Large-scale topography of the north Tibetan ranges as a proxy for contrasted crustal-scale deformation modes

Forming the northern margin of the Cenozoic Himalayan–Tibetan orogen, the area surrounding the Qaidam Basin is a key region to study the mechanisms that govern the long-term evolution of intracontinental orogenic processes. Indeed, tectonic deformation in this region involves newly formed and inherited crustal and lithospheric structures, partition between horizontal and vertical movements, and a strong influence from external factors such as climate and erosion. This study proposes a description of the topography and crustal structure of the north Tibet ranges as an insight into their tectonic deformation characteristics. The topographic study includes the integrated analysis of regional topographic profiles, river profiles, slope maps and thermochronology data. The available geophysical, geological and structural data are used to draw regional, crustal-scale geological sections across northern Tibet to describe the regional crustal structure. We show that three deformation modes are expressed (block uplift, distributed shortening and crustal buckling) in structural compartments that involve either complete mountain ranges or some specific regions inside those ranges. We propose that a new crustal structure may develop parallel to the Altyn Tagh fault, connecting the eastern tip of the Qimen Tagh with the Haiyuan fault, across the Qaidam Basin. Supplementary material: Supplementary figures are available at https://doi.org/10.6084/m9.figshare.c.5841461 Thematic collection: This article is part of the Fold-and-thrust belts collection available at: https://www.lyellcollection.org/cc/fold-and-thrust-belts

The topographic and tectonic evolution of the ranges that surround the Qaidam Basin in northern Tibet is central to understanding of the intracontinental evolution of the major Cenozoic Himalayan-Tibetan collisional orogen. These ranges have been, and are still, the object of a number of important studies (e.g. Tapponnier et al. 1990;Matte et al. 1996;Meyer et al. 1998;Mock et al. 1999;Jolivet et al. 2001;Yin et al. 2007aYin et al. , 2008Cheng et al. 2015a, b;Allen et al. 2017;Groves et al. 2020;Wu et al. 2020;L. Wang et al. 2021) (Fig. 1). For example, deciphering the modalities of the northward propagation of the Tibetan plateau requires understanding of the interaction and partitioning between large strike-slip and thrust faults in a single regional stress-field. Northern Tibet includes some of the largest strike-slip faults in Asia such as the left-slip Kunlun, Altyn Tagh and Haiyuan faults, directly associated with major fold and thrust belts such as the Qilian Shan (Molnar and Tapponnier 1975;Peltzer et al. 1989;Lasserre et al. 1999;Jolivet et al. 2001). The topographic evolution of the plateau and of orogenic domains in general is closely controlled by the complex interactions between tectonics, erosion, sedimentation and climate. The endorheic Qaidam Basin offers a unique sedimentary archive to address these processes (e.g. Heermance et al. 2013;Chang et al. 2015;Cheng et al. 2015bCheng et al. , 2021aJi et al. 2017).
All these topics are strongly dependent on a precise understanding of the regional structural pattern and deformation style. Structural, geomorphological and geochronological studies have therefore focused on establishing a precise timing of the deformation (e.g. Mock et al. 1999;George et al. 2001;Jolivet et al. 2001;Clark et al. 2010;Lease et al. 2011;Duvall et al. 2013;Yuan et al. 2013;B. Li et al. , 2020Staisch et al. 2020) and determining deformation rates and magnitudes along major structures (e.g. Meyer et al. 1998;Yue et al. 2003;Fu and Awata 2007;Yin et al. 2007aYin et al. , 2008Cheng et al. 2015a;Zuza et al. , 2018Wu et al. 2020;Liu et al. 2021). The kinematic evolution of northern Tibet in response to the Cenozoic India-Asia collision has been summarized in two end-member models: (1) a progressive, in-sequence migration of the deformation initiating in the East Kunlun Shan during the late Eocene-Oligocene and reaching the eastern front of the Qilian Shan during the mid-to late Miocene (England and Houseman 1985;Meyer et al. 1998;Tapponnier et al. 2001;Zheng et al. 2017) or (2) an initial, Eocene reactivation of inherited strike-slip structures such as the Altyn Tagh fault, rapidly transmitting the deformation to the northern Qilian Shan, followed by out-of-sequence thrusting in the East Kunlun during the Oligocene and progressive northward propagation and increase of this compression throughout the Miocene Yin et al. 2008;Duvall et al. 2011;Zuza et al. 2019). Many of the major faults in north Tibet, such as the Altyn Tagh, Kunlun or Haiyuan fault, are largely inherited from pre-Cenozoic orogenesis and were reactivated in response to the India-Asia collision (Jolivet et al. 1999Yin and Harrison 2000;Sobel et al. 2001;Cheng et al. 2016;Zuza et al. 2018). However, large differences exist in the topography of the various ranges or even inside a single range. These differences have been poorly addressed although they may reflect major mechanisms in crustal and/or lithospheric deformation processes. For example, in the East Kunlun Shan, the topography of the western part of the range (the Qimen Tagh) is shaped by several narrow, widely spaced subranges whereas further to the east the range is unique, mainly controlled by a frontal thrust and the Kunlun strike-slip fault ( Fig. 1) (Cheng et al. 2014).
In this study, we first compiled available geological, structural and geophysical data to build regional, crustal-scale geological cross-sections. We then used large-scale geomorphological analysis, including cross-range topographic profiles, river-bed longitudinal profiles and slope maps, to characterize the various topographic compartments in northern Tibet. Comparing the topography with the structural information and long-term exhumation rates derived from low-temperature thermochronology data we discuss the topographic compartments in terms of deformation mode. Our analysis suggests that three deformation modes, block uplift, distributed shortening and crustal buckling, are active in northern Tibet. We finally propose that a new crustal structure may develop parallel to the Altyn Tagh fault, running from the eastern tip of the Qimen Tagh across the Qaidam Basin and the Qilian Shan to finally merge with the Haiyuan fault. This result suggests that although most of the present-day major strike-slip faults in northern Tibet are inherited, new crustal-scale structures may be initiating to accommodate the continuously increasing compressive strain.
Geological setting and regional geological cross-sections A number of large-scale cross-sections have been drawn across northern Tibet (Meyer et al. 1998), the Qaidam Basin (Yin et al. 2007a(Yin et al. , 2008Cheng et al. 2016Cheng et al. , 2017 or the various ranges of northern Tibet (Wittlinger et al. 1998;Jolivet et al. 2003;Craddock et al. 2014;L. Li et al. 2021;L. Wang et al. 2021). However, only few have addressed the integrated regional crustal-scale structure and the relation between the various basins and ranges (Meyer et al. 1998;Yin et al. 2007aYin et al. , 2008Craddock et al. 2014). To propose a crustal-scale Cenozoic deformation model for northern Tibet we therefore compiled the geological, structural and geophysical data available for the various ranges into three representative regional cross-sections (Fig. 2).

Altun Shan range
The Altun Shan range forms the northern limit of the Tibetan plateau, separating the Tarim Basin from the Qaidam Basin (Figs 1,2a and 3). Except in the intra-mountain Xorkol Basin, which forms the central part of the range, the exposed lithologies mainly consist of Archean, Proterozoic and Paleozoic igneous and metamorphic rocks (Jolivet et al. 1999;Chen et al. 2003;Cowgill et al. 2003;Gehrels et al. 2003;Mattinson et al. 2007;Wang et al. 2013;Zhang et al. 2014). These include Early Paleozoic ophiolitic series (corresponding to a mid-Silurian-mid-Devonian suture zone; Sobel and Arnaud 1999) exposed along the Altyn Tagh fault, which limits the range to the south. The nature of the 150 km long triangularshaped Xorkol Basin (Fig. 1a) is still debated: it has been interpreted as a pull-apart basin (Guo et al. 1998), a fragment of the Qaidam Basin (Chen et al. 2004) or a Cenozoic transtensive basin accommodating oroclinal bending along the Altyn Tagh fault system (Wang et al. 2008). The basin is filled with Paleogene to Quaternary clastic deposits, sometimes unconformably on Jurassic series (Ritts et al. 2004;, E. Wang et al. 2008Li et al. 2018). North and west of the Xorkol Basin, large patches of Quaternary deposits cover the basement.
The deformation and topographic growth of the Altun Shan is driven both by the sinistral strike-slip motion along the major Altyn Tagh fault and by lithospheric subduction (Fig. 2a). However, the subduction polarity is debated, two opposing models proposing either southward subduction of the Tarim lithosphere beneath the Qaidam and Kunlun blocks (Wittlinger et al. 1998;Jiang et al. 2004;Zhao et al. 2006;Zhang et al. 2015) or northward subduction of the Kunlun lithosphere below the Tarim block lithosphere . The proposed onset time of movement along the Altyn Tagh fault is still highly debated from Eocene to Miocene and the suggested amount of offset varies between 90 and 1200 km with a major consensus between 300 and 400 km (Tapponnier et al. 1986;Ritts and Biffi 2000;Yin and Harrison 2000;Jolivet et al. 2001;Gehrels et al. 2003;Darby et al. 2005;Yin et al. 2008;Searle et al. 2011;Cheng et al. 2015aCheng et al. , 2016. This motion accommodated the northeastward displacement of the Qaidam block and the correlative shortening in the Qilian Shan Yin et al. 2002Yin et al. , 2008Cheng et al. 2015aCheng et al. , 2016Cheng et al. , 2021b.

East Kunlun Shan range
Separating the Qaidam Basin from the Bayan Har terrane to the south (Fig. 1), the East Kunlun Shan is composed of Proterozoic to Paleozoic igneous and metamorphic basement overlain by Triassic and Jurassic sediments (Ministry of Geology and Natural Resources 1980;Harris et al. 1988;Mock et al. 1999) (Figs 2b, c and 4). The basement also contains ophiolitic mélanges and suture zone materials, including the Permian-Triassic Kunlun-Anyemaqen suture, which corresponds to the closure of the Palaeo-Tethys ocean (Konstantinovskaya et al. 2003;Wu et al. 2016Wu et al. , 2019. Similarly to the Altun Shan, the deformation and topographic growth of the East Kunlun Shan is controlled by both the sinistral strike-slip motion along the Kunlun fault and the southward subduction of the Qaidam lithosphere under the Kunlun-Bayan Har blocks (Meyer et al. 1998;Yin and Harrison 2000;Tapponnier et al. 2001), at least in the western part of the range (Fig. 2c). The initiation age of deformation and uplift in the East Kunlun Range is still not fully constrained, with estimations ranging from Eocene to Miocene (Mock et al. 1999;Jolivet et al. 2001;Clark et al. 2010;Duvall et al. 2013). Although probably limited in altitude (L. , some pre-Paleocene relief is inferred in this region although the Qaidam Basin was still extending south, connecting to the Hoh Xil Basin on the Bayan Har terrane ( Fig. 1) (Cheng et al. 2015b;Staisch et al. 2020). Tectonic uplift increased during the Oligocene and again during the Miocene Duvall et al. 2013;Cheng et al. 2015b;Li et al. 2020). The total horizontal Cenozoic displacement along the central Kunlun fault is estimated at 100 ± 20 km (Kidd and Molnar 1988;Fu and Awata 2007). The structure of the range and the associated topography is divided into three distinct domains: east of 94°E, the range is narrow (about 75 km wide), mainly controlled by the strike-slip Kunlun fault along its southern limit and the South Qaidam fault, a frontal thrust separating the range from the Qaidam Basin. This geometry, associated with the very limited width of the range, suggests that southward subduction of the Qaidam lithosphere underneath the easternmost part of the East Kunlun range is still incipient and very limited (Fig. 2b). This assumption is further supported by wideangle seismic data, some researchers even suggesting northward motion of the East Kunlun lower crust below the Qaidam block (Karplus et al. 2011(Karplus et al. , 2019. To the west, the Qimen Tagh ranges spread over more than 200 km in width, and are formed by a series of arcuate NW-SE subranges developing along transpressive faults (Figs 1a and 4a). The vergence of these faults has been widely discussed (Meyer et al. 1998;Jolivet et al. 2001;Yin et al. 2007aYin et al. , 2008 and the debate remains open. Based on the interpretation of satellite images, Meyer et al. (1998) proposed that the Qimen Tagh range developed as a north-verging accretionary prism allowing the southward subduction of the Qaidam crust below the Kunlun-Bayan Har region (Fig. 1). Based on field mapping and seismic profile analysis, Yin et al. (2007aYin et al. ( , 2008 proposed that the Qimen Tagh is a southverging accretionary prism formed by wedging of the Kunlun crust between the Qaidam crust and the southward subducting Qaidam lithosphere. However, a recent model based on basin analysis, seismic data and analysis of satellite images proposed that the main Qimen Tagh faults are mostly north-verging, initiating as left-lateral strike-slip structures forming the western termination of the Kunlun fault (Cheng et al. 2014) (Fig. 2b). Miocene to present-day northeastward displacement of the Qaidam block along the Altyn Tagh fault induced the progressive migration of the faults, increasing their compressive component. This migration seems to be further attested by the progressive northward propagation of the strike-slip Baiganhu (or Kadzi) fault since the mid-Miocene as a transform fault, subparallel to the Altyn Tagh fault (Wu et al. 2020;Liu et al. 2021) (Fig. 4). Although the exact geometry of the Qimen Tagh ranges still remains to be documented, it corresponds to that of a wide basement-involved accretionary prism and suggests a well- Lithospheric subduction in the Kunlun is absent or very limited, but is more pronounced in the Qilian Shan. The crustal deformation is largely accommodated through crustal buckling between the Kunlun fault and Haiyuan fault. The latest structures are major shear zones superimposed on former suture zones (the Anyemaquen and Central Qilian Shan sutures). The Elashan fault probably corresponds to a Cenozoic structure. The oceanic palaeo-slab below the Qilian Shan is inferred from geophysical data Li et al. 2020). (c) Section from the Qimen Tagh range to the Yumen Basin. The occurrence of volcanism in the western part of East Kunlun, as well as the geometry of the ranges, suggests well-developed lithospheric subduction of the Qaidam Block below Tibet. A mid-crustal detachment, possibly developing along the fragile to ductile crust interface. propagates the deformation to the NE. The Danghe Nanshan thrust, superimposed on the Central Qilian Shan suture zone, limits this propagation and the Central Qilian Shan Block forms a backstop to a second crustal prism developing toward the North China Block. The oceanic palaeo-slab is poorly documented and may represent a continuation, toward the north, of the slab observed further south. developed southward subduction of the Qaidam block lithosphere underneath the Kunlun-Bayan Har blocks (Figs 1a and 2c). Such a subduction may explain the formation of late Cenozoic crustalderived magmas (shoshonites and rhyolite) on the northern edge of the Tibetan plateau in this region (Arnaud et al. 1992;Turner et al. 1996;Hacker et al. 2000;Jolivet et al. 2003). Some of these magmas include high-pressure granulite xenoliths indicating a sampling depth of 50-60 km (14-17 kbar, 1000-1100°C) (Jolivet et al. 2003). This volcanism does not occur east of the Qimen Tagh region, further supporting the idea that continental subduction to the east is very limited.
Finally, west of the Baiganhu fault, the sigmoidal-shaped Jianxia Shan thrust belt is formed by a series of stacked thrust faults. These accommodate both the strike-slip motion along the Altyn Tagh fault and NE-SW regional shortening (Wu et al. 2020) (Fig. 4).

Qilian Shan range
On the northeastern margin of the Tibetan plateau, the Qilian Shan is formed by a series of parallel sub-ranges extending over nearly 400 km in width between the Qaidam Basin and the North China craton (Figs 2b, c and 5). The crustal structure of the range was inherited from a long and complex geodynamic history. The Proterozoic basement is affected by several Early to Mid-Paleozoic suture zones marked by UHP rocks such as the North Qaidam eclogites at the limit between the Qaidam and Qilian Shan (Yang et al. 2000;Song et al. 2003;Yin et al. 2007b;Menold et al. 2009;Chen et al. 2012) or the North Qilian ophiolite separating the North Qilian Shan from the North China Block to the north (Song et al. 2009). Geophysical data indicate a significant downward inflection of the Moho below the range (Meng and Cui 1997) and crustal-scale thrusts are inferred on the eastern front of the range (Wu et al. 1996).
Recently acquired magnetotelluric and deep-seismic reflection data suggest that a south-verging (possibly bivergent) oceanic slab is preserved below the range, at least in the southern region Huang et al. 2021;B. Li et al. 2021) (Fig. 2b). The presence of such a slab further north as suggested in Figure 2c remains highly speculative although geometrically valid. Early Paleozoic metasedimentary series are extensively exposed throughout the range, unconformably covered by Permian and Triassic detrital series in the central region between Lake Hala and Lake Qinghai (Fig. 5). Early to Mid-Jurassic sinistral strike-slip movement along the Altyn Tagh fault induced extension in the northern part of the Qilian Shan and transpression further away from the fault, creating depocentres along the eastern margin of the Qaidam Basin and in the Hexi Corridor (Vincent and Allen 1999). Rotation in the regional stress-field during  the Late Jurassic-Early Cretaceous subsequently inverted these basins, creating topographic uplift (George et al. 2001;Sobel et al. 2001;Pan et al. 2013;Cheng et al. 2018Cheng et al. , 2019a. Cenozoic deformation is considered to have initiated in the South Qilian Shan during the Paleogene (Yin et al. 2008;He et al. 2017;Wang et al. 2017;Cheng et al. 2019a) before propagating northward until reaching the Northern Qilian Shan front during the mid-to late Miocene Bovet et al. 2009;Zheng et al. 2010;Li et al. 2019). In the northern part of the range, contraction along thrust faults, associated with strike-slip motion along the Altyn Tagh fault, accommodates the northeastward motion of the Qaidam Basin (Meyer et al. 1998;Cheng et al. 2015a; (Fig. 2c). The southeastern region of the range is further characterized by a number of large strike-slip faults such as the Haiyuan, Riyueshan and Elashan faults (Fig. 5), which accommodate the regional shortening through eastward displacement of crustal blocks (Cheng et al. 2015a(Cheng et al. , 2021b. Incipient lithospheric subduction of the North China craton southward along the northern margin of the range has long been inferred (Meyer et al. 1998;Tapponnier et al. 2001;Yin et al. 2008), with recent support from geophysical studies (Ye et al. 2015;Huang et al. 2021).

Methods
The aim of this study is to describe the mode of crustal deformation in northern Tibet using tectonic, geomorphological and thermal pattern indicators.

Crustal deformation modes and general geomorphological response
In general, crustal deformation in compressional setting can be described by three main models: distributed shortening, block uplift and crustal buckling (Fig. 6). Distributed shortening implies deformation along a series of compressive, transpressive or strikeslip faults (e.g. Meyer et al. 1998). This first deformation mode allows a large amount of convergence to be accommodated. The associated topography displays highly heterogeneous relief and exhumation, localized within narrow regions directly linked to the faults. The main rivers are generally parallel to the uplifting subranges, fed by secondary drainage perpendicular to the ridges (Beaumont et al. 2001;Burbank and Anderson 2001). One exception is the major rivers pre-existing the newly formed topography that may cut across the uplifting ranges if their incision rate is high enough to compensate the uplift rate. In intracontinental collisional settings, and in the absence of an underlying magmatic plume, homogeneous block uplift leading to plateau formation is produced by diffuse deformation (e.g. Dewey and Bird 1970). Erosion leading to strong incision but limited exhumation is initially confined to the edges of the block whereas internal surfaces are preserved as plateaux Vassallo et al. 2007). Incision then migrates inside the uplifted block through time. Crustal buckling forms through layer-parallel shortening (Turcotte and Schubert 1982) and leads to the development of antiformal basement domes associated with subsiding synformal basins (Burg et al. 1994;Burg and Podladchikov 2000). Relief and exhumation in the incipient dome are limited. Rivers flow radially from the highest part of the dome with generally limited incision. Exhumation is limited to the highest part of the dome. However, depending on the thermal state and rheology of the lithosphere, when shortening increases, doming will be replaced by thrusting, with major thrusts developing on each limb of the initial fold (Burg and Schmalholz 2008). Continuing deformation may lead to exhumation of the deep crust such as in the Damba dome on the southern termination of the Longmen Shan range (eastern Tibet) (Weller et al. 2013;Jolivet et al. 2015).

Apatite fission-track thermochronology as indicator of amount and age of denudation
Apatite fission-track (AFT) thermochronology (Hurford and Green 1983;Laslett et al. 1987;Gallagher 1995) has been widely used in northern Tibet to date the initiation of the deformation and evaluate exhumation rates on specific structures (e.g. Jolivet et al. 1999Jolivet et al. , 2001Jolivet et al. , 2003George et al. 2001;Sobel et al. 2001;Duvall et al. 2013;Wu et al. 2020). However, the temperature range in which the method applies (about 120-60°C, corresponding to a depth range of about 3-4 km with a geothermal gradient of c. 30°C km −1 ) does not allow a detailed description of the evolution of the surface topography and relief (Ehlers and Farley 2003). Furthermore, for the same reason, the method poorly constrains horizontal deformation that usually does not generate exhumation. Only exhumation in large restraining bends or satellite thrusts associated with strike-slip faults with a large displacement can be described. Nonetheless, AFT thermochronology provides highly valuable constraints on the age of the major exhumation events and, when enough data are available on a large area, of the regional exhumation pattern. Below we compile the mean basement rock AFT ages available from the literature. The mean age integrates the whole thermal history of a sample during its crossing of the 120-60°C temperature range (e.g. Laslett et al. 1987;Green et al. 1989) and we therefore used these data to characterize the large-scale exhumation pattern of northern Tibet (Fig. 1b) (Jolivet 2017).
Slope mapping, regional topographic profiles and river-bed longitudinal profiles To distinguish between high and low topographic gradient areas, topographic slopes were calculated (Horn 1981) using the Shuttle Radar Topography Mission (SRTM) 30 m digital elevation model (DEM) and the Generic Mapping Tools (GMT) software (Smith and Wessel 1990;Wessel and Smith 1991). In this study, slope maps are especially used to reveal fragments of pre-Cenozoic planation surfaces preserved within the various ranges (e.g. Craddock et al. 2014;Morin et al. 2019) and to distinguish areas of strongly compressive deformation (Clark et al. 2005(Clark et al. , 2006 (Fig. 1b). The pre-Cenozoic planation surfaces were defined as low-slope areas, generally below 10°(indicative of low relief ) developed in crystalline or metamorphic basement or pre-Cenozoic sedimentary series. Following this method, the planation can be related to either erosion (when developed in basement) or sedimentation, although the surfaces always correspond to pre-Cenozoic flat areas, undeformed during the Cenozoic orogeny (Morin et al. 2019). Using the same approach, strongly compressive areas are defined as high-slope regions (generally above 40°).
To describe the large-scale topographic evolution along-strike of the main ranges, topographic swath profiles were also drawn based on the SRTM data and using the GMT software. To allow estimation of lateral topographic variations in a given section of the ranges, each profile is composed of stacked individual parallel topographic profiles acquired every kilometre over a 20 km wide region with a 20 m elevation sampling frequency. To evaluate the impact of the stacking width on the profiles, similar profiles have been drawn along the same axis but with a 10 km stacking width (Supplementary material Fig. S1). The mean profiles do not differ significantly and only the dispersion envelope is reduced. Finally, river-bed longitudinal profiles were acquired from the SRTM DEM on major or characteristic rivers from each range. Both detachmentlimited rivers (where incision rate is controlled by the mechanics of river incision into bedrock) and transport-limited rivers (rivers showing incision or aggradation at a rate controlled by the sediment discharge) (Whipple and Tucker 2002) are represented. The shape of such profiles provides information on the equilibrium state of the rivers (Leopold and Maddock 1953;Whipple 2004;Peyron and Royden 2013;Wickert and Schildgen 2019). This equilibrium depends on multiple parameters including bedrock lithology, water and clastic material input, and channel slope. In this study we used river profiles as a proxy for tectonic deformation (river slope, occurrence of knickpoints) and relative incision rates. The general structure of the drainage system also provides information on the large-scale topographic evolution pattern.
Finally, combining all the above-mentioned information should allow estimation of the equilibrium state of the topography, which provides indications on the style and rate of deformation. This equilibrium state is often categorized as steady-state and nonsteady-state topography. A steady-state topography is defined by a landscape that maintains a stable shape and in which erosion compensates rock uplift (Kooi and Beaumont 1996;Beaumont et al. 2001;Burbank and Anderson 2001;Willett and Brandon 2002). In the framework of northern Tibet this generally corresponds to highly dissected topographies. In contrast, regions where remnants of flat erosion surfaces are preserved in the ranges and that are deeply incised by rivers will be considered as nonsteady-state topographies (Supplementary material Fig. S2). However, the structural pattern, kinematic setting, topographic heritage and long-term climate evolution of northern Tibet most probably prevent a near-steady-state topography being reached, even locally. Liu-Zeng et al. (2008) defined the concepts of 'positive' topography (a region showing a positive correlation between relief and altitude and local slope and altitude) and 'negative' topography (showing an inverse correlation between relief and slope). In this model, a strongly positive topography will correspond to rapidly deforming, fast-growing ranges (the distributed shortening mode) whereas a negative topography will correspond to an uplifted but poorly deformed plateau surface (the block uplift model). However, the crustal buckling mode defined above does not fit into these two end-members as it implies low slopes and low relief at all elevations (Fig. 6). Clark et al. (2006), working on Eastern Tibet, defined an 'active' landscape as areas affected by deep river incision leading to increased erosion rates and sediment export. They opposed this to a 'relict landscape' formed by poorly incised surfaces generally situated in the upper part of the drainage system of major rivers. In the following discussion, we will use the term 'positive' topography as defined by Liu-Zeng to characterize the areas affected by deep incision and devoid of preserved palaeo-surface remnants, and the term 'relict landscape' to describe the topography of areas showing extensive preservation of palaeo-surfaces. It should be noted that the drainage system in these 'relict landscape' may not be connected to a major fluvial system as in the initial definition of Clark et al. (2005Clark et al. ( , 2006.

Low-temperature thermochronology
Apatite fission-track data are summarized in Figure 1b. In the Altun Shan, AFT data indicate strong late Miocene exhumation in the western part of the range (mean AFT ages below 10 Ma) (Jolivet et al. 1999Li et al. 2015;Wu et al. 2020). However, data are available only along the North Altyn Tagh thrust fault, which controls the uplift of the frontal part of the range (Fig. 1). No data exist inside the massif. Exhumation remains stronger along the northern edge of the range but decreases eastward, with the poorly exhumed eastern part recording Mesozoic mean AFT ages (Jolivet et al. 1999Sobel et al. 2001;Li et al. 2015). Wang et al. (2008) reported a 5.1 Ma age from the planation surface west of the Xorkol Basin but no precise analytical data were given and the sample location was close to a large active fault that could generate localized exhumation. To the east of the Jinyan Shan AFT ages are all Mesozoic, including an Early Jurassic age of 167 ± 15 Ma on the northern front of the range .
In the East Kunlun Shan, AFT data record highly diachronous exhumation patterns (Jolivet et al. , 2003F. Wang et al. 2004Yuan et al. 2006;Duvall et al. 2013;Feng et al. 2017;Liu et al. 2017Liu et al. , 2021Y. Wang et al. 2018;Staisch et al. 2020;Tian et al. 2020;Wu et al. 2020). Mid-to Late Miocene ages indicate strong exhumation in the Jianxia Shan (Figs 1 and 5). In contrast, immediately east of the Baiganhu fault (Fig. 5), in the Qimen Tagh ranges the widespread occurrence of Mesozoic to early Cenozoic ages indicates very limited exhumation. Miocene ages are recorded only along the main Kunlun fault and very locally on some of the major thrust faults (Jolivet et al. , 2003Liu et al. 2017). Further east, the Eastern Kunlun Shan displays Mesozoic to early Cenozoic AFT ages associated with Oligocene to Pliocene ages along the major faults, indicating that exhumation is not distributed but is strongly localized along the major structures.
In the Qilian Shan, AFT data reveal an exhumation pattern largely controlled by the main thrust faults (Qi et al. 2016;Li et al. 2019) or very locally along a major strike-slip fault such as the Haiyuan fault (Yu et al. 2019;Wu et al. 2021). The distribution of the AFT ages is similar to that of the Altyn Tagh and Kunlun, with a large number of Mesozoic (mainly Cretaceous) ages associated, near the faults, with Oligocene and Miocene ages. However, the young, late Miocene to Pliocene ages observed in the western Altun Shan and central East Kunlun Shan are absent, suggesting low exhumation. However, it should be noted that a large region between Lake Hala and the Altyn Tagh fault (including some of the highest summits) as well as some areas east of Lake Qinghai are not documented. The AFT ages tend to increase in the southeastward part of the range, where only few Miocene ages are reported (Lease et al. 2011;Lu et al. 2012) among over-represented Mesozoic and Paleogene ages.

The Altun Shan
The range is arc-shaped, convex toward the Tarim Basin and limited by the Altyn Tagh fault to the south (Fig. 1a). Several large transpressive faults near-parallel to the Altyn Tagh fault divide the range into SW-NE elongated compartments (Fig. 2). The highest summits (up to about 6240 m) are situated in the western half of the range, along restraining bends directly linked to shearing along the Altyn Tagh fault. The western end of the Altun Shan range shows a high-relief, strongly dissected topography. A topographic profile perpendicular to the range in the westernmost area ( profile Topo 1, Fig. 7) suggests that, although dissected, the morphology north of the Altyn Tagh fault corresponds to a surface tilted northward toward the Tarim and offset by at least two major faults (F1 and F2 in Figs 3 and 7). The fault that shows the largest vertical offset (F1) shows a southward normal topographic throw of about 1000 m. However, this offset may be overestimated as it is calculated using the mean value of the DEM elevation data, which are very poorly resolved in this part of the range. River beds are narrow and largely free of clastic material (bedrock rivers). Although numerous no-data pixels are present in the SRTM elevation data, the river profiles in this region are concave without major knickpoint, suggesting that trunk channels are at near-equilibrium (Leopold and Maddock 1953;Goldrick and Bishop 2007;Robl et al. 2017) ( profiles R1 and R2, Fig. 7). However, we acknowledge that existing knickpoints may be hidden by the strong pixel noise. The deep incision of narrow gorges and the generally very high slopes (Fig. 1) characterize a region of positive topography associated with strong deformation. However, it should be noted that the concavity decreases eastward from profile R1 to R2, forecasting a lateral change in topographic balance.
The magnitude of river incision decreases progressively eastward until the topography becomes characterized by a wide planation surface worn into the Proterozoic and Paleozoic crystalline basement. This erosion surface is also affected by faults but the major throw observed to the west is not present ( profile Topo 2, Fig. 7). The tilt of the surface is generally limited to a few degrees, the northern compartment, covered by Quaternary deposits (Liu 1988), being horizontal. Only the northernmost part of the range shows strong slopes and incisions linked to uplift along the North Altyn Tagh thrust fault (Fig. 3). Overall incision is limited to a few rivers flowing perpendicular to the range. In the southern part of the range, satellite images show that rivers are gravel-bedded with wide river beds. However, in the central (on the horizontal plateau) and northern areas, they form deep gorges with little detrital material preserved in the river bed. The trunk channels have a generally lowslope profile with two obvious knickpoints associated with major faults (profile R3, Fig. 7). The slope of the river bed decreases at the onset of the gorges (see river segment below point K2 on profile R3, Fig. 7). The topography of this part of the Altun Shan range thus appears to correspond largely to the relict landscape type, erosion being largely outpaced by tectonic uplift.
The central part of the range is mainly characterized by the transtensive Xorkol Basin (Fig. 3). To the south, the Akatengmeng Shan massif, although affected by a relay zone between two segments of the Altyn Tagh fault, displays a generally flat northdipping topography ( profile Topo 3, Fig. 7). North of the Xorkol Basin, the Jinyan Shan massif (Fig. 3) again shows a flat mean topography, tilted to the north and split into two compartments by a major fault (F3 in Figs 3 and 7) that has a mean southward normal topographic throw of about 500 m. These offsets are consistent with the occurrence of large normal faults bordering the southern and northern edges of the Xorkol Basin (Wang et al. 2008). Fragments of planation surfaces are preserved especially in the eastern part of Fig. 7. Topographic swath profiles and longitudinal river profiles across the Altun Shan Range obtained from SRTM-1 data. Location is shown in Figure 3. On profiles Topo 1 and Topo 2, the very low altitude values (reaching 0 m) are linked to gaps in the SRTM data. ATF, Altyn Tagh fault; Ak., Aka., Akato Tagh, Akatengmeng Shan; Jin., Jinyan Shan; K1 and K2: knickpoints. F1, F2 and F3 refer to major faults (see Fig. 3 for location). NAT-FT, North Altyn Tagh thrust fault. the massif, supporting Eocene sediments, unconformably covered by Quaternary deposits (Liu 1988;Li et al. 2018). In this part of the Altun Shan range, the water divide is located along the northern edge of the Xorkol Basin, and no river crosses the entire range. The rivers flowing northward across the Jinyan Shan are gravel-bedded with generally wide beds and display a very regular, slightly concave, profile ( profile R4, Fig. 7). Being more complex owing to the occurrence of the Xorkol Basin, this section of the Altun Shan corresponds mainly to a relict landscape associated with some local areas of positive topography along major faults (Fig. 1).
Finally, to the extreme east of the Altun Shan, the topography is complex and no individual surfaces can be clearly distinguished. The topography has a general dome-like shape divided into three parts by two valleys ( profile Topo 4, Fig. 7). Incision is distributed among many small drainages, controlled by the numerous en échelon thrust faults that merge with the main Altyn Tagh fault. Rivers flowing north have a regular, slightly concave profile similar to those of the Jinyan Shan ( profile R5, Fig. 7). This topography is again largely associated with the relict landscape type.

The East Kunlun Shan
In the eastern part of the East Kunlun range, river incision is important and, except for those flowing along the Kunlun fault valley, most rivers flow northward, perpendicular to the range (Fig. 4). Transverse topographic profiles display a steep frontal relief along the thrust fault separating the Qaidam Basin from the range ( profiles Topo 1 and Topo 2, Fig. 8). This initial step in the topography is followed by a topographic compartment with altitudes varying between about 4000 m and slightly less than 5000 m, showing several large fragments of flat surfaces at altitudes above 4500 m and mainly worn into Permian-Triassic granites or Proterozoic gneisses   (Fig. 4). Some of these surfaces display deep incisions although the rivers mostly remain gravel-bedded. The Nuomuhong river (river R1, Fig. 8), flowing perpendicular to the range immediately east of one of the major planation surfaces, has a slightly concave, relatively steep profile with a potential knickpoint in its upper part. Further west, the Golmud river (river R2, Fig. 8), which flows parallel to the range before crossing it, has a regular, low-slope profile, slightly convex in its upper part, without knickpoints. The low-slope profile suggests a limited sediment-to-water discharge ratio (Wickert and Schildgen 2019). The river incises in its own alluvium along most of its length, which might be related to an increase in water discharge (Poisson and Avouac 2004), a drop in base level owing to Late Pleistocene drying of the Qarhan palaeo-lake system forming the base level in the Qaidam Basin (An et al. 2012) or an increase in tectonic uplift rates, or to a combination of these. Finally, river R3 (Fig. 8) flows perpendicular to the range in a poorly incised, lowslope surface at the limit between the eastern East Kunlun range and the Qimen Tagh. The river profile is similar to that of the Nuomuhong river, with at least one marked knickpoint in its upper reaches and probably a second downstream. Based on satellite images, the river bed seems to contain less sediment than the other rivers as bedrock is visible at least in the section between the two knickpoints. Both knickpoints correspond to faults affecting the surface. All three river profiles appear out of equilibrium. In general, this topography can be categorized as a relict landscape: tectonic uplift seems to outpace erosion as attested by knickpoints, Mesozoic to Paleogene AFT ages (Fig. 1b) and ancient erosion surfaces largely preserved at high altitude. However, the deep incision and higher exhumation in the frontal part of the range and along the major strike-slip faults as attested by late Cenozoic AFT ages indicate a transition toward active topography.
The topography of the Qimen Tagh is completely different (Fig. 4). As indicated above, the range widens rapidly, the topography including narrow thrust-controlled ranges separated by wide basins filled with Eocene to Quaternary clastic sediments to the east and mostly Neogene to Quaternary deposits to the west (Liu 1988). These basins host lakes in their western part, which form local base levels for the rivers, making for a complex series of nonconnected range-parallel drainage sub-systems. All the main rivers flow NW, suggesting a general slope in that direction, fitting with the presence of the lakes to the west. River profile R4 (Fig. 8) represents a characteristic example of one of these rivers although it crosses the northern sub-range to connect to Lake Gaskule in the SW Qaidam Basin (Fig. 4). From observation of satellite images, the river bed is wide, full of sediment and shows only a very limited incision. The longitudinal profile is regular with a relatively low slope. These two observations suggest limited tectonic disturbance of the river profile, even on crossing the frontal range, and a low sediment-to-water discharge ratio (Wickert and Schildgen 2019). The topographic profiles across the range indicate a progressive decrease of relief from east to west ( profiles Topo 3 to Topo 5, Fig. 8), consistent with the general low exhumation indicated by the AFT data. Part of that decrease in relief is linked to sedimentation in the multiple isolated basins. All these observations complemented by the Mesozoic and Paleogene AFT ages (Fig. 1b) suggest a relict landscape type topography associated with localized positive topography along the major thrust faults.
Finally, the Jianxia Shan displays a highly dissected, positive topography very similar to that of the westernmost part of the Altun Shan (Fig. 4). Most rivers are bedrock channels. This is consistent with the recent and strong exhumation suggested by the Mid-to Late Miocene AFT ages reported in this region (Wu et al. 2020). Groves et al. (2020) recently investigated the large-scale geomorphology of the Qilian Shan and divided the range into five domains based on the distribution of geomorphological indices. Below we have simplified this division to three main zones (Fig. 5). From the Altyn Tagh fault to Lake Hala, the topography is composed of a series of narrow, high ranges separated by sedimentary basins ( profile Topo 1, Fig. 9). The width of these ranges increases eastward and their altitude increases toward the axial part of the range before decreasing again eastward. The relief of the sub-ranges is steep with strong incision (Fig. 1b). However, some limited-relief basement surfaces are preserved, especially a large one in the Sule Nan Shan (Figs 5 and 9). The main rivers that flow in this part of the range are gravel-bedded drainages, largely parallel to the ranges, similar to those of the Qimen Tagh. All the main rivers flow north toward the Altyn Tagh fault, suggesting again a general slope in this direction. Except for those flowing into Lake Suhai, which forms a local base level in the Suganhu Basin, the rivers cross the Altyn Tagh fault to reach the regional base level formed by the much lower Tarim Basin. The Danghewan River (river R1, Fig. 9) flows from the southern summits of the Sule Nan Shan along the eastern side of the Tang He Nan Shan before crossing the Altyn Tagh fault. The upper part of the river profile is marked by a typical concave shape with possibly a first knickpoint on reaching the intra-mountain basin. The lower part of the profile is characterized by two well-marked knickpoints on crossing the ranges associated with the Altyn Tagh fault. Except for some limited bedrock-confined segments, the Changma River (river R2, Fig. 9) is also gravel-bedded. Sourced on a mountain ridge separating Lake Hala basin from Lake Qinghai drainage area, it flows inside the Ta Shue Shan sub-range, parallel to the range axis, before crossing it and the northern tip of the Qilian Shan sub-range northward to reach the eastern tip of the Altyn Tagh fault. On reaching the Tarim Basin it forms a huge alluvial fan. The river profile is flat, slightly convex in its upper-central part, suggesting that rock uplift in the Ta Shue Shan (Fig. 5) is not equilibrated by river incision. The fact that the river crosses the Ta Shue Shan and Qilian Shan sub-ranges instead of being guided along-strike suggests that it was antecedent to the ranges and was able to incise fast enough to compensate uplift. No knickpoint is observed on crossing the Altyn Tagh fault, indicating that river incision largely compensates the tectonic uplift. The topography of this region is categorized as positive topography although it is largely smoothed by sediment deposits between the main sub-ranges (Liu-Zeng et al. 2008). However, it is interesting to note that although the topography has been largely reworked by recent tectonics, part of the drainage system seems to be inherited from a previous landscape.

The Qilian Shan
A topographic profile immediately north of Lake Hala displays a general dome-like shape of the range, the lake being situated on the top of the dome ( profile Topo 2, Figs 5 and 8). The range is still formed by a series of parallel sub-ranges but their relief is lower than to the north, except along the eastern front of the range, which is marked by a sharp 3000 m drop in altitude on reaching the Hexi Corridor (Figs 1a and 5). Large fragments of low-relief erosion surfaces are exposed in this region, including a huge surface between Lake Hala and the Qaidam Basin (Fig. 5b). South of Lake Hala, in the northern part of Lake Qinghai drainage area, the width of the range decreases and the dome-like shape is well expressed in the central part, bordered by the high relief of the Qilian Shan to the east and the Delingha Basin (Fig. 5) to the west ( profile Topo 3, Fig. 9). A major part of the topography in the central part of this region is formed by poorly incised surfaces (Fig. 5c). It includes the most extensive, flat-lying Permian and Triassic outcrops within the whole range. AFT mean ages are mostly Late Cretaceous to Paleocene in the central region whereas available Neogene ages are restricted to the front of the Qinghai Nan Shan and to the vicinity of the Haiyuan fault (Lu et al. 2012;Qi et al. 2016) (Fig. 1b). This region is the source of many rivers diverging radially from the summit of the dome (including the Changma River described above). The Xiala-Heihe River flows eastward across the Qilian Shan (river R3, Fig. 9) and shows a relatively steep-slope profile with no or poorly marked knickpoints, suggesting a relative equilibrium between uplift and incision (it should be noted that the lower part of the profile is disturbed by the occurrence of a dam lake). The Buha River initiates in the same region but flows the opposite way toward Lake Qinghai with a constant, low-slope profile (river R4, Fig. 9). This reflects the low slopes prevailing in this region as the river mostly flows across the poorly incised surface supporting the Permian-Triassic series. The topography of this region is clearly of relict landscape type, except for the northern front of the range where positive topography dominates.
South of Lake Qinghai, the range narrows again down to about 250 km instead of 350 km north of Lake Hala. The dome-like shape observed in the previous profiles is no longer present ( profile Topo 4, Fig. 9). The topography is formed by a series of sub-ranges separated by narrow valleys and most rivers flow toward the SE. Incision is stronger, including the deep Longyang Gorge formed by the Yellow River (Lin et al. 2021, and references therein). Several low-relief erosion surfaces are preserved in this region but not to the extent observed further north (Craddock et al. 2014;Lin et al. 2021). AFT mean ages are widely spread between Early Jurassic (Lin et al. 2021) and Eocene (Wang et al. 2016) (Fig. 1b). No Neogene ages are reported, indicating a generally low Cenozoic exhumation although exhumation initiated around the Late Miocene along the western front of the range (Craddock et al. 2014) and in the Mid-to Late Miocene along the eastern front . Once again, the topography seems intermediate between a relict landscape type and a newly developing positive topography type.

Discussion
The geomorphological analysis presented above demonstrates that the deformation pattern varies greatly between and within the ranges that form the northern edge of the Tibetan plateau. After discussing and comparing the deformation in each range, we propose a synthetic model showing that the three main modes of deformation described above occur simultaneously in northern Tibet: (1) distributed shortening; (2) homogeneous block uplift; (3) crustal buckling.
Below we first discuss the occurrence of these three types of deformation in northern Tibet. We then show that their distribution suggests the occurrence of a transition zone, traversing the central parts of the East Kunlun range, the Qaidam Basin and the Qilian Shan, parallel to the Altyn Tagh fault.

Deformation pattern within individual ranges
The topography of the Altun Shan range corresponds to an ancient planation surface ( probably of Early Mesozoic age; Jolivet 2017) tilted northward along major normal faults and eastward as a general trend (Figs 1 and 3). The strongly positive topography and Pliocene AFT ages in the western part of the range attest to recent strong deformation and exhumation. However, except in this westernmost part, exhumation is low to very low as evidenced by the widely distributed Mesozoic and early Cenozoic AFT ages (Fig. 1b). The topography is mainly of relict landscape type showing a relief largely in disequilibrium as incision does not compensate uplift. Wang et al. (2008) proposed that the Altyn Tagh range is a restraining bend along the Altyn Tagh fault, with extension in the Xorkol Basin being related to oroclinal bending. However, such a model would require that, before extension takes place in the central part of the growing dome, exhumation should be important as observed in the smaller-scale restraining bends such as the Akato Tagh (Fig. 3) (Cowgill et al. 2004a, b). This is not recorded in the topography, which preserves flat palaeo-surfaces, nor in the lowtemperature thermochronology data. The only area where doming might be suggested is the easternmost part of the range (see profile Topo 4, Fig. 7). We propose that the Altyn Tagh range corresponds to a narrow, basement-involved accretionary prism accommodating the subduction of the Tarim lithosphere beneath the Qaidam and Kunlun blocks (Fig. 2a).
Although they represent the present-day deformation pattern and should be extrapolated back in time with caution, global positioning system (GPS) measurements of crustal motion available in the western part of the range indicate north-directed displacements (relative to a fixed Eurasian Plate reference), at a high angle to the faults (Calais et al. 2006;Gan et al. 2007; Wang and Shen 2019) (Fig. 1a). The general uplift of the range is controlled by the steep frontal thrusts such as the North Altyn Tagh fault that decouple the subducting and shortening Tarim crust from northern Tibet. To the south, the nearly purely strike-slip Altyn Tagh fault acts as a backstop structure to the Altun Shan terrane, decoupling it from the Qaidam and Kunlun blocks. The deformation of the upper crust in the Altyn Tagh range is thus accommodated by thrusting on the frontal North Altyn Tagh fault and transtensive movements leading to the formation of pull-apart basins along southward-verging faults inherited from the Mesozoic transtensional phase (Chen et al. 2003;Cheng et al. 2019a;Zhao et al. 2020). This structural pattern leads to vertical uplift of crustal blocks with only limited shortening to the west (most probably driven by the general sinistral component affecting the range), preserving the initial surface from deformation. The dry climate (less than 250 mm a −1 precipitation; Hong et al. 2005), associated with low slopes and recent onset of uplift, explains the limited incision and the highly unbalanced topography. Only to the east, where, as suggested by the available GPS data, the strike of the faults favours strike-slip motion, may oroclinal bending occur.
In the eastern part of the East Kunlun Shan, the GPS data indicate a nearly 45°angle between the direction of crustal motion and the strike of major faults (Fig. 1a). The range is narrow (about 80 km wide) and the deformation linked to the southward subduction of the Qaidam lithosphere is mainly accommodated by partitioning along the strike-slip Kunlun fault as well as on the near-unique frontal, steeply dipping South Qaidam thrust fault (Figs 2b and 4). As indicated by the occurrence of late Cenozoic AFT ages, exhumation is localized in the restraining bends along the strike-slip fault (Jolivet et al. , 2003) and the major rivers are incising deeply in the range, indicating strong regional rock uplift. However, in general the topography is of relict landscape type and, as attested by the Mesozoic and early Cenozoic AFT ages, regional exhumation remains limited, preserving large fragments of the Mesozoic erosion surface Roger et al. 2011;Duvall et al. 2013;Jolivet 2017;B. Li et al. 2021;). The setting is largely similar to that of the central Altun Shan. To the west, in the Qimen Tagh, the deformation pattern is controlled by strike-slip motion along both the Kunlun and Altyn Tagh faults as demonstrated by Cheng et al. (2014Cheng et al. ( , 2015a. The deformation is highly localized along major, basement-involved, thrust faults that propagate into the SW Qaidam Basin, most probably along a midcrustal décollement level (Meyer et al. 1998) (Fig. 2c). The GPS data indicate a crustal motion near perpendicular to the strike of the faults, suggesting the absence of a strike-slip component on these structures (Fig. 1a). However, seismic imaging analysis and river bed lateral shifts have revealed the occurrence of a noticeable strikeslip component on several of these faults (Cheng et al. 2014(Cheng et al. , 2015b. The Qimen Tagh thrusts are widely spread on a total rangewidth of about 200 km, which makes for an overall limited relief and exhumation as indicated by the low-temperature thermochronology data (Fig. 1b). Because the deformation is localized, subranges are narrow, impeding internal preservation of the Mesozoic planation surface. However, these surfaces are probably preserved between the ranges as suggested by the general low exhumation, implying that the large-scale topography is mostly of relict landscape type but includes positive topography in the narrow ranges.
As recently demonstrated by Allen et al. (2017) and Groves et al. (2020), the topography of the Qilian Shan evolves from SE to NW. South of the southeastern tip of the Qaidam basin, the range is narrower and, as suggested by the AFT data, the topography is largely inherited from a Cretaceous exhumation event (e.g. Lu et al. 2012;B. Li et al. , 2020Lin et al. 2021;Y. Wang et al. 2021) (Figs 1b and 5). In this region, the GPS data indicate crustal motion at a relatively low angle to the strike of the main faults, as attested by the strong strike-slip component along many of these faults. This major horizontal component limits the exhumation to a few restraining bends and thrust faults (Lease et al. 2011;Lin et al. 2021). Between Lake Qinghai and Lake Hala, the GPS displacement data show vectors near perpendicular to the strike of the faults, implying pure shortening (Fig. 1b). The topography largely preserved the Mesozoic surface, with very limited Cenozoic deformation and associated exhumation as attested by the extensive preservation of Permian to Triassic series and the Late Cretaceous to Paleocene AFT mean ages (Figs 1b and 2b). Active thrusts are restricted to the edges of the range and Allen et al. (2017) described this area as similar to the initial Tibetan plateau. The region includes the source of many of the large rivers flowing through the range in a diverging pattern (Fig. 5). Meng et al. (2020) suggested that this river pattern resulted from a Late Miocene complete reorganization of the river network linked to the onset of strike-slip motion along the Haiyuan fault. However, their model suggests that rivers flow parallel to major thrusts. Although this is largely the case in the northern region, it is not true for the central and southern regions. The topographic profiles reported in the present study suggest that the plateau-like topography observed in the central part of the Qilian Shan accommodates deformation through doming (Fig. 9). Lake Hala Basin marks a major limit: to the north, the topography is composed of a series of parallel sub-ranges developing on major thrust faults and separated by wide valleys (Fig. 2c). The uplift of the sub-ranges is largely controlled by strike-slip motion along the Altyn Tagh (Allen et al. 2017), in a similar way to the northward migration of the Qimen Tagh thrusts. However, the Paleozoic suture zones located in the Danghe Nanshan and the Daxue Shan and the contrasted basement-block lithology probably impede the development of a single, continuous crustal décollement level previously suggested by Meyer et al. (1998) (Fig. 2c). Few AFT ages are available inside this part of the Qilian Shan but the observation that some of the main rivers that cut across sub-ranges may be inherited from a previous drainage network suggests that the topography still contains some relict geomorphological elements.
Large-scale deformation pattern in northern Tibet: coexisting deformation modes The three modes of deformation described above, block uplift, distributed shortening and crustal buckling (Fig. 2), appear to coexist in northern Tibet. Below we propose a crustal-scale model of the deformation in northern Tibet, taking into account both the regional structural pattern and the surface morphology (Fig. 10).
Deformation is strongly partitioned across the lithospheric-scale Altyn Tagh fault. In the southern part of the Altun Shan, shortening and incipient subduction of the Tarim crust below the Qaidam block is accommodated by strike-slip NE-SW transpressive and ESE-WNW transtensive faults. However, some of the deformation is also accommodated along the steep frontal North Altyn Tagh thrust fault and, to a limited extent, on the Tanan rise system immediately to the north (Figs 1 and 2a). This structural pattern leads to a nearhomogeneous uplift and northeastward tilting of the Altun Shan block.
Immediately SSE of the Altyn Tagh fault, the Cenozoic NE-SW shortening of northern Tibet is accommodated by a large accretionary prism encompassing the Kunlun, the NW Qaidam Basin and the South Qilian Shan. NW-SE parallel thrusts, formed as en échelon thrusts accompanying the strike-slip motion, develop at a high angle to the main Altyn Tagh fault (Figs 1a and 2c). This angle increases eastward from about 40°in the Qimen Tagh to 55°f or the folds and thrusts in the north Qaidam Basin and 60°in the Qilian Shan. The increasing angle is correlated to a clockwise rotation in the direction of crustal motion as observed from GPS data Wang and Shen 2019;Cheng et al. 2021b) (Fig. 1b). Separating the South Qilian Shan block from the Central Qilian Shan block, the Danghe Nanshan thrust system presents some of the highest slopes, altitudes and denudation rates inside the thrust system ( Fig. 1b and profile Topo 1, Fig. 9). It is a major system associating thrusting and sinistral strike-slip (Meyer et al. 1998). Similarly, the Daxue Shan thrust system is superimposed on the north Qilian Shan suture zone and again displays steep slopes and high altitudes. Finally, the North Qilian Shan sub-belt extending along the whole eastern front of the range displays the strongest observed topographic gradients (Figs 1b and 9) (Allen et al. 2017; Groves et al. 2020;Meng et al. 2020). As already suggested (Meyer et al. 1998;Tapponnier et al. 2001;Cheng et al. 2019b), the Central Qilian Shan block, limited by the two weak zones formed by the Paleozoic suture zones, may act as a backstop to the development of a second accretionary prism propagating eastward in the North China Block in response to the incipient subduction of this block beneath the Central Qilian Shan (Fig. 2c). As mentioned above, the occurrence of a Paleozoic oceanic slab below the Central and North Qilian Shan is highly speculative. However, magnetic anomaly data from the EMAG2v3 grid (calculated at 4 km altitude; Meyer et al. 2017) display a large, long-wavelength NW-SE negative anomaly below the Central and Northern Qilian Shan that might reflect the occurrence of a deep body (Fig. 11) (Kang et al. 2012).
To the west, between the Altyn Tagh fault and the Qimen Tagh, the Jianxia Shan is a fast-exhuming wedge-like structure between the Altyn Tagh fault and the Baiganhu fault. Crustal material is forced into the corner formed by the two faults, leading to rapid exhumation (Wu et al. 2020).
Southeast of a limit running from the eastern tip of the Qimen Tagh to Lake Hala, the GPS data indicate crustal motion near perpendicular to the main tectonic structures. The deformation mode changes to vertical block uplift in the East Kunlun and crustal buckling in the Qilian Shan. Sedimentation and subsidence occur in the SE Qaidam Basin where shortening is accommodated only by a few folds (Yin et al. 2007a(Yin et al. , b, 2008. L.  suggested downward crustal flexure as a subsidence mechanism for this part of the basin. The axial direction of the basin varies from about N130E north of the supposed limit to N100E to the south. It should also be noted that the first Cenozoic deposits in the Qaidam Basin (Lulehe to Lower Xaiganchaigou formations) are largely restricted to the NW part (Yin et al. 2008;Cheng et al. 2018). This suggests that subsidence initiated first in the northern part of the basin (in relation to strike-slip movement on the Altyn Tagh fault) before crustal buckling initiated to the SE. Major thrusting, associated with high relief, is limited to only the eastern edge of the Qilian Shan, between the Haiyuan fault and the Hexi Corridor (Fig. 2b). We suggest that the Haiyuan fault, largely superimposed on the central Qilian Shan Paleozoic suture zone, plays a similar role to the Danghe Nanshan thrust system in segmenting the structural pattern of the range (Figs 2b and 10).
Compiling the initiation ages of the major faults in northern Tibet, Duvall et al. (2013) divided the regional deformation history into two periods: prior to 20 Ma mostly in the Altyn Tagh realm to the NW and after 20 Ma in the region from the East Kunlun to the Haiyuan fault zone as well as, to some extent, in the extreme NE Qilian Shan (see fig. 10 of Duvall et al. 2013). They related this southeastward propagation of the deformation to the initiation of the Haiyuan fault. Although they included the frontal thrust of the East Kunlun Shan, near Golmud, in the pre-20 Ma fault pattern, the limit they suggested between the two sets of faults is similar to the limit defined between the deformation modes in the present study. On the surface, no obvious geological or structural feature directly fits with the proposed limit. However, EMAG2v3 magnetic data display a negative west-east anomaly connecting the eastern tip of the Qimen Tagh range to the large NW-SE negative anomaly in the Lake Hala region (Fig. 11). This limit also corresponds to the limit between the numerous folds and faults accommodating the deformation in the western Qaidam and the few folds observed in the eastern part of the basin (Fig. 1a). This inferred shear zone would be parallel to the Altyn Tagh fault and ends to the east near the northern tip of the Haiyuan fault (Fig. 10). The structure would accommodate the differential movement between the intracontinental subduction of the Qaidam lithosphere in the Qimen Tagh area and the absence of such subduction (collision) in the Eastern Kunlun Shan (Fig. 10). The individualized SE Qaidam-South Qilian Shan block would accommodate deformation through crustal buckling but also southeastward motion along the East Kunlun-Haiyuan fault system and clockwise rotation.
If a strike-slip structure exists across the Qaidam Basin, many questions still remain to be addressed. For example, the deformation in the Qimen Tagh ranges, dated to the early Miocene (Cheng et al. 2014), marks the onset of the intracontinental subduction in this region. The onset of movement on the inferred lithospheric fault might thus be dated to the early-mid-Miocene and have accommodated several tens of kilometres of strike-slip displacement. Such a displacement on a single fault should be expressed at the surface, which is not the case. However, a wide, distributed shear zone may accommodate such an amount of strike-slip movement at depth without obvious surface expression. Similarly, the occurrence of a large detachment fault in the lower to middle crust of the Qaidam Basin as suggested by several researchers (e.g. Meyer et al. 1998;Yin et al. 2007aYin et al. , b, 2008 may distribute some of that deformation at depth.

Conclusions
The large-scale geomorphological analysis presented here, associated with the exhumation pattern derived from apatite fission-track thermochronology, suggests the occurrence of three deformation modes in northern Tibet. The distributed shortening mode observed in the Qimen Tagh, north Qaidam Basin and north Qilian Shan is directly related to strike-slip motion along the Altyn Tagh fault and may represent the oldest, pre-Miocene deformation. To the SE, deformation appears largely accommodated by crustal buckling in the Qilian Shan and southern Qaidam Basin, as well as vertical block uplift in the East Kunlun Shan. Although it is poorly expressed as a geomorphological structure, an incipient shear zone running from the eastern tip of the Qimen Tagh range to the Haiyuan fault may accommodate the differential movement between the intracontinental subduction to the NW and the collision in the Eastern Kunlun Shan. The existence of this inferred structure should be further documented by fieldwork in the Qilian Shan, especially SE of Lake Hala Basin.
Competing interests The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.
Data availability The dataset that support the findings of this study are available from cited references and no new data were generated.  . EMAG2v3 magnetic anomaly map of north Tibet (data from Meyer et al. 2017). The near east-west negative anomaly across the Qaidam Basin may correspond to the developing structure accommodating the differential motion between the NW and SE Qaidam-South Qilian Shan blocks. It merges with the palaeo-slab anomaly between Lake Hala and Lake Qinghai and may reach the western termination of the Haiyuan fault.