rate is overestimated due to the presence of a non-dated T5 terrace in between (Figure 3, ∼4 m higher than T4, ∼8 m lower than T6), which is >3.68 ka and could be ∼5.5 ka, assuming a constant incision rate between T6 and T4, therefore yielding a maximum slip-rate of ∼7 mm/yr. The smaller offset can only be matched with the age of its upper terrace (T4) since T2 was not dated. Matching the 14 ± 1 m offset with T4's age (3.68 ± 0.3 ka) yields a minimum slip-rate of 3.8 ± 0.4 mm/yr (underestimated because this offset is a minimum, as discussed above). We are also able to calculate vertical rates of 0.45 mm/yr (4.1 m of vertical offset and 9.08 ka for T6,Figure 4) or 0.3–0.49 mm/yr (1.1 to 1.8 m of vertical offsets and 3.68 ka for T4, Figure 4), bringing the Muji fault slip-rate to >5 mm/yr. This rate is consistent with GPS data [∼4–5 mm/yr,Zubovich et al., 2010] and with the long-term geologic rate (∼4.5–5.5 mm/yr) [Robinson et al., 2010] across the northern Kongur Shan normal fault.
 We speculate that, during the Holocene Climatic Optimum, the moraines were breached and rapid emplacement of fluvial terraces occurred due to the strong release of water. One attempt to check for fast terrace emplacement is to calculate an incision rate at this particular site. Matching the total T4/T0 riser height, ∼35 m, with its abandonment age of 3.68 ka (T4), yields a rapid incision rate of ∼9 mm/yr. Rapid fluvial transport is consistent with the short distance between the glacier terminus and the sampled surfaces and implies extremely rapid river catchment evolution and interaction between river dynamics, tectonics and climate in the Pamir.
 While current active tectonics along the southern half of the ∼1000-km-long right-lateral Karakorum fault (KF), which extends from the eastern Pamir to southwestern Tibet (Kailas), is well-established [e.g.,Banerjee and Burgmann, 2002; Brown et al., 2002; Chevalier et al., 2005a; Jade et al., 2004, 2010; Lacassin et al., 2004; Valli et al., 2007, 2008], its trace north of the Karakorum range is unclear and has recently been interpreted to be inactive since the Pliocene [Robinson, 2009a]. This has important implications for whether the KF is accommodating the northward convergence of India to Eurasia, as well as whether it is helping the eastward extrusion of the Tibetan crust and facilitating the northward movement of the Pamir salient relative to Tibet. According to Chevalier et al. [2005a, 2011], the late Quaternary slip-rate along the southern half of the KF is at least 5.5 mm/yr at Gar [Chevalier et al., 2005a] to at least 7–8 mm/yr in the Kailas [Chevalier et al., 2011], along one or two branches of the fault, respectively, and over a timescale of 10–200 ka. To the northwest, the fault becomes hard to follow, because it is mostly covered by glaciers of the Karakorum range. Some studies [e.g., Brunel et al., 1994; Burtman and Molnar, 1993; Strecker et al., 1995] interpret the northern end of the KF to become the East Pamir fault which then connects with thrust faults of the Rushan Pshart zone in the central Pamir. Just north of the Karakorum range and south of the Tashkorgan basin, the fault seems to split into two main strands [Robinson, 2009b; Robinson et al., 2007], as well as into several other smaller faults along which the dextral motion is partly transferred to, such as the NW striking right-lateral (transpressional) Aksu-Murgab, Karasu [Strecker et al., 1995], East Pamir and Aksu-Rangkul faults (the latter is geomorphologically less expressed) [Strecker et al., 1995] (Figure 1). However, the magnitude of right-slip on these faults is not well-determined [Burtman and Molnar, 1993].
Burtman and Molnar suggested that the accommodated displacement is relatively large (60 km) along the Aksu-Murgab fault and little (20 km) on the other splays, whileRobinson et al.  did not observe any significant displacement along the former on satellite images. Although these faults are poorly studied and constrained due to their remote location, Strecker et al. calculated a minimum 0.3 mm/yr and maximum of 0.8 mm/yr slip-rate along the right-lateral Karasu fault. This rate is based on the assumption that a 6 m offset of fluvio-glacial terraces is post Last Glacial Maximum (∼20 ka) and on carbonate ages (>350 ka) on the oldest fluvio-glacial terraces (135 m offset), as well as on solifluction that covers the fault scarp (possibly associated to the humid Holocene interval). If their younger terrace has the same age as our terraces at Muji (∼9 and ∼4 ka), the slip-rate along the Karasu fault becomes at least twice as fast, i.e. 0.6–1.6 mm/yr.
 As Strecker et al. pointed out, the slip-rate for the entire region is higher when slip distribution on several dextral faults is considered. Therefore, if each splay moves at ∼1 mm/yr (similar to the Karasu fault), the northern KF zone could move at a rate of ∼4 mm/yr, which is about half of what has been determined along the southern end of the fault [Chevalier et al., 2005a, 2011] across two branches.
 Most studies [e.g., Arnaud et al., 1993; Brunel et al., 1994; Liu, 1993; Murphy et al., 2000; Peulvast, 1992; Ratschbacher et al., 1994; Searle, 1996; Strecker et al., 1995; Tapponnier and Molnar, 1979] state that the Muji-Tashkorgan basin is a graben at the northern end of the right-lateral strike-slip KF, where a series of en-echelon rifts occur. The Kongur Shan and Mustagh Ata, located on the eastern side of the basin, are interpreted to be topographic anomalies at the northern tip of the KF, and correspond to the anomaly of the Gurla Mandhata (7728 m), which occupies the eastern side of the Pulan graben at the southern tip of the KF [Arnaud et al., 1993]. There, the Humla-Gurla Mandhata fault system [Murphy and Copeland, 2005; Styron et al., 2011] displays the same geometry, i.e. a segment of almost purely normal component (Gurla Mandhata detachment fault vs.Kongur Shan detachment fault) between two mainly strike-slip segments with a normal component (KF and Humla faultsvs.KF and Muji faults). Another example of such fault geometry is the Ashkule-Gozha fault system (western Altyn Tagh fault) where a segment of almost purely normal component (where the 2008 Ashkule earthquake, Mw7.2, occurred) [Li et al., 2011] is located between the left-lateral strike-slip Ashkule and Gozha faults. In addition,Liu found evidence for right-lateral displacement in addition to the main normal component (Figures S5 and S6 of theauxiliary material) just north of the Tahman fault (Figure 1) along the Kongur Shan detachment fault. He measured right-lateral offsets up to ∼44 m, created by several large earthquakes, including the M = 7.5, 1895 Tashkorgan earthquake.
 A way to reconcile the slip-rates along the northern and southern tips of the KF is to assume that it has stepped to the Muji fault (large-scale en echelon strands of right-lateral movement), which presents geometry and kinematic similarities with the southern half of the fault [e.g.,Chevalier et al., 2005a, 2011; Lacassin et al., 2004; Murphy et al., 2000, 2002], i.e. mainly strike-slip with a significant normal component. If this is the case (see supporting studies above), adding our Muji fault slip-rate (>5 mm/yr) to the ∼4 mm/yr rate inferred along the NW minor splays of the KF zone to the west, yields a total slip-rate of >9 mm/yr for the northernmost segment of the KF. This rate is similar to whatRobinson [2009b] suggested for the same region, on longer timescales, ∼7–11 mm/yr, by correlating the Aghil formation on each side of the fault. This rate is also similar to what we find along the southern segment of the KF in southern Tibet [Chevalier et al., 2005a, 2011].
 While our study is preliminary and more offset geomorphic features at different locations and on different timescales along the Muji and Tashkorgan basins are warranted, our well-clustered ages on two main offset terraces nevertheless yield a robust first range of slip-rate for the Muji fault between 4.5 ± 0.2 and 11.1 ± 0.9 mm/yr or most probably a maximum of ∼7 mm/yr, to which a ∼0.3–0.5 mm/yr of vertical movement should be added. Field and satellite image evidences of right-lateral movement along the Kongur Shan detachment fault, as well as geometry and kinematic similarities with the southern tip of the KF attest that the Muji fault belongs to the KF system, with minor splays towards the Central Pamir to the NW (south of the Tashkorgan basin) maybe contributing to as much as ∼4 mm/yr of movement altogether, and large-scale en echelon strands of right-lateral movement to the north. Therefore, the minimum slip-rate along the northernmost segment of the KF system might be as fast as ∼9 mm/yr, which is similar to whatRobinson [2009b] finds in the same region, as well as what we find along the southern third of the KF (>5–11 mm/yr at Gar) [Chevalier et al., 2005a], to >8–11 mm/yr at Kailas [Chevalier et al., 2011].
 This project was conducted under the auspices of the China Geological Survey (projects 1212010918036 and 1212011121267) and the Basic Outlay of Scientific Research Work from the Ministry of Science and Technology (J1126). We thank George Hilley for letting us use his cosmogenic facility at Stanford University. Thank you to W. Bohon and A. Robinson for their reviews, as well as Editor R. Harris for her help.
 The Editor thanks Alexander Robinson and an anonymous reviewer for their assistance in evaluating this paper.