No vertical axis rotations during Neogene transpressional orogeny in the NE Gobi Altai: coinciding Mongolian and Eurasian early Cretaceous apparent polar wander paths

In this paper, we test the role of vertical axis rotations during transpressional mountain building. To this end, we carried out a palaeomagnetic study in the NE Gobi Altai of southern Mongolia, sampling widely exposed lower Cretaceous lavas allowing comparison of rotation histories of the Ih Bogd, Baga Bogd and Artz Bogd restraining bends at the eastern termination of the Bogd strike-slip zone. We provide new 40 Ar/ 39 Ar ages to show that the stratigraphy of maﬁc lavas and ﬂuvio-lacustrine sediments on the southern ﬂanks of Mt Ih Bogd and Mt Baga Bogd have ages between ∼ 125 and ∼ 122 Ma, and a maﬁc sill that intrudes the sequence has an age of 118.2 ± 0.8 Ma. The lavas are older than previously dated lavas south of Artz Bogd, with ages of 119–115 Ma. Palaeomagnetic results from the 119–115 Ma lavas south of Artz Bogd show a signiﬁcant steeper inclination than both results from 125 to 122 Ma lavas of Baga Bogd and Ih Bogd, as well as from newly sampled and previously published younger lavas and necks of the 107–92 Ma Tsost Magmatic Field and Shovon and Khurmen Uul basalts. We explain this result by insufﬁcient averaging of secular variation and small errors induced by overcorrection of bedding tilt. We show that individual lavas in the SE Artz Bogd locality represent individual spot readings of the Earth’s magnetic ﬁeld and integrate all results obtained from lower Cretaceous lavas in the Gobi Altai. We present a pole, or rather, an apparent polar wander path without signiﬁcant plate motion from the reference positions of Eurasia, from ∼ 125 to 95 Ma, with n = 126, λ = 80.8 ◦ , φ = 158.4 ◦ , κ = 25.3, A95 = 2.5, palaeolatitude = 48.2 with a scatter S λ = 16.7 (Sl = 15.3, Su = 17.8) and a regionally consistent direction for the Gobi Altai of D / I = 11.1 ◦ /65.9 ◦ , (cid:5) D / (cid:5) I = 3.8 ◦ /1.9 ◦ . This is one of the best-determined palaeopoles/APWP’s for Asia. Formation of the Ih Bogd, Baga Bogd and Artz Bogd restraining bends was thus not associated with vertical axis rotations larger than our error margin of ∼ 10 ◦ . From this we conclude that the Bogd strike-slip zone is a weak fault zone, in which shear is localized.

deformation in central Asia, which extended progressively northward through Tibet and the Tien Shan into Mongolia (Tapponnier & Molnar 1979;Cobbold & Davy 1988). In western and Central Mongolia, two transpressional orogenic belts formed around a passive indentor formed by the Hangay dome (Cunningham 1998(Cunningham , 2005Fig. 1): the right-lateral Mongolian Altai in the west (Cunningham et al. 1996b(Cunningham et al. , 2003a, which formed since the Oligocene (Howard et al. 2003(Howard et al. , 2006 and the left-lateral Gobi Altai in the south (Cunningham et al. 1996a(Cunningham et al. , 1997, which formed since the late Miocene (∼8 Ma: Vassallo et al. 2007). In both ranges, deformation continues today (Baljinnyam et al. 1993;Ritz et al. 1995;Bayarsayhan et al. 1996;Carretier et al. 1998Carretier et al. , 2002Philip & Ritz 1999;Ritz et al. 2003;Vassallo et al. 2005). Both belts contain linked strikeslip faults, thrust ridges and restraining bends, the latter leading to lenticular, well-exposed mountain ranges with elevations up to 4 km, as much as 2.5 km higher than the surrounding basins.
The origin of intracontinental transpressional mountain ranges is debated: if there is a compressional component across the master strike-slip fault system, it may lead to localized thrusting and vertical axis rotation of the fault, such as proposed in the concept of discrete block rotations within a distributed fault system (Garfunkel 1988;Nur et al. 1988;Taylor et al. 1998;Bayasgalan et al. 1999;Cunningham 2005). Bayasgalan et al. (1999) proposed that restraining bend formation at the northeastern termination of the Gobi Altai resulted from clockwise rotation of 'slats' bound by parallel left-lateral strike slip faults, comparable to scenarios in-voked for Aegean region fault rotations (Taymaz et al. 1991). At present, the angle between the transpressional segments bounding the restraining bends and purely strike-slip fault segments in between restraining bends in the NE Gobi Altai is approximately 20-30 • (Fig. 1). Alternatively, strike-slip segments may initially form in en echelon pattern and across contractional segments. Both the strike-slip segments and the contractional segments may form due to reactivation of favourably oriented basement faults and structural 'grain'. In this paper, we test by means of palaeomagnetic analysis whether transpressional mountain building in the NE Gobi Altai mountain range of southern Mongolia was associated with vertical axis rotations.
To this end, we targeted the youngest stratigraphic interval with widespread exposure across the NE Gobi Altai region, with suitable lithologies for palaeomagnetic analysis that predates the late Miocene (∼8 Ma: Vassallo et al. 2007) onset of uplift of the restraining bends of Ih Bogd, Baga Bogd and Artz Bogd (Vassallo et al. 2007; Fig. 1). The Phanerozoic geological history of southern Mongolia preceding the formation of the modern transpressional mountain ranges consists of multiple tectonic events broadly reducible to three major phases: (1) Mainly Palaeozoic accretion of terranes of continental and oceanic affinity (Sengör et al. 1993;Buchan et al. 2001Buchan et al. , 2002Badarch et al. 2002;Dijkstra et al. 2006;Helo et al. 2006;Windley et al. 2007) ending in the late Jurassic with the closure of the last intra-Asian 'Mongol-Okhotsk' ocean between Mongolia and Siberia, the last intra-Asian seaway within the Altaids (Zhao et al. 1990;Enkin et al. 1992;van der Voo et al. 1999;Zorin 1999;Cogné et al. 2005;Windley et al. 2007). Terrane accretion caused a roughly E-W trending basement grain in the Gobi Altai region; (2) Mesozoic rifting and widespread basin formation, clastic sedimentation and volcanism, notably in the late Jurassic and early Cretaceous (Berkey & Morris 1927;Graham et al. 1993;Traynor & Sladen 1995;Hendrix et al. 1996;Meng et al. 2003;Johnson 2004;Wang et al. 2006) and (3) scattered occurrences of Cenozoic basaltic magmatism and local sedimentary basin formation (Whitford-Stark 1987;Devyatkin & Smelov 1980;Höck et al. 1997; Barry & Kent 1998;Cunningham 2001; Barry et al. 2003). In the NE Gobi Altai region, all elements of this general history are found in the rock record. The restraining bends of Ih Bogd, Baga Bogd and Artz Bogd expose the pre-Mesozoic basement in their cores and are flanked by lower Cretaceous alternating lavas and fluvio-lacustrine sediments, which reach an exposed thickness up to 1500 m or more, notably along their southeastern sides (Fig. 1). Furthermore, scattered occurrences of shallow intrusive basalts, basalt plateaux and younger lavas have been reported (Devyatkin & Smelov 1980;Barry & Kent 1998;Barry et al. 2003;Hankard et al. 2007a). However, the occurrences of the Cenozoic volcanic rocks are too few and scattered for regional correlation and assessment of vertical axis rotations. Therefore, we targeted the lower Cretaceous stratigraphy, which consists dominantly of basalts for our palaeomagnetic and geochronologic investigations.
Recently, Cogné et al. (2005) concluded that the closure of the Mongol-Okhotsk ocean occurred in the late Jurassic, a finding that was confirmed by Hankard et al. (2007b). Therefore, we tested for regional vertical axis rotations by comparing our palaeomagnetic results with apparent pole positions of stable Eurasia in the early Cretaceous published by Besse & Courtillot (2002), Schettino & Scotese (2005) and Torsvik et al. (in press).

G E O L O G Y A N D S A M P L I N G
We collected 871 samples from 123 lava and magmatic neck sites, with typically seven samples per site (Fig. 1). The southeastern parts of these three mountain ranges expose the Mesozoic stratigraphy that was the target of this study.
The Mesozoic deposits comprise some lavas sampled near Dulaan Bogd (DR-Jurassic) predating the early Cretaceous, and which are estimated as Jurassic. Along their southern edges, the basement exposed in the Ih Bogd, Baga Bogd and Artz Bogd restraining bends is overlain by lavas and fluvio-lacustrine deposits of early Cretaceous age. Palaeoflow directions in fluvial sediments, as well as the lateral disappearance of lavas to the east and west suggest a half-graben geometry with north-facing normal faults and southward palaeoslopes. This has an important implication: lavas may have had an initial non-tectonic tilt of a few degrees to the south. However, tilting of lacustrine mudrocks in between lavas, as well as observed tilted Tertiary peneplains suggest that at least part of the generally 10-15 • of southward tilt, which characterizes the lava successions is of tectonic, post-depositional origin. We thus chose to correct the palaeomagnetic directions obtained in this study for bedding tilt and test the validity of this with fold tests, although we note that this may lead to a structural overcorrection of a few degree to the north if the successions were not deposited purely horizontally.
The southeastern part of Ih Bogd exposes a southeastward tilted and locally tectonically folded sequence of lower Cretaceous lavas and sediments, continuously exposed over 20 km, from Dulaan Bogd (DR-Cretaceous) to Jaran Bogd (JI), sometimes also transcribed  Figure 2. Logs of the Jaran Bogd and Bulgantiin Uul sections with palaeomagnetic and stratigraphic sampling levels.
as Jiran or Dziran Bogd (Fig. 1). Chilled margins at the base of the lavas, pahoehoe flow structures, local pillow structures, and upward increasing vesicularity leading to distinct erosion profiles show that the basalts in this section are largely extrusive, although a few bedding-parallel sills are also suspected. From the Jaran Bogd section, we collected three samples JB26, JB57 and JB94 for 40 Ar/ 39 Ar dating (Fig. 2). Alluvial and fluvial sediments and lavas of late and post-Cretaceous age unconformably overlie the sequence. The region east of Baga Bogd contains a stratigraphic succession of Cretaceous lavas and sediments, which were sampled at four locations (BA, BL, BN, KK). We collected two samples from the Bulgantiin Uul (BL) section-samples BU23 and BU56-for 40 Ar/ 39 Ar geochronology (Fig. 2). In between the eastern end of Baga Bogd and Artz Bogd to the south, we collected samples from seven steeply dipping lavas in the footwall of a NE-verging thrust at Unegt Khairkhan (UK; Fig. 1).
South of Artz Bogd, lower Cretaceous lavas, alternating with light blue and white lacustrine clays and redbeds are exposed. These sequences are intruded by the Tsost Magmatic Field basaltic necks (Whitford-Stark 1987;Devyatkin & Smelov 1980;Barry 1999;Hankard et al. 2007b). Barry (1999) reported 40 Ar/ 39 Ar ages of the lavas ranging from 119 to 115 Ma, and ages from the intrusions in the Tsost Magmatic Field ranging from 107 to 96 Ma. We sampled the lavas in the continuous 119-117 Ma Tsagaan Tsav section (TT) of Barry (1999) and in its lateral equivalent Kharaat Uul (KU), as well as near the top of the stratigraphy, at Khatavch (115 Ma: Barry 1999). West of these sections, we sampled 10 sites in the Tsost Magmatic Field. Recently, Hankard et al. (2007b) also reported palaeomagnetic results from the Tsost Magmatic Field. They described their sampling locations as both necks and flows and reported K/Ar ages of ∼104-94 Ma, with one outlier at 118 Ma. These ages suggest that the majority of their sites were taken from the basalt intrusions of 107-96 Ma (Barry 1999) or 104-94 Ma (Hankard et al. 2007b), with the 118 Ma age probably obtained from the underlying lavas equivalent to the series east of the Tsost Magmatic Field with ages of 119-115 Ma (Barry 1999 ; Fig. 1).

G E O C H RO N O L O G Y
Six basalts were sampled for 40 Ar/ 39 Ar dating (Fig. 1). Microscopic inspection showed no or hardly any signs of alteration. Groundmass was separated using standard mineral separation techniques. All samples were leached with 1 N HNO 3 during 1 hour in an ultrasonic bath. Circa 50 mg of material form the size fraction 250-500 μm of the groundmass of each sample was wrapped in Al-foil and loaded in a 15 mm ID quartz vial. The in-house Drachenfels sanidine standard (25.26 ± 0.03 Ma; modified from Wijbrans et al. (1995) was used as neutron fluence monitor and loaded at top and bottom positions between each set of four unknowns. Samples and standards were irradiated for 18 hr in the Cd-lined RODEO P3 position of the HFR Petten, the Netherlands. 40 Ar/ 39 Ar incremental heating experiments were performed at the Vrije Universiteit Amsterdam. Standards were fused with a Synrad 48-5 50W CO 2 laser. Samples load in a sample tray with 6 mm diameter holes in which the sample was spread out evenly and were incrementally heated using a Raylase scan head as a beam delivery and beam diffuser system. After purification the gas was analysed with a Mass Analyzer Products LTD 215-50 noble gas mass spectrometer. Beam intensities were measured in a peak-jumping mode in 0.5 mass intervals over the mass range 40-35.5 on a Balzers 217 secondary electron multiplier. System blanks were measured every four steps. Mass discrimination was monitored by frequent analysis of aliquots of air. The irradiation parameter J for each unknown was determined by interpolation using a second-order polynomial fitting between the individually measured standards. Ages are calculated using the inhouse developed ArArCalc software (Koppers 2002). All 40 Ar/ 39 Ar ages are calculated with Steiger & Jäger (1977) decay constants at the 2 level and include the analytical error and error in J . Corrections factors for neutron interference reactions are 2.7 ± 0.03 × 10 −4 for ( 36 Ar/ 37 Ar) Ca , 6.99 ± 0.13 × 10 −4 for ( 39 Ar/ 37 Ar) Ca and 1.83 ± 0.2 × 10 −2 for ( 40 Ar/ 39 Ar) K . Table 1 summarizes the 40 Ar/ 39 Ar results. The full analytical data are given in the supplementary Table S1. The general criterion for definition of a reliable plateau age is that the plateau includes more than three concordant incremental heating steps representing more than 50 per cent of total 39 Ar K released (e.g. Fleck et al. 1977). The MSWD over the plateau steps was calculated as an additional test for plateau homogeneity. Acceptable plateaus yield MSWD values of less the 2.0. Five samples yield excellent age plateaus according to these criteria and do not show indications of thermal disturbances. One samples yields no reliable age spectra. Its total fusion age of 166 Ma (JB46) does not have any geologically meaning and this age is not included in further analysis. The 40 Ar/ 36 Ar intercepts of the inverse isochrons do not deviate from the atmospheric 40 Ar/ 36 Ar ratio of 295.5 at the 2σ level and show no indications of inherited argon components (Table 1 and Fig. 3).". The radiogenic 40 Ar contents are generally high with all plateau steps >80 per cent 40 Ar * , Notes: MSWD is mean square weighted deviates, N is the number of steps included (excluded) in the plateau age. 39 Ar K (per cent) is the percentage of 39 Ar K released by plateau steps. 40 Ar * is the radiogenic amount of 40 Ar. Total fusion ages and inverse isochron intercepts are given for reference. Errors are given at 95 per cent confidence level.   but most even >90 per cent 40 Ar * . Therefore, we conclude that alteration, if present has been successfully removed by leaching and interpret the weighted plateau ages as the best age estimates for the 5 basalts (Table 1, Fig. 3).

PA L A E O M A G N E T I C P RO C E D U R E S
Samples were collected with a water-cooled gasoline powered motor drill and cut in pieces of an inch across and length in the laboratory. The orientation of all samples was measured with both a magnetic and a sun compasses. To exclude any local magnetic deviations, all palaeomagnetic interpretations are based on the solar azimuths. The samples were demagnetized using either thermal (TH) or alternating field (AF) progressive demagnetization. Heating took place in a magnetically shielded, laboratory-built furnace using small temperature increments of 20-80 • C up to temperatures of 640 • C. The AF demagnetization was carried out with 5-20 mT increments up to 120 mT. For all samples, the natural remanent magnetization (NRM) of the specimens was measured on a 2G Enterprises horizontal DC SQUID cryogenic magnetometer (noise level 3 × 10 −12 Am 2 ). For AF demagnetization, the instrument was interfaced with an in-home developed robot-assisted automated measuring device.
To identify the magnetic carriers and their characteristics, acquisition of isothermal remanent magnetization (IRM) was carried out (Fig. 4). Samples were carefully selected based on the demagnetization diagrams. They served as being representative for lightning struck samples (JP60), samples that were entirely demagnetized at ±580 • C (BP66) and at ±560 • C (JI292). Additionally, hysteresis loops were determined (Fig. 4). The plots of M r /M s against H hc /H c in a Day-plot (Day et al. 1977), indicates the presence of pseudo-single-domain grains in all three analysed samples; (Fig. 5). Curie temperatures were determined using a modified horizontal translation type Curie balance that applies a cycling field (Mullender et al. 1993). Stepwise increasing temperatures of 450, 550, 600, 650 and 700 • C were used revealing blocking temperatures of 560-580 • C (Fig. 4). These Curie temperatures, together with the hysteresis loops closing at or below ∼0.3 Tesla, indicate (Ti-poor) magnetite to be the main carrier of the natural remanent magnetization (NRM). Finally, these rock magnetic experiments give us confidence that we measured a stable NRM. The demagnetization diagrams of the NRM were plotted in orthogonal vector diagrams (Zijderveld 1967). In addition, a number of multicomponent samples were plotted on equal-area projections. Representative examples are given in Fig. 6. Identification of the characteristic remanent magnetization (ChRM) was done by principal component analysis (Kirschvink 1980). When vector end points showed a trend towards the origin of the diagram, we determined this component to be the ChRM.
Initially, we demagnetized one sample of every site by thermal demagnetization. The remaining samples of the sites were either demagnetized on a robotized alternating field demagnetizer or demagnetized thermally. Initial intensities range typically from 0.5 to 2.0 A m −1 . The AF demagnetized samples show good agreement with the TH samples of the same site (Figs 6c and d). At a number of sites, the ChRM was not recognized in the TH demagnetization but could clearly be determined by AF demagnetization (Figs 6k and l).

ANALYSIS
Samples were rejected from further analysis if the maximum angular deviation exceeded 15 • . Because lava sites represent spot-readings of the Earth's magnetic fields, within-site errors can be assumed to be random, and averages and cones of confidence on lava site level were determined by Fisher statistics (Fisher 1953). The resulting site means and statistical values are given in Table 2. We excluded lava sites with k < 50 from further analysis, since variation between samples within a single lava should be minimal. Remaining lava sites were screened for remagnetization (e.g. caused by lightning). Complete remagnetization of sites was recognized in 15 cases. The declinations of the lowermost sites of section JI show remarkable deviations from north after tectonic correction. These sites show consistent reversed directions before tectonic correction and we interpret this as a post-folding obtained NRM, which is confirmed by the clearly negative foldtest of Tauxe & Watson (1994;Fig. 7a). The results of this foldtest indicate that the best fit is found much below 0 per cent untilting. This may be the result of various degrees of overprinting or uncertainties in bedding attitudes. These sites were excluded from further analyses. Nine sites show complete remagnetization caused by lightning. The NRM of rocks struck by lightning can be entirely or partly overprinted (Hallimond & Herroun 1933).
The identification of lightning-induced NRM is based on abnormal high magnetic intensity, together with within-site randomness of NRM directions (Fig. 6i). In some cases, an overprint caused by lightning or gyro-remanent remagnetization (Dankers & Zijderveld 1981) could be identified and eliminated. This was done using the great-circle approach of (McFadden & McElhinny 1988;Fig. 6).
Because scatter of palaeomagnetic directions induced by secular variation of the Earth's magnetic field is near-Fisherian at the poles, but ellipsoid elsewhere on the globe (Cox 1969;Tauxe & Kent 2004), we calculated our lava site means into poles and determined locality means on the poles using Fisher-statistics (Fisher 1953). Lava sites were excluded from determination of locality means when their poles exceeded the maximum angular standard deviation of Vandamme (1994; Table 2). The average pole was then calculated into an average declination and inclination, with an errorbar D and I following equations A.60 and A.57 of Butler (1992), respectively ( Table 2). The resulting average directions per sampling locality are shown in Figs 8 and 9.
Although the lava site averages of locality Dulaan Bogd Jurassic are very accurate, site averages are scattered and do not pass the reversal test of McFadden & McElhinny (1990). The resulting k-value of 2.7 indicates that there is no sensible palaeomagnetic direction obtained from these 10 lavas and they are excluded from further analyses. The average direction obtained from the lower Cretaceous lavas at Unegt Khairkhan suggest very large rotations which can by no means be confirmed through structural geological observations, and we interpret these results (based on only three of the seven lavas) as bearing of no regional importance. These directions are excluded from further analyses.
In the following analysis, we integrate our results from lower Cretaceous with the recently published results of Hankard et al. (2005Hankard et al. ( , 2007b, and recalculated the average pole since our cut-off of k = 50 for individual lava sites eliminates a few of their sites, and mainly because we use a fundamentally different statistical approach by calculating averages on poles and calculating D and I, instead of determining a dp/dm error ellipsoid around the (Fisherian) average pole. Based on age and location, we identify five localities: SE Ih Bogd (n = 21, 124.7-118.2 Ma), SE Baga Bogd (n = 29, 124.3-122.7), SE Artz Bogd (n = 24, 119.3-115.4 Ma; Barry 1999), the Tsost magmatic field, from which we obtained seven reliable site means, and which is combined with the recently published 'Artz Bogd' locality of Hankard et al. (2007b) (n = 25, 107-94.7 Ma) and finally the combined Shovon/Khurmen Uul locality of Hankard et al. (2005Hankard et al. ( , 2007b; n = 20, 94.7-92 Ma). Within all three localities SE Ih Bogd, SE Baga Bogd and SE Artz Bogd we have sampled limbs of steep folds and the resulting positive fold tests (Fig. 7) show that folding and tilting (largely) post-dates deposition and acquisition of the NRM.
Testing of common true mean directions (McFadden & Lowes 1981) shows that there is neither significant difference at the 95 per cent confidence level between the SE Ih Bogd and SE Baga Bogd localities, nor between the Tsost Magmatic Field and the Shovon/Khurmen Uul localities (Table 3). Therefore, we have recalculated the palaeopoles and average directions and combined the above localities (Table 4). However, locality SE Artz Bogd does differ from the older and younger combined localities of Ih Bogd/Baga Bogd and Tsost/Shovon/Khurmen Uul (Table 3). The combined localities of Ih Bogd/Baga Bogd and Tsost/Shovon/Khurmen Uul, however, do not significantly differ at the 95 per cent confidence level. In Table 4, we have combine all our successful data to end up with a combined early Cretaceous pole of the Gobi Altai (n = 126, ∼125-94 Ma).   To determine whether post-Cretaceous rotations have occurred in the Gobi Altai, we compare our directions with palaeomagnetic poles constructed for stable Eurasia. In recent years, three apparent polar wander paths (APWP's) have been reconstructed for Eurasia by Besse & Courtillot (2002), Schettino & Scotese (2005) and Torsvik et al. (in press). We compare our results to the published poles and directions predicted for the Gobi Altai from these three reference models at 90, 100, 110, 120 and 130 Ma (Table 5). All localities in the Gobi Altai except for SE Artz Bogd share a common true mean direction with their relevant reference poles (  Figure 7. Results of the fold-test of Tauxe & Watson (1994). Equal area projections show site means prior to and after tectonic tilt correction of (a) the lower seven sites of the Jaran Bogd section (negative); (b) the remaining 21 sites of the Jaran Bogd section that pass our selection criteria (see Table 2 When comparing the declinations and inclinations of our localities with those expected based on the reference paths (Table 7), it is clear again that the SE Artz Bogd locality has a deviating inclination at the 95 per cent confidence level, which is approximately 9 ± 5 • steeper than expected for Eurasia. However, the declinations of all localities do not significantly differ from the reference directions. Our analysis, therefore, clearly shows that no vertical axis rotations of more than the confidence limit (∼10 • ) have occurred since the early Cretaceous (Table 7).

Geochronology and stratigraphy
Our newly obtained 40 Ar/ 39 Ar results for the Bulgantiin Uul section give an age of 124.3 ± 0.9 Ma for sample BU23, which is located approximately 300 m stratigraphically below sample BU56 with an age of 122.7 ± 0.8 Ma (Fig. 2). If sedimentation rates were constant, then the base of the section, which is laterally equivalent to the Baga Bogd locality, would be approximately 126 Ma old. Locality           Table 5) are indicated as dotted orange line.
Khalzan Khairkhan occurs at the top of the stratigraphic section and its age is probably about 120 Ma or slightly older. A comparable time-span may also represent the Tsagaan Tsav and Khatavch sections where approximately 750 m of lavas and sediments were deposited between 119.3 ± 0.4 and 115.4 ± 0.4 Ma (Barry 1999). In the Jaran Bogd section, sample JB57 with an age of 124.7 ± 0.8 Ma was taken approximately 400 m below sample JB94 (121.9 ± 0.8 Ma), which was collected near the top of the section. Again, comparable sedimentation rates are calculated. However, sample JB26 was sampled approximately 400 m stratigraphically below JB57, yet gives an age of 118.2 ± 0.8 Ma, significantly younger than the higher samples. We therefore, conclude that this sample was obtained from a younger sill. However, because since the majority of the samples we collected were obtained from basalts with clear extrusive characteristics such as upward increasing vesicularity, pahoehoe structures or occasionally pillows, the palaeomagnetic results obtained from this section are representative for the ∼125-122 interval or slightly older. Assuming a constant sedimentation rate, the base of the section would be approximately 129 Ma old.

Anomalous inclination at Artz Bogd
As clearly shown in our statistical analyses, locality Artz Bogd yields a significantly steeper inclination than both the older and younger localities, as well as the three reference paths (Tables 6 and 7). This may be explained by (1) insufficient averaging of secular variation, by chance leading to an anomalously steep inclination, (2) overcorrection of bedding tilt or (3) palaeolatitudinal drift.
To test whether secular variation is averaged out, we compare the scatter of directions in our localities with the expected scatter at the palaeolatitude of Mongolia in the interval of 110-80 Ma of McFadden et al. (1991; Table 8). The scatter in the combined Ih Bogd/Baga Bogd locality, the combined Tsost/Shovon/Khurmen Uul locality and the combined Cretaceous data set do not significantly differ from the expected scatter is Sλ = 16.9 (Sl = 14.9, Su = 19.5), whereas the scatter in the SE Artz Bogd locality is significantly lower: Sλ = 14.0 (Sl = 11.5, Su = 16.0). Even though 24 lavas were averaged, the low scatter shows that secular variation is not averaged out. This may be caused by a very rapid outflow of lavas, such that a series of consecutive lavas represent the same spot reading of the Earth's magnetic field influencing the average. We tested this for the seven lavas of Khatavch, which together yield a κ-value of 241, much higher than expected in a secular variation scatter (which has a k-value of typically 20-50 for directions (Tauxe & Kent 2004), which corresponds to a κ of 10-30 for the poles). However, none of the lavas in Khatavch share a common true mean direction with the next or previous one, showing that each direction obtained from this section is a separate spot reading, which by chance happen to cluster tightly.
An important point of discussion before one can conclude palaeolatitudinal drift from palaeomagnetic directions of lavas is the issue of tilt correction. As stated earlier, we choose to correct for bedding tilt, because we obtain positive fold tests and the region has tilted Tertiary peneplains and tilted fine-grained sediments, which proves that the majority of the tilt is post-depositional. However, the early Cretaceous basin configuration concluded from sedimentological     Besse & Courtillot (2002), Schettino & Scotese (2005) and Torsvik et al. (in press)  observations indicates that all our samples were collected from northern, southward dipping slopes of the basins, which may, therefore, lead to an overcorrection to the north of a few degrees. Moreover, this overcorrection is not necessarily consistent and the original tilt of the lavas may vary locally due to eruption and deposition over slightly uneven topography. Hankard et al. (2007b) also addressed this issue. In their earlier paper, Hankard et al. (2005) interpreted an anomalously shallow inclination from the six lavas at Khurmen Uul of 10 • and suggested palaeolatitudinal drift as the explanation. In their later paper, Hankard et al. (2007b) argued that this inclination anomaly may have been caused by correction of ∼10 • southward bedding tilt at Khurmen Uul, and chose in their re-analysis not to correct for bedding tilt, which led to no significant deviation from the Besse & Courtillot (2002) reference path of Eurasia. In most cases, one can argue for and against bedding tilt, and in this case, one should allow an extra uncertainty of at least a few degrees, and maybe as much as 10 • for which interpreting the palaeo-inclination. Especially in the case reported here, where the early Cretaceous basin configuration implies a consistent tilt direction, which prevents averaging out of this error. In conclusion, we do not interpret the anomalously steep inclination in terms of palaeolatitudinal drift, because insufficient averaging of the secular variation, and small bedding tilt errors may explain the anomaly. However, because we showed that the lavas do represent individual spot readings of the Earth's magnetic field, we integrate all our results of the lower Cretaceous lavas and construct a pole, or rather, an apparent polar wander path without significant plate motion from ∼125 to 95 Ma, with n = 126, λ = 80.8, φ = 158.4, κ = 25.3, A95 = 2.5, palaeo-latitude = 48.2 with a scatter Sλ = 16.7 (Sl = 15.3, Su = 17.8) and a regionally consistent direction for the Gobi Altai of D/I = 11.1/65.9, D/ I = 3.8/1.9. This is one of the statistically best-determined palaeopoles/APWP's for Asia.

Tectonic implications
In their study of terminations of strike-slip faults, Bayasgalan et al. (1999) suggested a scenario for the formation of the Ih Bogd, Baga Bogd and Artz Bogd restraining bends in which clockwise rotation of the mountain ranges was accommodated by their bounding   left-lateral strike-slip faults (Fig. 10). Also, Cunningham (2005) discussed the possibility of vertical axis rotation of strike-slip faults to create an increasing component of thrusting during progressive transpressional deformation, as is suggested in the northern Altai by a palaeomagnetic study of the Zaisan basin indicating a Neogene 40 • clockwise rotation (Thomas et al. 2002). The angle between the transpressional faults and the strike-slip faults in the NE Gobi Altai at present is 20-30 • , and the absence of rotations larger than our error bar of ∼10 • shows that vertical axis rotations have not played a significant role in the formation of the Ih Bogd, Baga Bogd and Artz Bogd restraining bends in the NE Gobi Altai. On a large scale, Schlupp (1996) suggested a model in which maximum SW-NE shortening at the junction between the Gobi Altai and Altai mountains, decreasing east and northwest, respectively, would lead to opposite rotation of the strike-slip faults of the Gobi Altai and western Altai (Fig. 11). Our error margin of ∼10 • and a E-W width of the Gobi Altai of approximately 600 km allows a shortening of ∼100 km at the junction that would remain undetectable for our palaeomagnetic analysis, which would mean that the suggestion of Schlupp (1996) may be valid for the Gobi Altai region (although we note that the Altai widens to the NW and has accommodated more shortening there, not less and appears, therefore, inconsistent with scenario of Schlupp (1996). However, the presence of segments of pure strike-slip without transpression between the restraining bends in the Gobi Altai indicates that it is unlikely that this scenario is of great importance. The presence of these pure strike-slip segments is not demonstrated in this paper. Indeed they appear to be subordinate and less important than the thrusts in Fig. 1.

10∫ error margin
Hangay Dome W e s t e r n A l t a i Go bi-Al tai ± 60 0k m ± 1 0 0 k m Direction of shortening Figure 11. Schematic representation of the scenario of Schlupp (1996), in which opposite rotations in the Western Altai and Gobi Altai accommodate maximum convergence at their junction and no covergence at their northwestern and eastern limits, respectively. Our 95 per cent confidence limit of ∼10 • allows 100 km of shortening to be undetectable through palaeomagnetic analysis. The presence of pure strike-slip segments in the NE Gobi Altai suggests that this mechanism is too simplistic.  The absence of >10 • vertical axis rotations during transpressional mountain building has important implications for the mechanics and kinematics deformation in the NE Gobi Altai. First, we conclude that the restraining bends are thrust-bound ranges that grew in their present orientation. They may have formed between linked en echelon strike-slip segments, or be the result of strain partitioning into pure strike-slip and compressional segments (Fig. 12). Both cases have been documented in Mongolia and elsewhere (Cunningham et al. 1996a(Cunningham et al. , 2003aCunningham 2005;Mann 2007). Restraining bend border thrusts in the Gobi Altai trend subparallel to the pre-Mesozoic basement grain, thus they probably reactivate pre-existing weak zones such as prevailing metamorphic fabric, older faults and lithological boundaries. Conversely, the strike-slip segments crosscut the pre-existing basement grain and are probably newly formed in the late Cenozoic. We suggest that as restraining bends have grown outwards through time, linking E-W strike-slip faults have either persisted and grown in length, or developed components of oblique-slip displacements at their tips where they link with NW-striking thrusts (and thus the pure strike-slip segments have shortened), or the early formed strike-slip faults have been locally abandoned as new master faults evolve within a strikeslip duplex (although this has not been observed, more structural mapping is needed to test this idea). The implication is that if the strike-slip faults have persisted without rotating, then they must be parallel to the regional slip vector. We note that GPS vectors for the northeast Gobi Altai region are consistent with this interpretation (Calais et al. 2003).
Secondly, absence of vertical axis rotations also shows that drag folding, which would cause counter-clockwise vertical axis rotation adjacent to left-lateral faults, has not played a significant role during motion along the Bogd strike-slip zone, even though we have sampled very close to the main fault zones (e.g. Dulaan Bogd Cretaceous, Bulgantiin Uul, Bulgantiin Uul North and Tsagaan Tsav were sampled within hundreds of metres of the main strikeslip faults). From this we infer that the Bogd strike-slip faults are weak, and that strike-slip displacements are not distributed, but localized along the faults. As a consequence, estimates of cumulative strike-slip displacements along different strike-slip faults is likely a good measure for the entire E-W left-lateral displacement associated with late Cenozoic transpressional orogeny in the NW Gobi Altai.

C O N C L U S I O N S
To test the vertical axis rotation history associated with transpressional mountain building, we have carried out a palaeomagnetic study in the NE Gobi Altai mountains of southern Mongolia. We sampled lower Cretaceous lavas, which are widely exposed and allow comparison of rotation histories of the Ih Bogd, Baga Bogd and Artz Bogd restraining bends at the eastern termination of the Bogd strike-slip zone. Our conclusions can be summarized as follows: 1. We provide new 40 Ar/ 39 Ar ages to show that the stratigraphy of mafic lavas and fluvio-lacustrine sediments on the southern flanks of Mt Ih Bogd and Mt Baga Bogd have ages between ∼125 and ∼122 Ma, with a sill of ∼118 Ma old. The lavas are older than previously dated lavas south of Artz Bogd, with ages of 119-115 Ma.
2. Palaeomagnetic results from the 119 to 115 Ma lavas south of Artz Bogd show a significant steeper inclination than both the 125-122 Ma lavas of Baga Bogd and Ih Bogd, as well as the younger lavas and necks newly sampled and previously published from the 107 to 92 Ma Tsost Magmatic Field and Shovon and Khurmen Uul localities. We explain this result by insufficient averaging of secular variation and small errors induced by overcorrection of bedding tilt. We show that individual lavas in the SE Artz Bogd locality represent individual spot readings of the Earth's magnetic field and integrate all results obtained from lower Cretaceous lavas in the Gobi Altai.
4. There is no significant deviation of the 125-95 Ma pole position of the Gobi Altai from the reference positions of Eurasia.
5. Formation of the Ih Bogd, Baga Bogd and Artz Bogd restraining bends was not associated with vertical axis rotations larger than our error margin of ∼10 • 6. The Bogd strike-slip zone is a weak fault zone, in which shear is localized. 7. Transpressional deformation in the uplifted ranges of the northeastern Gobi Altai could have developed by linked, but partitioned oblique-thrust and strike-slip fault displacements. Both strike-slip and oblique thrust displacements could follow the regional slip vector without requiring vertical axis rotations

S U P P L E M E N T A R Y M A T E R I A L
The following supplementary material is available for this article: Table S1. Full analytical data are given. The isotope ratios are corrected for blanks, mass discrimination, 37 Ar and 39 Ar decay and nuclear interference reactions. Ages are calculated relative to Drachenfels sanidine of 25.26 Ma (and recalculated from there relative to 28.34 Ma for TCR of Renne et al. (1998)). 40 Ar * (per cent) is the percentage radiogenic 40 Ar in each step. 39 Ar K is the percentage of 39 Ar K released in each step with a total of 100 per cent over the full experiment (Excel format).
This material is available as part of the online article from: http://www.blackwell-synergy.com/doi/abs/10.1111/j.1365-246X.2006.03712.x (this link will take you to the article abstract).
Please note: Blackwell Publishing are not responsibe for the content or functionality of any supplementary materials supplied by the authors. Any queries (other than missing material) should be directed to the corresponding author for the article.