Underthrusting and duplexing beneath the northern Tibetan Plateau and the evolution of the Himalayan-Tibetan orogen

The Cenozoic Qilian Shan thrust belt is the northern margin of the Tibetan Plateau, which developed in part due to progressive India-Asia convergence during Himalayan-Tibetan orogeny. Available geologic observations suggest that this thrust belt started deforming shortly after initial India-Asia collision at 60–55 Ma, and thus its kinematic development is intrinsically related to the construction and evolution of the Tibetan Plateau. Here, we present new field observations from a geologic traverse across the Qilian Shan to elucidate the style of deformation across the active thrust belt. In particular, we infer protracted out-of-sequence deformation here that is consistent with this thrust system remaining a stationary northern boundary to the Tibetan Plateau since the early Cenozoic. We present a lithosphere-scale model for this region that highlights the following: (1) coupled distributed crustal shortening and underthrusting of the North China craton beneath Tibet, which explains the spatial and temporal distribution of observed crustal shortening and thickness, (2) this underthrusting exploited the south-dipping early Paleozoic Qilian suture paleo–subduction mélange channel, and (3) development of a lower-crustal duplex at the lithospheric underthrusting ramp. This last inference can explain the relatively high elevation, low relief, and thickened crust of the central Qilian Shan, as well as the comparative aseismicity of the region, which experiences fewer earthquakes due to less upper-crustal faulting. Both the northern and southern margins of the Himalayan-Tibetan orogen appear to have developed similarly, with continental underthrusting and crustal-scale imbrication and duplexing, despite vastly different climatic and plate-velocity boundary conditions, which suggests that the orogen-scale architecture of the thrust belt is controlled by neither of these forcing mechanisms. Instead, strength anisotropies of the crust probably control the kinematics and style of deformation, including the development of northern Tibet, where thrust systems are concentrated along pre-Cenozoic suture zones.


INTRODUCTION
How the Tibetan crust accommodates India-Asia convergence and deforms to develop the Tibetan Plateau are outstanding questions in the study of the Himalayan-Tibetan orogen (e.g., Yin and Harrison, 2000;Clark and Royden, 2000;Tapponnier et al., 2001; van Hinsbergen et al., 2011avan Hinsbergen et al., , 2011bYakovlev and Clark, 2014;Ingalls et al., 2016;Zheng et al., 2017). The evolution of the Tibetan Plateau directly impacts our knowledge of continental tectonics (Molnar, 1988;Yin, 2010), and examination of the northernmost margin of the Tibetan Plateau is key to unravelling the deformation mechanisms acting in northern Tibet (Fig. 1). Since the 1970s (Dewey and Burke, 1973;Molnar and Tapponnier, 1975), intense study of the Tibetan Plateau has led to a series of end-member models for its formation, including: (1) Cenozoic vertically coherent shortening of the Tibetan lithosphere (England and Houseman, 1986;Dewey et al., 1988), (2) Cenozoic underthrusting of the Asian lithosphere beneath the Tibetan Plateau (Willett and Beaumont, 1994;Kind et al., 2002;Zhou and Murphy, 2005;Zhao et al., 2011;Feng et al., 2014;Ye et al., 2015), (3) Cenozoic vertical inflation of Tibetan crust by lateral channel flow (Zhao and Morgan, 1987;Bird, 1991;Royden et al., 1997Royden et al., , 2008Clark and Royden, 2000), (4) discrete Cenozoic intracontinental subduction coupled with lateral extrusion along major strike-slip faults , (5) Cenozoic uplift of the central plateau via shortening and mantle processes (e.g., England and Houseman, 1989;Harrison et al., 1992;Molnar et al., 1993), followed by outward plateau expansion to the north, east, and south (Wang et al., 2008), and (6) pre-Cenozoic crustal thickening (e.g., Worley and Wilson, 1996;Murphy et al., 1997;Wallis et al., 2003;see Fig. 2). An important difference between many of these models is the way in which the continental crust and mantle lithosphere accommodate the convergence between India and Asia, and the degree of coupling between these two layers (e.g., Willett and Beaumont, 1994).
These tectonic models make specific predictions for the timing of deformation, location and kinematics of crustal structures, and the strain distribution in northern Tibet (e.g., Meyer et al., 1998;Lease et al., 2012;Zuza et al., 2016a). Thus, systematic geological observations ranging from  The Cenozoic Himalayan-Tibetan orogen resulted from the collision of the Indian and Asian plates (e.g., Yin and Harrison, 2000;Royden et al., 2008) at 60-55 Ma (see recent discussions in DeCelles et al., 2014;Hu et al., 2015Hu et al., , 2016. The orogen has a north-south width (i.e., nearly convergence-parallel direction; Zhang et al., 2004) of >2000 km, from the Himalayan fold-thrust belt in the south to the Qilian Shan thrust belt in north (Fig. 1). Most of the present-day topography, crustal thickness, and presently active structures were generated during India-Asia collision (e.g., Yin and Harrison, 2000;Tapponnier et al., 2001;Royden et al., 2008), although there are notable exceptions of inherited crustal thickness and structures as documented for the southern and eastern portions of Tibet (e.g., Worley and Wilson, 1996;Murphy et al., 1997;Kapp et al., 2003Kapp et al., , 2007Robinson et al., 2004;Tian et al., 2016). That said, preexisting structures often control the distribution of Cenozoic deformation, including the Paleozoic-Mesozoic suture zones across Tibet (e.g., Yin and Harrison, 2000;Wu et al., 2016Wu et al., , 2017Zuza et al., 2018).

Q il ia n S h a n -N a n S h a n th ru s t b e lt N o rt h Q a id a m th ru s t s y s te m Q im e n T a g h th ru s t b e lt
The Tibetan Plateau is bounded by the northeast-striking left-slip Altyn Tagh fault in the northwest and three other thrust systems: the westnorthwest-trending Himalaya to the south, the northwest-trending Qilian Shan in the northeast, and the north-trending Longmen Shan in the east (Fig. 1). A cross section across the Tibetan Plateau parallel to India-Asia convergence involves the Himalaya to the south and the Qilian Shan to the north, and thus an understanding of the kinematics and/or dynamics of the Tibetan Plateau requires comprehensive knowledge of the structural framework of these two distinct thrust systems. Investigations across the Himalaya have been robust (e.g., Burchfiel et al., 1995;Johnson, 2002;DeCelles et al., 2002;Robinson et al., 2006;McQuarrie et al., 2008;Webb et al., 2011Webb et al., , 2017Webb, 2013;Gao et al., 2016; and many others), and our knowledge of Cenozoic crustal deformation across the northern plateau margin is growing (e.g., Gaudemer et al., 1995;Meyer et al., 1998;Lease et al., 2012;Gao et al., 2013;Craddock et al., 2014;Cheng et al., 2015;Zuza et al., 2016aZuza et al., , 2017aZuza et al., , 2018Allen et al., 2017;Yu et al., 2017). Herein, discussion of this orogen is concerned primarily with the geology of northern Tibet, north of the Kunlun left-slip fault (Fig. 1).

Geology of Northern Tibet
The northeastern Tibetan Plateau has an average elevation of ~4.5 km, and this high topography decreases rapidly to <1.5 km in the Hexi Corridor foreland basin to the north (Fig. 1). Geophysical observations show that the crust is locally 55-65 km thick (Zhao et al., 2001;Yue et al., 2012;Gao et al., 2013;Xu et al., 2018), and the relatively stable crust beneath Ordos Basin to the north is ~42 km thick ( Fig. 1; Liu et al., 2006;Pan and Niu, 2011).

Paleozoic Geology
The early Paleozoic Qilian orogen, as summarized by Song et al. (2013) and Zuza et al. (2018), involved the collision of the Proterozoic Qaidam microcontinent with the Archean-Proterozoic North China craton to the north in the Ordovician-Silurian. Widespread Cambrian-Ordovician arc volcanic and plutonic rocks represent the Qilian arc, which accommodated the convergence between these two continents (Wu et al., 2017) and the closure of the Qilian ocean, although the number and polarity of these subduction systems are debated (e.g., Xiao et al., 2009;Song et al., 2013;Zuza and Yin, 2017;Zuza et al., 2018). Dismembered mélange-ophiolite complexes are found throughout the Qilian Shan, and they represent the Qilian suture(s) (Yin and Harrison, 2000;Yin et al., 2007b;Menold et al., 2009Menold et al., , 2016Song et al., 2013Song et al., , 2014. As previously outlined in Zuza et al. (2018), we believe the simplest model that explains available field observations involves a single southdipping subduction system that was constructed along the northern margin of the Qaidam continent, as originally described and discussed by Sobel and Arnaud (1999) and Gehrels et al. (2003aGehrels et al. ( , 2003b. Subsequent contractional deformation during early Paleozoic Qilian and Cenozoic Himalayan-Tibetan orogenies reorganized the original tectonic configuration of the early Paleozoic Qilian arc, leading to an apparent multitude of suture zones (e.g., Xiao et al., 2009;Yin et al., 2007b). Qilian orogeny ceased by the Devonian, and Carboniferous strata were deposited unconformably over amphibolite-grade arc and basement rocks (Xiao et al., 2009;Wu et al., 2017).

Mesozoic Geology
Jurassic and Cretaceous extensional and transtensional basin deposits are observed across northern Tibet and the Qilian Shan, from Xining Basin in the south to the Hexi Corridor in the north Pan et al., 2004). These deposits have been related to the Mesozoic closure of the Paleo-Tethys and Meso-Tethys Oceans south of Qaidam Basin (Pullen et al., 2008;Zhang et al., 2014), and associated slab rollback to the south may have led to regional extension that affected much of the continent to the north (Huo and Tan, 1995;Vincent and Allen, 1999;Chen et al., 2003;Yin et al., 2008aYin et al., , 2008b. Extensional faults and related basins have been imaged in the upper crust via seismic profiling (Yin et al., 2008a(Yin et al., , 2008b. Although there are extensive Jurassic and Cretaceous strata and basin deposits, actual Mesozoic normal faults are rarely exposed in the field (e.g., Chen et al., 2014;Cheng et al., 2019). As discussed later, Mesozoic normal faults do not appear to be significantly reactivated by Cenozoic thrusting in either seismic images or field observations.
Starting at 20-15 Ma, northern Tibet experienced a pulse of rapid cooling and exhumation that overlapped with the through-going initiation of the major left-slip Kunlun, Qinling, and Haiyuan faults ( Fig. 3; e.g., Jolivet et al., 2001;Craddock et al., 2011;Duvall et al., 2013;Yuan et al., 2013;Zuza et al., 2016b;Li et al., 2019). Here, we note that although left-slip strike-slip faulting is inferred to have begun at ca. 20-15 Ma in northern Tibet, there are relatively few age constraints for strike-slip fault initiation, especially along the Haiyuan fault ( Fig. 3; e.g., Yuan et al., 2013;Duvall et al., 2013). Recent apatite (U-Th)/He and fission-track (AFT) analyses across a restraining bend of the Haiyuan fault at 37.6012°N and 101.5468°E suggest that rapid exhumation inferred to be related to Haiyuan fault motion occurred at ca. 15-10 Ma (Li et al., 2019), which is consistent with previously proposed initiation ages for the Haiyuan fault ( Fig. 3; see summary in Duvall et al., 2013).

KINEMATICS OF THE TIBETAN PLATEAU'S NORTHERN MARGIN
We modeled shortening rate, crustal thickening, and elevation changes through time across the northern Tibet region using external convergencerate constraints across the entire Himalayan-Tibetan orogen. The goal of this exercise was to generate simple shortening strain predictions to compare against shortening observations from our own work and published data. More specifically, the simple observation that convergence rates across the Himalayan-Tibetan orogen have decreased since initial collision at ca. 58 Ma (Molnar and Tapponnier, 1975;Molnar and Stock, 2009;Copley et al., 2010;DeCelles et al., 2014;Hu et al., 2016) effectively requires shortening rates across the intraplateau thrust belts to have fluctuated accordingly. That is, shortening rates probably decreased through time, although we note that variable strain partitioning among thrust belts may have led to more complex shortening-rate relationships. Therefore, we attempted to provide some external bounds for comparison with field-based shortening data. The present-day north-northeastdirected convergence rate across the Qilian Shan thrust system is 5-7 mm/yr (Zhang et al., 2004;Gan et al., 2007). Here, we take the approach of applying an orogen-scale bulk strain rate-which is inferred to have remained constant throughout the Cenozoic (Clark, 2012)-to northern Tibet to estimate bounds on local deformation magnitude and rate. This approach assumes that the northern boundary of the Himalayan-Tibetan orogen has remained stationary with time (Clark, 2012).
The velocity field across the Himalayan-Tibetan orogen yields an orogen-scale bulk strain rate ( ) of 6.6 × 10 −16 s -1 (i.e., the rate of change of the orogen's width divided by its width, ~40 mm yr -1 /~1900 km; Fig. 1; Clark, 2012). The strain rate at the time of collision can be estimated using plate-circuit reconstructions for the India-Asia convergence rate at the time of collision (ca. 58 Ma) and the estimated orogen width derived from these reconstructions (Molnar and Stock, 2009;van Hinsbergen et al., 2011avan Hinsbergen et al., , 2011bHuang et al., 2017): At ca. 58 Ma, the convergence rate and orogen width were ~120 ± 10 mm/yr and 6050 ± 450 km, respectively, which yields a bulk strain rate of 6.3 ± 0.9 × 10 −16 s -1 . This value overlaps the present-day strain rate of 6.6 × 10 −16 s -1 (Clark, 2012). This rate applied to the Qilian Shan thrust belt can be used to estimate shortening rate and magnitude through time according to the following relationships (Clark, 2012;Zuza et al., 2017b): ( where L(t) is the width of the orogen at time t (where t = 0 m.y. is today, and t = -58 m.y. is the time of collision), L f is the final width of the orogen or L(t = 0 m.y.), x(t) is the width of a deforming segment of the orogen, such that x(t) + L f = L(t), and v(x) is the rate at which a point at position x moves north toward stable Eurasia and is the orogen-scale convergence rate at x = L(t). Total shortening is defined by L(t = -58 m.y.) minus L f . We assumed the final length of this region (i.e., from the northernmost Qaidam Basin to the southernmost Hexi Corridor; Fig. 1) was ~350 km. We also assumed that most of the crust was ~35 km thick prior to the Cenozoic, which we justify because: (1) the relatively undeformed Ordos Basin to the north is ~42 km thick (Liu et al., 2006), and (2) modern extended or rifted continental crust thickness is between 30 and 40 km thick (Christensen and Mooney, 1995), which applies to northern Tibet following Mesozoic extension (Vincent and Allen, 1999). Note that this analysis assumes plane-strain deformation, where "extrusion"-type processes are not accounted for (e.g., Tapponnier et al., 1982;Cheng et al., 2015).
In Figure 4A, we plot shortening rate and magnitude over a range of constant strain rates from 5 × 10 −16 s -1 to 7 × 10 −16 s -1 , assuming either (1) deformation initiated at ca. 58 Ma, or (2) alternatively, deformation started at ca. 40 Ma, as geologic evidence suggests (e.g., Craddock et al., 2011). If the initiation timing is between these values, one can extrapolate between the curves (Fig. 4A). To further demonstrate parameter sensitivity in our model, we modeled two L f values: 350 km and 300 km. For the purposes of this discussion, we primarily use the ca. 40 Ma initiation age for deformation. Using this time and assuming pure-shear crustal thickening, isostatic compensation, a total ~5 km of erosion, and a constant 6.6 × 10 −16 s -1 strain rate, Figure 4B shows the expected crustal thickening and elevation through time. We also plotted the modeled elevation through time of the Qaidam Basin from Yu et al. (2015), which assumes that basin sedimentation leads to crustal uplift following isostasy (Fig. 4B). Note that the Qilian Shan elevation remains at or higher than the estimated elevation of Qaidam Basin, which is required for continuous Cenozoic Qilian Shan-derived sedimentation into the basin and for the Qilian Shan to trap sediments in the endorheic basin . ZUZA ET AL. | Development of the northern Tibetan Plateau RESEARCH These models, which are only based on external kinematic constraints and not on our geologic observations, predict that shortening rates have decreased from ~16 mm/yr in the Eocene at ca. 40 Ma to the present-day value of ~6 mm/yr, which matches global positioning system (GPS)-based estimates ( Fig. 4A; Zhang et al., 2004). The model predicts that shortening magnitude increases to 400-450 km over a 40 m.y. period. These values are slightly higher than existing minimum shortening magnitude estimates across the thrust system (i.e., 250-350 km; Zuza et al., 2018). Both crustal thickness and elevation increase through time to reasonable estimates of 70 km and ~5 km (Fig. 4B), respectively, which is comparable to present-day values (Fig. 1).

FIELD OBSERVATIONS
Detailed field observations and mapping across the central Qilian Shan have previously been reported in Zuza et al. (2018) and Wu et al. (2017). This present study focused on new study sites along the southern and northern margins of the Qilian Shan-Nan Shan thrust belt. Specifically, our observations focused on the Wulanshan (also referred to as the Zongwulong Mountains; Yu et al., 2017), which bounds the Qilian Shan thrust belt to the south, the central thrust belt near the city of Qilian, and the northern range-front thrust system ( Fig. 1). By focusing on these three locations, the goal was to relate Cenozoic deformation across the entire width of the Qilian Shan-Nan Shan thrust belt to provide an integrated view of the development of the northern margin of the Tibetan Plateau. Unit age assignments used here come from existing geologic maps, including those from Pan et al. (2004), Gansu Geological Bureau (1989, and Qinghai BGMR (1991). We did not attempt to date or confirm stratigraphic age assignments, but in the past we have found that previous ages of Paleozoic-Mesozoic units are generally accurate and have been corroborated by detrital zircon maximum depositional ages (e.g., Wu et al., , 2017Zuza et al., 2018).

North Qilian Shan
The North Qilian Shan defines the northernmost extent of the Tibetan Plateau. The predominantly north-directed North Qilian thrust system places Proterozoic-Paleozoic-Mesozoic strata over Cenozoic basin deposits of the Hexi Corridor (Li and Yang, 1998;Zhuang et al., 2011;Zheng et al., 2010;Yang et al., 2018). Typically, one major north-directed thrust system bounds the northern extent of the North Qilian Shan (e.g., Yang et al., 2007;Zheng et al., 2010;Yang et al., 2018), with some prominent exceptions, such as the Yumu Shan (e.g., Tapponnier et al., 1990), which is a diverging fault splay of the main North Qilian Shan thrust. The Hexi Corridor foreland basins to the north of these thrusts are variably deformed by discrete south-directed back-thrust systems to the north, off of the Tibetan Plateau. Geologic mapping and industry seismic reflection profiles show ~6 km of Mesozoic and Cenozoic strata in the Hexi Corridor and two major south-dipping thrust fault systems (Zuza et al., 2016a). This previous study estimated a minimum of ~33 km shortening (53% strain) over a deformed section length of 29 km (Zuza et al., 2016a). Extrapolation of these estimates across the entire North Qilian Shan suggests ~50 km of shortening along the northernmost Tibetan Plateau margin.
The North Qilian range front just southeast of the Yumu Shan rises abruptly from the relatively flat Hexi Corridor (Fig. 5A). Exposure along this thrust-bounded range is often poor, but at location NQ1 in Figure 1, the major thrust fault is observed in drainages placing Carboniferous strata over Cenozoic deposits (Fig. 5B). North-and south-dipping (40°-60° dips) faults are observed in the thrust hanging wall with dip-slip striations ( Fig. 5B). Despite numerous conjugate-fault sets, map-view and field relationships require north-directed thrusting of Carboniferous strata over Cenozoic foreland basin deposits (e.g., Yang et al., 2018). At another location-NQ3 in Figure 1-fault splays are observed in the footwall rocks of the North Qilian thrust system, duplicating Cenozoic strata. An ~100-m-wavelength, southeast-trending, open-close anticline is observed in the hanging wall of a minor north-directed thrust fault splay located entirely within Cenozoic strata (Fig. 5C). These observed minor faults may represent incipient range-bounding structures. Yang et al. (2018) reported that the west-trending Hujiatai anticline to the west of the Yumu Shan is shortening at a rate of 0.8 mm/yr.
Between sites NQ1 and NQ3, at site NQ2 in Figure 1, we observed an uplifted and exposed hanging wall-footwall flat structure, where Ordovician metasedimentary rocks are thrust over Cenozoic basin deposits, and this entire system is truncated by more recent higher-angle thrust faults (Fig. 6). Age constraints on these local Cenozoic sediments are lacking, but they are probably Oligocene or Miocene (Li and Yang, 1998;Pan et al., 2004;Zhuang et al., 2011;Fang et al., 2012). We infer that these exposures represent out-of-sequence deformation along the northern margin of the Tibetan Plateau, based on the following observations and logic. The horizontal contact between Ordovician strata and Cenozoic deposits is a thrust fault that is parallel to both the hanging-wall and footwall rocks (i.e., a thrust flat). This subhorizontal fault is truncated by higher-angle thrust faults that appear to be more recent because they control topography along this range front (Fig. 6). Earlier movement along the high-angle fault would have disrupted a subhorizontal thrust sheet, which would make it difficult to form the observed thrust flat. We also observed relict erosional surfaces that are subparallel to the thrust flat, but that are offset vertically from each other by >670 m by the more recent, higher-angle thrust faulting (Fig. 6A). We interpret an early-stage of thrusting that produced the thrust-flat geometries, when Ordovician strata overthrust onto early Cenozoic (e.g., Eocene-Oligocene; Li and Yang, 1998;Zhuang et al., 2011) foreland basin deposits. Sometime after this faulting, the relict erosional surfaces developed, parallel to the fault flat. More recently, these faults were crosscut by out-of-sequence thrusting by two high-angle faults: the southern one juxtaposed Ordovician rocks against Ordovician   rocks and uplifted the relict erosional surface, whereas the northern one placed the earlier hanging wall-footwall flat over Quaternary deposits of the Hexi Corridor (Fig. 6E).

Qilian City Ranges
Some of the best exposures of high-pressure blueschist rocks of the Qilian suture are located just west of Qilian City in the central Qilian Shan (e.g., Song et al., 2007Song et al., , 2009. Here, along the east-trending Black River Valley, a belt of north-dipping, high-pressure ophiolitic mélange is observed in north-trending drainages. Other ophiolite and blueschist complexes to the southwest, including the Aoyougou and Yushigou localities (e.g., Song and Su, 1998;Xiang et al., 2007), are of similar age but dip south. Thus, we interpret the north-dipping mélange rocks near Qilian City to be the northern limb of a broad anticlinorium of the Qilian subduction complex.
Mesozoic normal faulting was directly observed in the field along the Black River (Fig. 7). Although Cretaceous basin-fill deposits related to normal faulting are prevalent in northern Tibet (e.g., Vincent and Allen, 1999), relatively few normal faults are directly observed except in seismic data (e.g., Yin et al., 2008b;Chen et al., 2014). This may suggest that many of the Mesozoic normal faults were reactivated in the Cenozoic as part of Himalayan-Tibetan orogeny. However, near Qilian City, we observed several Mesozoic normal faults that have not been reactivated by any Cenozoic contractional deformation (Fig. 7). West of Qilian City, along the Black River valley, a north-striking, dip-slip normal fault places Cretaceous strata over Ordovician metasedimentary rocks (Figs. 7 and 8C). The normal fault is orthogonal to the east-northeast-striking north-directed thrust faults that are mapped in this area, and there is no evidence for contractional deformation or reactivation (Fig. 7). Just north of Qilian City, a northwest-striking dip-slip normal fault places Cretaceous strata over Cambrian metavolcanic rocks (Fig. 7). Shear fabrics along the fault zone only show evidence for hanging-wall-down motion, and thus thrust-related reactivation seems unlikely. These observations suggest that reactivation of Mesozoic normal faults was not locally significant in the central Qilian Shan. Elsewhere across the Qilian Shan, we never see Mesozoic strata thrust over younger Cenozoic deposits (e.g., Wu et al., 2017;Zuza et al., 2018).
About 20 km west of these localities, Mesozoic sedimentary rocks are south-dipping, and geologic maps show that these rocks were thrust northward over the Black River valley along a south-dipping hangingwall-flat thrust fault (Fig. 8A;Gansu Geological Bureau, 1989;Pan et al., 2004). Here, Cretaceous medium-grained red-bed sandstone and pebble conglomerate rocks overlie cross-bedded, white-gray, fine-to mediumgrained Triassic sandstones. The beds are subparallel. Ordo vician metasedimentary rocks are thrust over, and crosscut, the package of Mesozoic strata (Figs. 8A and 8B). A north-verging hanging-wall anticline is observed, suggesting an overall hanging-wall ramp at this locality in the Ordovician rocks (Fig. 8B).

Wulanshan Thrust System
The Wulanshan thrust system, located just north of the city of Wulan to the west of Qinghai Lake (Fig. 1), represents the southernmost thrust system of the active Qilian Shan-Nan Shan thrust belt (Wang and Burchfiel, 2004). The south-directed Wulanshan is approximately correlative along-strike to the south-directed Qinghai Nanshan thrust system to the east (Craddock et al., 2014) and the Zongwulong thrust system to the west (e.g., Yu et al., 2017). The thrust-bounded ranges in the Wulanshan surround a Cenozoic basin, in which the city of Wulan is located. Well data and seismic profiles across nearby basins show that they are floored by Eocene Xiaganchaigou and Oligocene Shangganchaigou Formation rocks . Growth-strata relationships of the Xiaganchaigou Formation against Cenozoic thrust faults (i.e., the Olongbulak thrust) imply that deformation initiated here in the Eocene .
The Wulanshan range-bounding north-northeast-dipping thrust fault places Proterozoic-Paleozoic rocks over Miocene basin deposits in the Wulan Basin (e.g., Lu et al., 2012;see also Fig. 8E). Where observed, this fault and minor related fault structures dip 55°-60°N. Within the range, ~10-m-wavelength open fold pairs are observed in the Proterozoic rocks, with fold hinges parallel to the major Cenozoic range-bounding thrust fault. Although the age of this folding is not robustly constrained, based on parallelism of the fold hinge to the range-bounding thrust fault, we assume that folding also occurred in the Cenozoic. Coupled magnetostratigraphy and AFT thermochronology suggest that basin sedimentation and range uplift and exhumation, respectively, initiated at ca. 22-20 Ma (Lu et al., 2012). The Wulan sediments (Fig. 8E) were deposited between ca. 22 Ma and 9 Ma (Lu et al., 2012).
To the east, the Wulanshan thrusts are truncated and offset by the right-slip Elashan fault (Fig. 1), which has late Quaternary slip rate of ~1.1 mm/yr (e.g., Wang and Burchfiel, 2004;Yuan et al., 2011). Given that the Qinghai Nan Shan, which is truncated to the west by the Elashan fault (Fig. 1), has a very low estimated late Quaternary shortening rate of ~0.1 mm/yr (Craddock et al., 2014), we speculate that the Wulanshan accommodates higher shortening rates on the order of ~1 mm/yr. This rate is necessary to produce the observed right-lateral faulting on the Elashan fault at a rate of 1.1 mm/yr, given the fact that the right-slip fault terminates within the central Qilian Shan (Fig. 1).
The Miocene strata in Wulan Basin are variably tilted ~15-20°N ( Fig. 8E; e.g., Lu et al., 2012). The most recently active thrusts crosscut and displace the Miocene sediments, and these faults are observable in satellite images parallel to the prominent range front. This relationship suggests that the active thrusting is deforming the footwall of the rangebounding thrust fault, and the thrust system is advancing to the south along this newly formed fault system. In the interior of the Wulanshan, north  of the southern range front, there are isolated Oligocene-Miocene basin deposits (Qinghai BGMR, 1991) that are steeply tilted today (~80°NE; Fig. 8D). This suggests that following an episode of Oligocene-Miocene exhumation and basin deposition, significant syn-and post-Miocene deformation affected the interior of the Wulanshan.

DISCUSSION
Using a combination of new field observation and existing published data sets, we discuss the tectonic evolution of northern Tibet, from Qaidam Basin in the south to the Hexi Corridor in the north (Fig. 1). Ultimately, the Cenozoic development of this region bears on orogenic plateau formation mechanisms and processes of continental tectonics and plate convergence.

Integrated Shortening Across the Qilian Shan
We integrated available shortening rates across the Qilian Shan thrust belt to see how they match both present-day geodetic convergence rates and our modeled time-dependent shortening rates (Fig. 4). The purpose of this exercise is (1) to demonstrate whether available geologic observations support protracted deformation at a progressively decelerating rate (i.e., ~16 mm/yr to ~6 mm/yr; Fig. 4A) in northern Tibet since the Eocene, and (2) to test our shortening-rate model over the last ~10 m.y. to identify whether it may be valid in the Miocene and older time. We focused on geologic shortening rates (greater than million-year rates) rather than Quaternary rates, which may reflect local spatial or temporal variations of strain during the interseismic loading cycle. Our stain-rate-derived shortening rate curve predicts present-day shortening rates of ~6 mm/yr, which match geodetic convergence rates of 5-7 mm/yr ( Fig. 4A; Zhang et al., 2004;Gan et al., 2007). We first tested whether more recent geologic shortening rates (ca. 10-0 Ma) across the Qilian Shan-Nan Shan thrust belt sum to 5-7 mm/yr. There have been recent shortening rate estimates across the North Qilian Shan thrust belt of 3.3 ± 0.6 mm/yr (Zuza et al., 2016a; see also similar but slightly lower estimate by Champagnac et al., 2010) and across the integrated Gonghe Nan Shan and Qinghai Nan Shan of ~0.9 mm/yr (Craddock et al., 2014), which is comparable to the probable ~1 mm/yr shortening rate estimated in this study across the Wulanshan (Fig. 9D). Presently, there are no well-resolved shortening rates estimated across the central Qilian Shan. Based on cross-section restorations across an ~120 km traverse, Zuza et al. (2018) provided some preliminary bounds, although the timing of deformation initiation is uncertain: Rate estimates range from ~1.2 mm/yr assuming ca. 50 Ma deformation initiation to ~4.1 mm/yr assuming ca. 15 Ma initiation. Although speculative, here we assume that the decrease in shortening rate across this central segment follows the same exponential decay as Equation (3) (Fig. 4)-which mirrors the decay in overall plate-convergence velocities (Molnar and Stock, 2009;Copley et al., 2010)-and thus the local shortening rate across the central Qilian Shan decreased from ~5.3 mm/yr at ca. 40 Ma to ~2.9 mm/yr at ca. 10 Ma to ~2.3 mm/yr today. Summing these average rates across the entire width of the Qilian Shan over the last ~10 m.y. yields geologic shortening rates of 6.2-7.4 mm/yr (Fig. 9D), which broadly agree with geodetic convergence and predicted thrust-belt-scale shortening rates (Fig. 4).
Next, we investigated whether available geologic observations suggest an exponential decrease in shortening rate across the thrust system through time, from ~16 mm/yr at ca. 40 Ma to ~10 mm/yr at ca. 10 Ma (Fig. 4). Importantly, this scenario suggests and predicts that more than half of the total Cenozoic shortening strain across the Qilian Shan occurred by the early Miocene. Eocene cooling ages across the central Qilian Shan (Qi et al., 2016;Jian et al., 2018;Du et al., 2018;Li et al., 2019) suggest local exhumation distributed across the region at this time. The Cenozoic Lulehe Formation of northern Qaidam Basin was sourced from the southern Qilian Shan and records the first clastic sedimentation derived from deformation-related uplift. It is traditionally assigned an early Eocene age (Yin et al., 2008b;Yu et al., 2017;Ji et al., 2017;Du et al., 2018;, although Wang et al. (2017) recently suggested a ca. 25.5 Ma age for basal Lulehe strata. We acknowledge that the Lulehe Formation age is debated, but based on early Cenozoic cooling ages observed from across the Qilian Shan, we prefer an Eocene age for the Lulehe Formation. Paleosols sampled from the Lulehe Formation yield δ 18 O values that suggest cooling and >1 km uplift of the northern Qaidam Basin since Lulehe deposition (Song et al., 2018). However, these isotopic data should be interpreted with some caution because the Lulehe samples were probably buried deep enough to be significantly heated, and thus potential alteration is a concern (e.g., Jesmok et al., 2018;Song et al., 2018).
Constraints on the magnitude of this early Cenozoic, probable Eocene, shortening are limited. There are two pieces of indirect field evidence that suggest that shortening strain was greater prior to ca. 10 Ma. First, integrating the above deduced ~6 mm/yr over the last ~10 m.y. yields 60 km of shortening across the Qilian Shan thrust belt, which leaves at least ~240 km of remaining shortening (e.g., Meyer et al., 1998;Zuza et al., 2016aZuza et al., , 2018 to have accumulated prior to ca. 10 Ma, at a bulk average shortening rate of ~8 mm/yr from 40 Ma to 10 Ma (Fig. 9D). This is a minimum rate, given that methods of cross-section restorations are minimum estimates (e.g., Judge and Allmendinger, 2011). This is higher than the 10-0 Ma rate, and, depending on the severity of shortening underestimation, the shortening rate could be higher (e.g., >400 km total shortening yields an average rate of >11 mm/yr).
In addition, field observations from the central Qilian Shan suggest that middle Miocene and younger contractional deformation primarily occurred along south-directed thrust faults (Reith, 2013;Zuza et al., 2018). These faults crosscut north-directed thrusts that are presently undated and appear to comprise at least ~60% of the stain across the mapped traverse , which equates to ~1.5 mm/yr shortening from 40 Ma to 10 Ma. Reconnaissance thermochronology (Zuza et al., 2016b(Zuza et al., , 2017b suggested local exhumation at ca. 50 Ma. Elsewhere across the Qilian Shan, major thrust faults are crosscut by Miocene and younger structures, which obscure the earlier phases of deformation . For example, Qi et al. (2016) reported Eocene to middle Miocene exhumation of the central Qilian Shan (~38°02′N, ~99°14′E) as interpreted from AFT track-length modeling. The folds and thrusts in their cross section (Qi et al., 2016, their fig. 2) are consistent with approximately ~30% shortening strain, which, applied over the South Qilian range (i.e., just northwest of Qinghai Lake), would correspond to a rate of ~1.3 mm/yr during this time period (Fig. 9D). Additional pre-middle-Miocene shortening to the south across the Wulanshan/Qinghai Nan Shan is entirely unconstrained. However, given that Wulan Basin sedimentation spanned 22-9 Ma and Wulanshan exhumation initiated at ca. 20 Ma (Lu et al., 2012), and our ~1 mm/yr rate for shortening across the integrated Wulanshan-Qinghai Nan Shan was from ca. 10 Ma to present (Craddock et al., 2014), appreciable shortening and exhumation must have occurred prior to ca. 10 Ma to provide detritus for Wulan Basin. We tentatively suggest at least ~1 mm/yr of shortening occurred across the Wulanshan from ca. 20 Ma to ca. 10 Ma (Fig. 9D), although this is speculative. Finally, approximately ~0.5 mm/ yr was estimated across the North Qilian Shan (Zuza et al., 2016a) prior to ca. 10 Ma (Fig. 9D). These observations are consistent with the Qilian Shan having a shortening rate of at least 4-5 mm/yr prior to the middle Miocene (Fig. 9D). This corresponds to 120-150 km shortening from ca. 40 Ma to ca. 10 Ma, roughly half of the ~240 km shortening that has been observed , which implies that this rate tabulation RESEARCH is missing deformation. Whether the residual ~10 mm/yr of shortening predicted by our shortening-rate model occurred across the Qilian Shan remains unclear and should be the basis for future field studies.

Role of Southward Underthrusting
How the convergence between the India and Asia lithosphere is accommodated in northern Tibet depends on whether or not the upper mantle and crust are coupled (e.g., Burchfiel et al., 1989;Willett and Beaumont, 1994;Gao et al., 1999). Field observations show that the Tibetan crust between the Kunlun fault and the Hexi Corridor (Fig. 1) has deformed by ~30% to 50% shortening strain, but the impact of continental plate convergence on the mantle lithosphere of Tibet or Eurasia is less well constrained. Models of vertically uniformly distributed shortening (e.g., Dewey and Bird, 1970;Dewey and Burke, 1973;England and Houseman, 1986) suggest thickening of the mantle lithosphere; for example, 50% shortening strain should double the thickness of the mantle lithosphere, assuming pure-shear thickening. Although central Tibet may have thickened mantle lithosphere, northern Tibet does not appear to have been extensively thickened (e.g., Priestley and McKenzie, 2013).
Alternatively, as postulated by some geophysical studies, including deep-seismic reflection profiling and more recent receiver function analyses, the mantle lithosphere of the North China craton may be underthrusting the northern Tibetan Plateau (e.g., Willett and Beaumont, 1994;Gao et al., 1999;Feng et al., 2014;Ye et al., 2015). That is, the crust and upper mantle along the northern boundary of the Tibetan Plateau are decoupled. More recent anisotropy and seismic velocity studies also corroborate underthrusting (e.g., Huang et al., 2017;Ding et al., 2017), although this hypothesis remains debated (Wei et al., 2017). These models suggest that the southern edge of the North China craton's mantle lithosphere is underthrusting from the northern plateau margin to a position underneath northern Qaidam Basin (Fig. 1A). The southward motion of North China's mantle lithosphere beneath the Tibetan crust and mantle lithosphere would presumably be associated with equivalent-magnitude shortening in the middle-to-upper crust as it resisted subduction (i.e., crustal accretion; see Himalayan examples: Capitanio et al., 2010;Webb, 2013;Gao et al., 2016). We envision that this process would result in focused deformation along the underthrusting interface, such that higher strain magnitudes should exist along the boundary between North China and Tibet. Indeed, relatively large-magnitude upper-crustal shortening (i.e., >50 km shortening or >50% shortening strain) has been documented along the interface between the Tibetan Plateau and North China in the North Qilian Shan (Zuza et al., 2016a). Roughly two thirds of total present-day convergence across the thrust belt is accommodated along this northern frontal thrust zone.
There are two caveats to the above interpretation: (1) Upper-crustal shortening may represent a combination of underthrusting accommodation and vertically coherent shortening strain of the Tibetan lithosphere. That is, the total convergence magnitude should equal  the sum of underthrust magnitude and internally distributed vertically uniform lithospheric shortening. Thus, the magnitude of underthrusting is less than or equal to total lithospheric shortening.
(2) The initiation of underthrusting and crustal shortening may have occurred at different times. The timing of underthrusting cannot be constrained from geophysical methods, and it may have occurred anytime during the Cenozoic or potentially earlier (Allen et al., 2017). Earlier underthrusting initiation implies a slower bulk rate of underthrusting, whereas more recent initiation requires faster rates.

Duplex Development Beneath the Qilian Shan
As stated above, the underthrusting process predicts that the magnitude of underthrusting of the mantle lithosphere should be equal to or less than observed crustal shortening strain as North China's crust resists subduction beneath the Tibetan mantle lithosphere. Shortening strain may occur via crustal-scale faults that accommodate vertically coherent horizontal shortening from the upper to lower crust. Alternatively, if the lower and middle-upper crust is decoupled (e.g., Burchfiel et al., 1989;Gao et al., 1999), different styles of deformation are expected between crustal layers, which may include lower-crustal duplex development or channelized flow of low-viscosity layers in response to horizontal pressure gradients (e.g., Clark and Royden, 2000;Gao et al., 2016). We prefer a model with several detachment horizons and duplexes for the following reasons: (1) Several detachment horizons in the lower and middle crust have been imaged seismically (Gao et al., 1999), which may have favored decoupled deformation at different crustal levels (Burchfiel et al., 1989). In addition, the preexisting weaknesses associated with the early Paleozoic Qilian orogen (Allen et al., 2017;Zuza et al., 2018) may have facilitated local duplexing in horizons of concentrated subduction-mélange material, which would involve friction-reducing clays, hydrated phyllosilicates, and relatively low-viscosity sediments (e.g., Collettini et al., 2009;Behr and Becker, 2018).
(3) There is a local crustal thickness maximum in the location of our inferred duplex ( Fig. 1C; Yue et al., 2012;Tian et al., 2014;Xu et al., 2018). Although there is a smooth decrease in crustal thickness to the east along strike of the Qilian Shan, the north-south variations are more erratic. The crust in the central Qilian Shan is ~10 km thicker than to the north and south (Tian et al., 2014), and as noted previously, this location has fewer surface-breaching thrust faults, particularly active faults, than other areas of the Qilian Shan (e.g., Allen et al., 2017). The crustal thickness of Qaidam Basin is ~10 km thinner than the Qilian Shan, despite the magnitude of crustal shortening being comparable (e.g., Yin et al., 2008aYin et al., , 2008bZuza et al., 2016aZuza et al., , 2018. Therefore, it appears that crustal thickening here locally requires some other source, such as duplexing or channelized flow of the lower crust. (4) Recent receiver function analyses by Xu et al. (2018) showed a complicated pattern of thickened crust beneath the Hexi Corridor, varying from ~45 km to 55 km thick, which does not appear to match surface fold or fault structures. Middle-crust detachments and passive thickening via duplexing and lower-crustal accretion could explain these local variations in crustal thickness.
(5) Lastly, this region is the location of the lithospheric ramp that accommodates the southward underthrusting of the Asian mantle lithosphere (Feng et al., 2014;Ye et al., 2015;this study). This structure could focus our inferred duplex development as the lower crust resists subducting deeper and is underplated and/or accreted to the Tibetan crust.
Although the above evidence suggests a duplex structure, we recognize that alternative scenarios are possible. For example, there could have been passive thickening of the crust around the Haiyuan fault latitude via lower-crustal flow; the same regions in which we envision duplexing could have thickened via channel flow processes. However, we prefer the duplex model on the basis of seismic reflection imaging, which does not reveal subhorizontal shear zones as predicted by channel flow models, but rather warped and folded reflectors suggesting folding and faulting in the middle and lower crust, such as folded duplex thrust sheets (Wang et al., 2011;Gao et al., 2013;Guo et al., 2016). In addition, horizontal crustal shortening values observed across northern Tibet are more than enough to vertically thicken the crust (Gaudemer et al., 1995;Meyer et al., 1998;Yin et al., 2008aYin et al., , 2008bLease et al., 2012;Zuza et al., 2016aZuza et al., , 2018, which means additional crustal thickening due to a lower-crustal channel flow process would be expected to overthicken the crust. Given erosion magnitudes are <5 km, it is hard to imagine how this overthickened crusted could be thinned to observed present-day crustal thickness values.

Integrated Kinematic Model for the Development of the Qilian Shan
Based on our new field observations, existing geophysical data, and the inferences discussed above, we present a new kinematic model for the evolution of the northern Tibetan Plateau that is summarized in the crosssection model in Figure 9. Prior to the Cenozoic, the Tibetan and North China lithospheres were separated by the early Paleozoic Qilian suture, and the crust was ~35 km thick (Fig. 9A). North Tibet started deforming quickly after initial India-Asia collision at ca. 58 Ma (e.g., DeCelles et al., 2014;Hu et al., 2015Hu et al., , 2016, broadly exploiting the mechanically weak Qilian suture zone as a lithosphere-scale thrust ramp (Fig. 9B). Tibetan crust deformed and thickened around the North Qaidam and Wulanshan thrust systems (Fig. 9B; and related Olongbulak thrust; Yu et al., 2017). We envision that underthrusting of North China's mantle lithosphere beneath Tibet initiated sometime during this time interval, although the exact timing remains unresolved. If underthrusting was delayed, partial pure-shear thickening of the mantle lithosphere must have occurred to accommodate convergence (e.g., Priestley and McKenzie, 2013). Underthrusting occurred along the relict suture zone, and the middle and upper crust above this underthrust ramp deformed to accommodate North China's southward motion (Fig. 9B). Duplexing of the lower crust occurred at the mantle lithosphere ramp and passively thickened the crust around Qinghai Lake and the future Haiyuan fault trace (Fig. 1). From the Eocene to early Miocene, we envision this region accommodated at least ~150 km of Tibet-Asia convergence.
From the Miocene to present, underthrusting and crustal shortening continued along the same lithospheric ramp, and North China's lower crust continued to form duplex structures at this ramp as it resisted being subducted beneath Tibetan mantle lithosphere (Fig. 9C). The Tibetan crust around this duplex was passively thickened, forming a high-elevation and low-relief landscape  with minimal active surfacebreaching thrust faults (Allen et al., 2017). The Haiyuan left-slip fault, and the relatively minor Riyueshan and Elashan right-slip faults, initiated at around ca. 15-10 Ma near this ramp structure, facilitated by a combination of the preexisting weakness of the early Paleozoic suture zone and overthickening of the crust (Fig. 9C). The highest magnitude of crustal shortening was concentrated along the northern plateau margin, at the interface between the underthrusting North China and overriding Tibetan RESEARCH lithosphere. Net underthrusting of North China is ~150 km, as suggested by geophysical studies (Fig. 9C; Ye et al., 2015).
What is not shown because of the scale of this cross section and its schematic nature are crosscutting relationships. However, it should be noted that deformation was concentrated in the north Tibetan crust over at least ~40 m.y., with minimal foreland advancement. The prominent exception is the northward-propagating thrusts underlying the Yumu Shan that advanced into the Hexi Corridor, which has occurred only over the last several million years (e.g., Tapponnier et al., 1990;Hu et al., 2017). That said, the North Qilian Shan remained the approximate northern plateau boundary during Cenozoic deformation (Clark, 2012).

Location of Underthrust Ramp
The position of the lithospheric ramp that accommodates southward underthrusting of North China beneath Tibet (Fig. 9) to approximately the latitude of Qaidam Basin ( Fig. 1; e.g., Ye et al., 2015) is critical for the location and style of crustal deformation, including our inferred lower-crustal duplexing (Fig. 10). This is because we infer that the lower crust of North China resists subduction beneath the Tibetan Moho and mantle lithosphere. Therefore, the simplest scenario is for both lower-and upper-crustal deformation to concentrate toward the foreland of this large underthrust system (Fig. 10), similar to the Himalaya (e.g., Long et al., 2011;Webb, 2013;Haproff et al., 2018). The position of the lithospheric ramp may be located at one of two end-member locations, assuming net underthrusting of North China's mantle lithosphere to approximately the latitude of northern Qaidam Basin (Fig. 1): either (1) a northern ramp position near the present-day margin of the plateau, or (2) a southern position underneath the Haiyuan fault (Fig. 10). If the ramp is in the south, a similar magnitude of upper-and lower-crustal shortening, S 1 , is predicted to the north of this ramp location in the Qilian Shan thrust belt (left panel in Fig. 10). If the ramp is located to the north by the plateau's northern margin, a higher magnitude of mantle-lithosphere underthrusting, and therefore crustal shortening, S 2 , is predicted (i.e., S 2 > S 1 ). This upper-crustal shortening should occur north of the ramp location, in the relatively undeformed Hexi Corridor foreland basin (central panel in Fig.   10). We envision one mechanism to reconcile a northern ramp location: wedge tectonics, where southward underthrusting of North China's mantle lithosphere is fed to south-directed faults in the central Qilian Shan (right panel in Fig. 10). This requires a high magnitude of shortening, S 2 , which matches a high magnitude of North China underthrusting.
Through these geometric arguments, we can better refine the magnitude of continental underthrusting along the plateau's northern margin. The northern ramp option, located beneath the North Qilian Shan (i.e., the Tibet-North China interface), suggests that upper-crustal shortening that accommodates southward underthrusting should have occurred at this interface and to the north in the Hexi Corridor (Figs. 1 and 10). However, Cretaceous-Cenozoic strata in the Hexi Corridor are essentially undeformed (see seismic profiles in Zuza et al., 2016a), which means that all of the deformation associated with underthrusting is accommodated to the south of the plateau's margin. Instead, we suggest that the mantle-lithosphere ramp is located around the latitude of the Haiyuan fault ( Figs. 9 and 10). This position may explain deformational features here, as discussed above, and requires a relatively lower magnitude of total underthrusting of ~150 km. We note that this ramp could be conceivably located beneath the North Qilian Shan and Hexi Corridor if a wedge-tectonic scenario is operating (Fig. 10).

Initiation Mechanism for Strike-Slip Faulting in Northern Tibet
Strike-slip fault initiation in the Qilian Shan occurred around ca. 15-10 Ma, roughly 30 m.y. after shortening initiated there Duvall et al., 2013;Li et al., 2019). A two-stage development of the Qilian Shan has been proposed, with early thrusting transitioning to mixed-mode thrustand-strike-slip faulting (Lease et al., 2012;Yuan et al., 2013) in the middle Miocene. However, if the strain rate across the orogen has remained constant (Clark, 2012), this implies a constant stress, and this thrust-to-strike-slip faulting transition may instead have resulted from protracted deformation.
Based on the kinematic model presented in this study, we hypothesize that the initiation of strike-slip faulting may have occurred due to progressive overthickening of the crust, leading to a switch of the intermediate compressive stress orientation from horizontal to vertical (Fig. 11).  Specifically, this may have been driven by lower-crust duplexing near the main lithospheric ramp for southward underthrusting, as discussed above. This process would have passively thickened the crust via duplex underplating, causing a rotation of the intermediate compressive stress orientation to vertical. Our model implies ~150 km underthrusting of North China beneath Tibet, and if we restore this underthrusting at the time-varying rates modeled in the section on "Kinematics of the Tibetan Plateau's Northern Margin" (Fig. 4), we find that underthrusting initiated at ca. 20-15 Ma. This is a maximum rate, as it assumes that all of convergence was accommodated via underthrusting; it instead may have been partitioned between underthrusting and distributed vertically coherent lithospheric shortening. However, it is interesting to note that the estimated 20-15 Ma initiation of underthrusting coincides with the estimated timing when strike-slip faulting became a dominate deformation mechanism in northern Tibet. We speculate that the process of underthrusting may control the style of deformation in the upper crust, as has been postulated elsewhere in Tibet (e.g., Molnar et al., 1993;Styron et al., 2015). Alternatively, initiation of both underthrusting and strike-slip faulting may be responses to protracted crustal thickening.

Implications for the Evolution of the Himalayan-Tibetan Orogen
A cross section drawn across the Himalayan-Tibetan orogen and Tibetan Plateau, parallel to north-northeast-trending India-Asia convergence, involves the Himalaya to the south and the Qilian Shan to the north (Fig. 1). These two thrust belts balance body forces associated with the high gravitational potential energy (GPE) of the ~70-km-thick Tibetan Plateau (England and Houseman, 1989;Molnar et al., 1993;England and Molnar, 1997). Therefore, the evolution of this entire system and the growth of the plateau depend on the structural framework of these thrust belts. The overall architectures of the Himalaya and Qilian Shan are comparable: Both involve continental underthrusting, crustal-scale imbrication, and duplexing along the convergence-perpendicular margins of the Himalayan-Tibetan orogen (Argand, 1924;Powell and Conaghan, 1973;DeCelles et al., 2002;Gao et al., 1999Gao et al., , 2013Gao et al., , 2016Zuza et al., 2016aZuza et al., , 2018this study). Interestingly, these similarities between the orogen-scale architectures persist despite vastly different climatic and plate-velocity boundary conditions (Table 1; Figs. 12 and 13), as discussed below. This begs the question: What factors control the style of orogenesis (e.g., Mouthereau et al., 2013)? For example, there is a long-standing debate whether climate and erosion affect the style and magnitude of continental deformation (e.g., Avouac and Burov, 1996;Beaumont et al., 2001;Whipple and Meade, 2004;McQuarrie et al., 2008).   *Thermochronometric data from apatite fission track, zircon (U-Th)/He, or higher-temperature systems; lower-temperature systems, including apatite (U-Th)/He, are excluded because they are significantly affected by topographic relief and may vary with elevation.   Overall present-day convergence rates across the Himalaya are four times greater than that of the Qilian Shan, which correlates with four times greater magnitudes of crustal shortening and continental underthrusting (Table 1; Fig. 13; Powell and Conaghan, 1973;DeCelles et al., 2002DeCelles et al., , 2011Schulte-Pelkum et al., 2005;Nábělek et al., 2009;Yin et al., 2010;Zhao et al., 2010;Long et al., 2011;Webb et al., 2011;Styron et al., 2015;Zuza et al., 2016aZuza et al., , 2018. Given the exponential decrease in India-Asia convergence rates (Molnar and Stock, 2009;Copley et al., 2010), these rates and shortening magnitudes have also probably decreased proportionally throughout the Cenozoic. Mean annual precipitation across the Tibetan Plateau decreases from meters per year across the Himalaya to <35 cm/yr across the Qilian Shan (Hijmans et al., 2005;Bookhagen et al., 2005;Bookhagen and Burbank, 2006;Bookhagen, 2014;see also Fig. 12). Isotopic and paleontological data suggest that northern Tibet has been similarly dry since the start of the Oligocene (e.g., Wang et al., 2003;Dupont-Nivet et al., 2007). The Eocene climate in northern Tibet is less well constrained, including whether it was a warmer and wetter time period (e.g., Dupont-Nivet et al., 2008;Jesmok et al., 2018). The across-strike Tibetan Plateau precipitation variations correlate with denudation and exhumation. AFT and higher-temperature thermochronometers from samples across the Himalaya yield Cenozoic ages of <15-10 Ma, corresponding to exhumation magnitudes of >20 km (e.g., Thiede and Ehlers, 2013;McQuarrie and Ehlers, 2015). Conversely, similar system cooling ages from the Qilian Shan are generally Mesozoic or older, and exhumation is estimated to be <5 km (George et al., 2001;Jolivet et al., 2001;Zheng et al., 2010;Qi et al., 2016; see also Table 1).
In summary, these observations suggest that while the absolute magnitude of crustal shortening and underthrusting is controlled primarily by total convergence across each thrust system, the orogenic architecture of the northern and southern margins of the Himalayan-Tibetan orogen (Fig. 13) is not strongly affected by climate or plate-velocity forcing mechanisms. Despite different convergence rates and climate-related erosion/exhumation rates, the two thrust belts involve similar deformational processes, including underthrusting and duplexing. Instead, the overall thrust-belt structure may be controlled by the age, temperature, and related strength of the deforming lithosphere (e.g., Kong et al., 1997;Mouthereau et al., 2013). A compilation of global orogens by Mouthereau et al. (2013) revealed continental underthrusting is predicted for orogens with a relatively old, cold, and strong foreland, whereas thrust belts in young warm lithosphere are expected to involve vertically coherent pure-shear deformation. The Tibetan crust, with a relatively low elastic thickness (T e ; e.g., T e < ~20 km; Braitenberg et al., 2003), is bounded to the north and south by the Archean-Proterozoic North China and Indian cratons, respectively, which predicts underthrusting-style deformation around the Tibetan Plateau margins (Mouthereau et al., 2013). Furthermore, deformation focused along preexisting weaknesses in northern Tibet related to the early Paleozoic Qilian orogen, such that this south-dipping subduction channel was favorably oriented to serve as a ramp structure to initiate and facilitate south-directed continental underthrusting. Ultimately, the orogenic structural style in the Himalayan-Tibetan orogen may be attributed to strength heterogeneities in predeformational crust (e.g., Kong et al., 1997;Chen and Gerya, 2016;Mouthereau et al., 2013).
That said, the kinematic development of the Qilian Shan-Nan Shan thrust belt is distinct from the Himalaya for two reasons. First, the Qilian Shan is relatively wide (~350 km), especially given its lower convergence rate and magnitude, compared to the Himalaya (~250 km). Second, shortening across the Himalaya is presently concentrated almost entirely along the Main Frontal thrust (Lavé and Avouac, 2000;Burgess et al., 2012), whereas the Qilian Shan has active shortening distributed across the thrust belt (Meyer et al., 1998;Allen et al., 2017; this study). The lower precipitation (Fig. 12) and limited focused erosion in the Qilian Shan compared to the Himalaya may facilitate these differences (e.g., Avouac and Burov, 1996;Willett, 1999;Beaumont et al., 2001;Whipple and Meade, 2004;McQuarrie et al., 2008). Wide distributed deformation in northern Tibet may reflect a mechanically weak basal detachment beneath the Qilian Shan, which also predicts the observed low-angle taper and substantial back thrusting (e.g., Burg et al., 1994;Malavieille, 2010). This weak detachment horizon would be either a subductionmélange channel of the early Paleozoic Qilian orogen (Song et al., 2013;Zuza et al., 2018) or weak and flowing low-viscosity lower crust (Bird, 1991;Clark and Royden, 2000). The early Paleozoic Qilian suture zone may serve as this preexisting weakness with friction-reducing clays and hydrated phyllosilicates (e.g., Collettini et al., 2009;Behr and Becker, 2018). This south-dipping paleo-subduction system  was thus favorably oriented to serve as a ramp structure to initiate and facilitate south-directed continental underthrusting. On the other hand, evidence for lower-crustal flow in this region of the plateau is inconclusive (Lease et al., 2012;Zuza et al., 2016a).
The Qilian Shan has persisted as the northern boundary of the Himalayan-Tibetan orogen and Tibetan Plateau since shortly after India-Asia collision. Given that deformation initiated here shortly after collision Clark et al., 2010;Yin et al., 2008aYin et al., , 2008bCraddock et al., 2011), a typical foreland-propagating thrust belt would be expected to migrate northward through progressive footwall/foreland accretion and/or overthrusting, thus moving the northern orogen boundary northward. However, the Qilian Shan thrust belt did not propagate northward during the Cenozoic, which is an inference that is supported by the undeformed nature of the Hexi Corridor, with ~5-10 km of parallel Cenozoic-Cretaceous sediments (e.g., Li and Yang, 1998;Zhuang et al., 2011;Zuza et al., 2016a). Over the last several million years, several thrust splays have advanced into the basin, as indicated by thermochronology studies along basin-bounding thrust panels . Our observations of the exposed hanging-wall flat in the Hexi Corridor (Fig. 6) corroborates some northward growth of the plateau, which may be related to local plateau expansion due to critical overthickening of the central Qilian Shan to the south. A stationary northern boundary suggests that the Qilian Shan-Nan Shan thrust belt has persisted as an out-of-sequence thrust system to preserve North China as a relatively undeformed block.

CONCLUSIONS
In this study, we presented field observations from a geological traverse across the Qilian Shan thrust belt in northern Tibet that demonstrate that protracted and overprinting Cenozoic contractional deformation has affected this region since the Eocene. We also developed a strain-ratebased shortening model to compare against field-based shortening estimates, which suggests an overall decrease in shortening rate in northern Tibet from ~16 mm/yr to ~6 mm/yr that correlates with the decreasing India-Asia convergence rates through the Cenozoic. This requires that most of the shortening strain in northern Tibet occurred prior to the middle Miocene. Available geologic shortening rate observations and crosscutting relationships discussed herein are consistent with this model, although more constraints of pre-middle Miocene deformation are required.
By integrating new field observations with published shortening studies, thermochronology, and geophysical surveys, we developed a lithosphericscale tectonic model for the development of the northern margin of the Tibetan Plateau. Key factors in this model are progressive convergence between Tibet and Asia since the Eocene, ~150 km of southward underthrusting of North China's mantle lithosphere along an early Paleozoic subduction mélange complex, and duplexing of North China's lower crust and related passive uplift of the central Qilian Shan. Duplexing may have occurred along the underthrusting ramp as North China crust resisted subducting beneath Tibetan lithosphere, which can explain the high elevation, low relief, relative aseismicity, and thickened crust of the central Qilian Shan. Ultimately, our integrated model for the development of the northern Tibetan Plateau explains available field and geophysical observations. Some aspects of our modeled tectonic history and time-varying shortening evolution are speculative, but these provide an updated framework to be tested via future field, geophysical, and geochemical studies.
The south-dipping early Paleozoic Qilian suture acted as an important preexisting weakness in northern Tibet to focus deformation, and its position probably assisted in the decoupling of the crust and mantle lithosphere during southward underthrusting. This suture zone thus helped to define the northern extent of the Cenozoic Tibetan Plateau. The Miocene initiation of strike-slip faulting along the Haiyuan and Kunlun faults paralleled pre-Cenozoic suture zones, which further suggests that strength anisotropies are critical in controlling deformation across the Himalayan-Tibetan orogen.
The Cenozoic evolution of northern Tibet is similar to that of the Himalaya; both involving large-scale underthrusting of mantle lithosphere beneath Tibet. These similarities persist despite different climatic and platevelocity boundary conditions, suggesting that the orogen-scale architecture of these thrust belts bounding the Tibetan Plateau is controlled by neither of these forcing mechanisms. The style and location of deformation around the plateau are instead controlled by local strength anisotropies, such as preexisting weaknesses like pre-Cenozoic sutures, which focus deformation.