The topographic evolution of the Tibetan Region as revealed by palaeontology

Abstract

The Tibetan Plateau was built through a succession of Gondwanan terranes colliding with Asia during the Mesozoic. These accretions produced a complex Paleogene topography of several predominantly east–west trending mountain ranges separated by deep valleys. Despite this piecemeal assembly and resultant complex relief, Tibet has traditionally been thought of as a coherent entity rising as one unit. This has led to the widely used phrase ‘the uplift of the Tibetan Plateau’, which is a false concept borne of simplistic modelling and confounds understanding the complex interactions between topography climate and biodiversity. Here, using the rich palaeontological record of the Tibetan region, we review what is known about the past topography of the Tibetan region using a combination of quantitative isotope and fossil palaeoaltimetric proxies, and present a new synthesis of the orography of Tibet throughout the Paleogene. We show why ‘the uplift of the Tibetan Plateau’ never occurred, and quantify a new pattern of topographic and landscape evolution that contributed to the development of today’s extraordinary Asian biodiversity.

Introduction

The modern Tibetan Plateau (Fig. 1) is the highest and most extensive elevated surface on Earth covering an area of 2,500,000 km²at an average elevation above 4500 m. The boundaries of the plateau extend 1000 km northwards from the Yarlung–Tsangpo suture zone (YTSZ), south of which is the Himalayan thrust belt, to the Altyn Tagh fault. Westwards the plateau boundary is marked by the Karakoram strike–slip fault, while 2000 km to the east the plateau morphs into the Hengduan Mountains and down into Yunnan and Sichuan.

Fig. 1

a Global locator (inset) and orographic map of the Tibetan region showing major bounding features, sutures and major fossiliferous basins (red circles) referred to in the text. Yellow triangles mark the positions of mountain peaks featured in the text. YTSZ—Yarlung-Tsangpo Suture Zone; BNSZ—Bangong–Nujiang Suture Zone; JSZ—Jinsha Suture Zone; KSZ—Kunlun Suture Zone. b Modern topographic profile along the transect A–B shown in (a)

In conjunction with the adjacent Himalaya and Hengduan Mountain systems (Fig. 1), the Tibetan Plateau is often referred to as ‘The Third Pole’ (Qiu 2008) and regarded as the ‘Water Tower’ (Viviroli et al. 2007; Immerzeel et al. 2020) of Asia. This is because its hydrological services, delivered through the Yangtse, Yellow, Red, Mekong, Salween, Tsangpo–Brahmaputra, Indus and Ganges rivers that all originate on or near the plateau, dominate the lives and livelihoods of people throughout most of Asia. By understanding the evolution of The Third Pole, we also obtain a better understanding of significant components of the modern Earth system and how they may change in the future.

For many years, the height and extent of the Tibetan Plateau and surrounding mountains have been regarded as the main drivers of the Asian monsoon, with summer heating of the atmosphere over the plateau creating a deep low-pressure system that draws in moist air from the Indian Ocean, while winter cooling creates a strong high-pressure system reversing the air flow and flooding the surrounding regions with dry air (Flohn 1968; Yanai and Wu 2006). The plateau has therefore been envisioned to act as a sensible heat pump. However, because the location of the greatest heating is not over the plateau, but over NW India and eastern Pakistan, the Himalaya are also regarded as having a key orographic influence on the monsoon (Boos and Kuang 2010; Molnar et al. 2010), although recent modelling shows that monsoon circulation would exist even in the absence of an orographic barrier (Acosta and Huber 2020).

The topographically complex domain of the Tibetan region contributes to, and experiences, a monsoon climate over much of its extent. The Asian monsoon systems also impact several of the world’s great centres of biotic richness, namely the Indo-Burma, the South-Central China and Western Ghats biodiversity ‘hotspots’ (Myers et al. 2000). In addition to the region’s climatic and biotic significance, Tibet, and the geology of the associated mountain systems, provides unique insights into collision tectonics and geodynamic processes. Despite this importance, and even after decades of research, the orographic evolution of the region remains imperfectly known.

Our conceptual framework for the evolution of the Tibetan region has been compromised by commonly held myths embedded in the literature of the geosciences and the wider research community (Spicer et al. 2020a, b). One such myth is exemplified in the commonly used phrase ‘the uplift of the Tibetan Plateau’, which implies that a flat-surfaced (low relief) Tibet rose as a coherent entity. Moreover, it is not unusual to find that the Tibetan Plateau and the Himalaya are aggregated into a single entity, which is unfortunate because this conflation obscures the distinctly different processes underlying their formation. Similarly, the monsoon climates experienced throughout the region represent not one system, but an aggregation of interacting monsoon types, each with its own pattern of evolution and causal drivers, and each with different heterogeneous impacts on biodiversity in time and space (Spicer et al. 2017; Wang et al. 2017a).

In this work, the Himalaya–Tibet–Hengduan Mountains area will be referred to as the ‘Tibetan region’ because using any term relating to the modern topography, and particularly references to a ‘plateau’, prejudices our view of the past. Similarly, the terms ‘Tibetan Plateau’ or ‘Qinghai Tibetan Plateau’ are retained just for the present day. To refer to a ‘plateau’ in the Paleogene is incorrect and conceptually misleading. The area occupied by the present plateau will simply be referred to as ‘Tibet’ without any real or implied topographic or administrative connotations.

Linked to this simple monolithic concept of Tibet is the idea that the uplift took place very recently in the late Neogene (Miocene–Pliocene) (Xu et al. 1973; Wang et al. 2006; Harrison et al. 1992; Molnar et al. 1993; Coleman and Hodges 1995; Zhou et al. 2007; Chang et al. 2010; Sun et al. 2014a, 2015b). Part of the problem here lies in models that treat Tibet and the adjacent mountain systems as a single block, leading to the assumption that surface height determined at a particular location indicates uplift of the whole Tibetan region. This clearly need not be the case, and while individual measurements may (or may not) be valid they only apply to the specific locations where they were made.

Another misconception is that the mechanism for this monolithic rise was purely the collision of India with Eurasia beginning early in the Paleogene (see review by Wang et al. 2014). There can be no doubt that the collision had far-reaching consequences, and is even invoked to explain re-activation of the Central Asian Orogenic Belt from Lake Baikal into Siberia (Molnar and Tapponier 1975; Dobretsov et al. 1996), but not all Tibet’s topography was formed by the India–Eurasia collision. Significant uplands, created by earlier terrane accretions, existed across the Tibetan region before the arrival of India.

In the last few years, improvements in our understanding of regional geology, fossil biotas and, crucially, radiometric dating has transformed our ideas of how ‘The Third Pole’ topography evolved, and that a wide variety of tectonic modes have operated over the past ~ 220 Ma to produce it (e.g. Dewey et al. 1988; Liu et al. 2016; Kapp and DeCelles 2019). We attempt to bring together these different strands of evidence, focusing on topographic evolution because it is topography, and not the underlying geology, that has most impact on atmospheric dynamics and the biota. We highlight recent palaeontological finds that have been crucial to bringing about a transformation in the way we envisage the Tibetan region developing, and explore why palaeoaltimetry based on the biota differs so much from that derived from stable isotopes. Finally, we synthesise the Paleogene topographic development of the Tibetan region and discuss its relationship to monsoon characteristics and the evolution of Asian biodiversity.

The assembly of Tibet

There is long-standing evidence that Tibet evolved in a piecemeal manner (England and Searle 1986; Dewey et al. 1988; Fielding 1996; Tapponnier et al. 2001), and that Tibet is not a single monolithic block but an amalgam of Gondwanan terranes accreted successively onto the Eurasia plate. This assembly spanned more than 200 million years beginning in the early Mesozoic (Allègre et al. 1984; Burg and Chen 1984; Şengör 1984; Dewey et al. 1988; Yin and Harrison 2000; Metcalfe 2006; Guillot et al. 2019), India being the most recent terrane to accrete to Eurasia early in the Cenozoic. From north to south, the modern Tibetan Plateau comprises the Kunlun–Qaidam, Hoh Xil–Songpan Ganzi, Qiangtang and Lhasa terranes (Fig. 2). At the junctions of these tectonic blocks, again going from north to south, are major suture zones: the Ayimaqin–Kunlun Suture Zone (AKSZ) between the Kunlun–Qaidam and the Hoh Xil–Songpan Ganzi terranes, the Jingsha Suture Zone (JSZ) between the Hoh Xil–Songpan Ganzi and the Qaidam terranes, and the Bangong–Nujiang Suture Zone (BNSZ) between the Qaidam and Lhasa blocks. The Yarlung–Tsangpo Suture Zone (YTSZ) marks the junction between the Lhasa block and the Indian Plate (Fig. 2).

Fig. 2

a Map of the Tibetan region showing the major terranes, suture zones (dotted red lines), thrusts (black lines) and normal (extensional) faults (red lines); blue arrows of different sizes indicated the modern primary directions and relative rates of present-day surface motion derived from GPS measurements (Gan et al. 2007); KSZ—Kunlun Suture Zone; JSZ—Jingsha Suture Zone; BNSZ—Bangong–Nujiang Suture Zone; YTSZ—Yarlung–Tsangpo Suture Zone; compiled from Wang et al. (2014) and Guillot et al. (2019). b Inferred vertical profile along the transect A–B shown in (a); simplified from Kapp and DeCelles (2019)

The assembly of Tibet began in the Late Triassic to Early Jurassic with the arrival of the Kunlun–Qaidam and Qiangtang terranes (Roger et al. 2003; Gehrels et al. 2011), followed by the Qiangtang and Lhasa block collision in the Early Cretaceous (Kapp et al. 2005). The early Paleogene (55 ± 10 Ma) seems to be the preferred date for the onset of the India–Lhasa terrane collision for many authors (Beck et al. 1995; Rowley 1996; Acton 1999; Sigoyer et al. 2000; Aitchison et al. 2002; Leech et al. 2005; Zhu et al. 2005; Garzanti 2008; Guillot et al. 2008; Copley et al. 2010; Liebke et al. 2010; St-Onge et al. 2010; Cai et al. 2011; Wang et al. 2011; Hu et al. 2012; Meng et al. 2012; White et al. 2012; Zahirovic et al. 2012; Zhang et al. 2012b; Bouilhol et al. 2013; Li et al. 2013; Wang et al. 2014), but there are significant outliers in these estimates that span 65 Ma (Yin and Harrison 2000; Ding et al. 2005) to 20 Ma (van Hinsbergen et al. 2012).

The lack of consensus over the timing of the onset of the collision is in part due to confusion over what is meant by the ‘onset of collision’. For biologists, the important moment is when a land connection is first established allowing free exchange of terrestrial biota, while for a geologist it may be when oceanic lithosphere finally disappears even though a marine ‘moat’ may persist long after. Another source of confusion arises from disparities in palaeomagnetic data. To attempt to reconcile these differences, the concept of a Greater Indian Basin (GIB) has been proposed (van Hinsbergen et al. 2012, 2019; Huang et al. 2015). This envisages rifting and seafloor spreading from ~ 120 to ~ 70 Ma within an oceanic basin, the GIB, between what is now the Tethyan Himalaya and stable India (van Hinsbergen et al. 2012). This affects the amount of subducted material and also the timing of the onset of continental collision. In a recent re-examination of palaeomagnetic data, Rowley (2019a) found no support for the GIB concept, although Rowley’s analysis has been challenged (van Hinsbergen 2019) and that challenge rebutted (Rowley 2019b).

The timing of the establishment of a land bridge enabling easy biotic exchange between India and Eurasia has long been argued over (Chatterjee and Scotese 1999; Prasad and Sahni 1999; Whatley and Bajpai 2000; Briggs 2003; Sahni and Prasad 2008; Datta-Roy and Karanth 2009; Samant et al. 2013; Klaus et al. 2016), particularly as Cretaceous sediments in India host fossils with non-Gondwanan affinities (Jaeger et al. 1989; Sahni and Bajpai 1991; Prasad and Sahni 1999; Samant et al. 2013). A recent molecular phylogenetic study using dated fossils to constrain an analysis of 50 mammalian lineages argues for free exchange between India and Eurasia as early as 64.8–61.3 Ma in the Danian (Ni et al. 2020). This is at the oldest extreme of the likely collisional age (55 ± 10 Ma) that emerged from the literature review of Wang et al. (2014), but it does not mean a land connection existed along the whole length of the southern margin of Tibet because the most recent marine sediments in the central part of southern Tibet only disappear at ~ 50 Ma (Ding et al. 2005; Najman et al. 2010).

Each of the accretion events that built Tibet would have thickened the crust, subducted ocean and continental lithosphere, and resulted in surface relief. This means that when India arrived, Tibet already exhibited an inherited complex topography and underlying geology, so subsequent compression from the northward passage of India will have idiosyncratically produced different responses in different parts of the Tibetan region at different times (Kelley et al. 2019), and not automatically produced a plateau. During accretion and compression, a combination of fault/suture re-activation, local subduction, slab break-off, lower lithospheric ductile flow and even localised thermal thinning of the lower lithosphere (England and Houseman 1988) would have produced a complex surface relief interacting with atmospheric dynamics and biodiversity. It is essential to quantify the evolution of topography throughout the Tibetan region if we are to disentangle properly the complex interactions between surface height, climate systems (particularly monsoons) and biodiversity (Spicer 2017).

Palaeoaltimetry

Measuring past surface height is not straightforward, and always has to be inferred via various assumptions as to how a geological observation or measurement can be transformed into a quantitative estimate of elevation. Before reviewing these assumptions, we need clarity regarding terminology because the poor use of terms has contributed significantly to misconceptions regarding orogeny, and its consequences, in the Tibetan region.

Uplift and exhumation

The term ‘uplift’ has a particular meaning in geology and relates to a process in which work is done against gravity, and is defined as a vector opposite to that of gravity. Any height change of a rock body ‘must specify both a reference frame and the object that moves’ (Molnar and England 1990, p. 30). That reference frame is usually the geoid (sea level corrected for eustatic changes) or some other point on Earth’s surface whose height is known (relative to a reference frame). In a geological context, the object that moves relative to the reference frame need not be Earth’s surface, but can be a sub-surface rock body. When a rock body rises against gravity (for example, the ascent through the crust of a granite pluton) and it approaches the surface, it cools, so sometimes cooling rates are interpreted in terms of uplift. However, this is potentially very misleading because erosion can lead to a reduction in surface height relative to that rock body and thus an increase in cooling rates. In this case, no work has been done against gravity.

The process of exhumation (rock removal from the Earth’s surface) should never be confused with uplift; the two processes are quite different and have totally opposite implications for tectonics, climate, and biodiversity. Note, however, that erosion (removal of rock) often leads to isostatic rebound and thus uplift, which can result in further erosion. Such uplift is usually quite modest, and erosion is a function of climate as well as surface uplift. In the remainder of the present paper, we always refer to rocks at Earth’s surface because it is the surface that provides the substrate for biodiversity and interacts with the atmosphere.

Erosion means removal of rock material, which is then usually transported away from the erosional site to be deposited elsewhere. This deposition can infill an enclosed basin and in so doing raise the height of the sediment surface occupying the basin lowlands. Even though the height of that surface has risen above the depositional basement, the underlying rocks forming that basement may not have moved with respect to the geoid and, if anything, are likely to have subsided due to sediment loading. In this case, no work has been done against gravity, and so the process of raising the surface height cannot be termed ‘uplift’.

Lapse rates

Palaeoaltimetry is the measurement of past surface height, and changes thereof, against a reference frame or datum. Ideally, this datum should be the geoid, but eustatic sea level changes are often difficult to quantify, and therefore correct for, and are usually small compared with the uncertainties in palaeoaltimetric techniques. For an example of the potential complexity of sea level and other corrections in palaeoaltimetry, see Meyer (1992).

To measure past surface height, we need to measure changes in a physical parameter with elevation above a datum (e.g. sea level), and those changes need to be captured and preserved unchanged in the rock record. A common physical parameter that changes with increasing height above sea level (a lapse rate) is atmospheric temperature, and this gives rise to what are called ‘thermal lapse rates’. Past surface temperatures can be archived in the form of geochemical proxies such as clumped isotopes (e.g. Huntingdon et al. 2015), biomarkers (Inglis et al. 2019; Naafs et al. 2019), the intrinsic thermal tolerances of specific plants or animals (e.g. Deng et al. 2019; Su et al. 2019), populations of plants and animals (Kershaw and Nix 1988; Mosbrugger and Utescher 1997; Utescher et al. 2014), or temperatures encoded in plant form (e.g. using leaf physiognomy—Wolfe 1979, 1993; Spicer et al. 2009; Yang et al. 2015). Thermal lapse rates can be based on mean annual temperatures, seasonal means or even monthly mean temperatures. Cold month mean lapse rates are particularly useful because many plants such as palms are sensitive to freezing (Reichgelt et al. 2018), whereas mean annual temperatures can disguise large variations in seasonal temperature, both extremes of which may be crucial to both biological (as captured in fossil remains, including biomarkers) and geological processes (e.g. mineralisation and clumped isotopes (Eiler 2007, 2011)).

For meteorological purposes, a thermal lapse rate is measured in a column of free air and such ‘free air thermal lapse rates’ are important for many atmospheric processes. Their ready availability means that free air lapse rates also have a long history of use in palaeoaltimetry (reviewed in Meyer 1992). However, free air lapse rates are not relevant to determining land surface heights because absorption and re-radiation of heat, convective processes and air mass deflection at a surface lead to local lapse rates that are markedly different from those in a column of free air (Meyer 1992, 2007). In intermontane basins, where sediments accumulate, such lapse rates can even be negative (Wolfe 1992). These surface lapse rates, or ‘terrestrial thermal lapse rates’, are therefore the only ones that should be used for determining surface heights (Spicer 2018), but they have a crucial drawback in that they are location (e.g. distance from a shoreline) and, more importantly, topography dependant. This means the only realistic way of applying terrestrial thermal lapse rates is through empirical climate model mediation (e.g. Su et al. 2019).

Lapse rates are also used in stable isotope palaeoaltimetry. Here, the physical parameter that changes with height is the rate at which isotopic composition of atmospheric moisture changes with elevation when an air parcel is forced upwards on the windward side of a mountain range. Initially, this was investigated empirically (Garzione et al. 2000), but it was soon realised that a reasonably predictable process of Rayleigh fractionation takes place on the windward side of a mountain range, and that this provides a useful generic model for measuring palaeoelevation (Poage and Chamberlain 2001; Rowley et al. 2001; Rowley and Garzione 2007). As before, the elevation measurement is made with respect to a datum, and in stable isotope palaeoaltimetry that datum is the isotopic composition of water at a known height (usually sea level) at the start of the air parcel trajectory.

Rayleigh fractionation describes the effect of heavier isotopes being more likely to rain out from a parcel of air as it rises and cools than lighter isotopes, with the result that at progressively higher elevations the composition of rainwater gets more and more enriched with the lighter isotopes. A common isotope system used is that of oxygen (¹⁸O/¹⁶O), but the deuterium/hydrogen system can also be employed. Eva-poration and plant transpiration (evapo-transpiration) also result in fractionation, and here the lighter isotopes tend to evaporate to the atmosphere more than the heavier ones, leaving the source water (or plant) enriched with the heavier isotope, and ideally such vegetation effects should be factored into palaeoelevation estimates (Mulch 2016). The reverse happens when water condenses: the heavier isotopes tend to rain out preferentially. It is worth noting that climate modelling has shown the Rayleigh fractionation model is temperature dependant (Poulsen and Jeffery 2011), such that in warmer climates the rate at which ¹⁸O rains out of an air parcel is less than it would be in a cooler climate. In effect, this means that the isotopic lapse rate is reduced in warm climates.

In practice, there may be multiple cycles of evapo-transpiration and precipitation along an air parcel trajectory (Blisnuik and Stern 2005). This produces the so-called continental effect (Winnick et al. 2014) that is important for inland sites (such as over central Tibet), but is difficult to quantify accurately unless the analysis is done in conjunction with an isotope-enabled climate model (Sturm et al. 2010) and the air parcel traced back to its source. Moreover, the isotopic composition of rainwater is not preserved in its raw state, but as minerals or organic materials formed in lakes, plants or soils (e.g. Quade et al. 2007) into, and upon which, the rain fell. Mineralisation and the formation of organics (often wood cellulose or leaf waxes) (e.g. Kim and O’Neil 1997; Polissar et al. 2009; Luo et al. 2011) introduces further fractionation, as does isotopic exchange with the surrounding host rocks during burial (diagenesis) and with any later meteoric waters that might penetrate the host system (Garzione et al. 2004; Leier et al. 2009). Nevertheless, with careful petrological examination, diagenetic effects can be detected, and compensated for, and highly altered samples rejected (Garzione et al. 2004).

Moist enthalpy—a thermodynamic palaeoaltimeter

Because thermal lapse rates are spatially and temporally variable and, crucially, dependant on the topography that is being investigated, their application in palaeoaltimetry is problematic. Coupled with several palaeothermometers being prone to diagenetic alteration (e.g. biomarkers, Naafs et al. 2019, and isotopes Rowley and Garzione 2007), there has long been a desire for a proxy that is more robust to these limitations. The proxy that has gained most widespread use in this respect exploits the encoding of an atmospheric property known as moist enthalpy in the architecture (physiognomy) of woody dicot leaf fossils (Forest et al. 1995, 1999; Spicer et al. 2003; Khan et al. 2014; Su et al. 2018).

Moist enthalpy (H) is given as

$$ H=\left(c^{\prime } pT+ Lvq\right) $$

where c′p is the specific heat capacity of moist air at a constant pressure, T is temperature (in K), Lv is the latent heat of vaporisation of water and q is specific humidity.

Moist enthalpy (H) and potential energy (gZ) together make up the total energy contained in a parcel of air (kinetic energy is minimal in most circumstances and can be ignored) and this is known as moist static energy (h)

$$ h=H+ gZ $$

where g is the acceleration due to gravity (a constant) and Z is height.

Because h is mostly conserved as a parcel of air rises up the windward side of a mountain range, any difference in enthalpy between two locations, one low and one high (H_low − H_high), divided by the acceleration due to gravity (the constant g) gives the height difference (∆Z).

$$ \Delta Z=\left({H}_{low}-{H}_{high}\right)/g $$

Leaf fossils encode well the moist enthalpy they were exposed to when alive (Forest et al. 1999), but leaves are intrinsically vulnerable to being damaged if they are transported far from their growth site (Spicer 1981; Ferguson 1985; Spicer and Wolfe 1987). However, once buried and preserved, their features are not subject to the subtle diagenetic modification that can compromise isotope palaeoaltimetry. Character loss can occur through taphonomic processes and collecting, but provided a large enough assemblage of leaves representing different species are utilised, these losses only minimally affect analytical precision (Povey et al. 1994; Spicer et al. 2005). Moreover, even catastrophic loss of some characters, such as leaf size, does not degrade the ability to determine past surface height in any significant way because moist enthalpy is encoded in multiple leaf characters (Spicer and Yang 2010). Leaf susceptibility to pre-burial transport damage is actually advantageous in that it means that a diverse leaf assemblage inevitably represents vegetation growing close to, and crucially at the same elevation as, the depositional site.

Each proxy has its own inherent strengths and weaknesses so, ideally, they should be used in combination; each proxy tends to reflect different aspects of the palaeotopography of a region and so complement one another (Su et al. 2019). This complementarity is no better expressed than in the palaeoaltimetry of the Tibetan region.

Paleogene topography of Tibet

Prior accretions of the Songpan–Ganzi/Hox–Xil, Qiangtang and Lhasa terranes bequeathed a complex topography across Tibet at the start of the Paleogene, and parts of that relief in north-western Tibet have been preserved for at least 40 Ma (Law and Allen 2019). In the early Paleogene, the southern margin of the Lhasa terrane may have been at a latitude of ~ 10°N (Cao et al. 2017; Li et al. 2017b; Tong et al. 2017), but other estimates put it at 16.5 ± 4°N in the Early Cretaceous (Huang et al. 2015), and with the northward movement of India it is unlikely that the Lhasa terrane subsequently drifted southwards. By the early Eocene, it is estimated to have been at 21–24°N, moving another 8–10°N northward by the latest Oligocene (Meng et al. 2017).

The rise of the Gangdese Mountains

At the time of closure between the Lhasa and Qiangtang terranes in the late Early Cretaceous (Kapp and DeCelles 2019), a mountain chain or ‘Cordilleran-style orogen’ (Burg et al. 1983; Chang et al. 1986; England and Searle 1986; Coward et al. 1988; Ratschbacher et al. 1992; Kapp et al. 2007a, b; Kapp and DeCelles 2019) was already forming on the southern margin of the Lhasa terrane. To the north of these mountains, known as the Gangdese Arc, a shallow marine retro-arc foreland basin existed across the northern part of the Lhasa terrane, and this accumulated the sediments of the Takena Formation until ~ 92 Ma (Ding et al. 2005; Leier et al. 2007; Zhang et al. 2012a, b) (Fig. 3). The overlying Campanian (~ 83–78 Ma) lacustrine-fluvial, and slightly evaporative, units were mainly derived from the south, i.e. the rising Gangdese (Leier et al. 2007). This transition from marine to non-marine is not localised (Sun et al. 2015a, b; Orme and Laskowski 2016) and, coupled with an intense pulse of magmatism at 95–90 Ma (Zhu et al. 2017), suggest that the Gangdese arc exhibited a significant rise in mid-Cretaceous time.

Fig. 3

South and Central Tibet in the late Early Cretaceous. The Gangdese formed an elevated land surface at this time, but the crest height, although not yet quantified, is likely to have been ~ 2 km. The blue arrow illustrates summer air flow drawn northward by an Asian interior low-pressure system and the colour gradations indicate the relative moisture content (dark blue represents wet air and light blue relatively drier air)

Although the mid-Cretaceous crest height of the Gangdese has not yet been quantified, northward subduction of the Neotethyan oceanic lithosphere formed an ‘Andean’ type Gangdese range that had reached ~ 4.5 ± 0.4 km high by 56 Ma, possibly due to Indian slab break-off (Ding et al. 2014), but surface uplift was confined to the southern margin of the Lhasa terrane. In the northern part of the Lhasa block, the elevation remained low (≤1 km) despite rapid rock uplift between 80 and 70 Ma. This was due to fluvial erosion and peneplane formation between 70 and 50 Ma (Haider et al. 2013; Hetzel et al. 2011; Strobl et al. 2012).

The height of the Gangdese mountains appears to have remained above 4.5 km until the present, although locally there is evidence of ~ 1-km subsidence since the mid-Miocene (Saylor et al. 2009; Khan et al. 2014; Ding et al. 2017; Xu et al. 2018) (Fig. 7). Both phytopalaeoaltimetry (plant-based height proxies) and stable isotope palaeo-altimetric proxies consistently yield similar results along the Gangdese mountain chain (Spicer et al. 2003; Currie et al. 2005, 2016; Khan et al. 2014; Ding et al. 2017; Xu et al. 2018) (Fig. 4).

Fig. 4

Diagram showing the slowdown in the northward convergence of India with respect to Eurasia over time (dotted lines) from Molnar and Stock (2009), and palaeoaltimetry of the southern margin of Tibet as determined by stable isotope palaeoaltimetry and phytopalaeoaltimetry. Modified from Xu et al. (2018)

The Qiangtang uplands

Quantifying the elevation history of these uplands, today’s Tangula Mountains (Fig. 1), is more challenging than for the Gangdese because of the continental interior position of the Qiangtang terrane. The continental effect on isotopic composition is complex in this setting and can only be properly resolved using isotope-enabled climate models, but this is only starting to be done. Nevertheless, a quantitative estimate that has been attempted without model-mediation returned an elevation of ~ 5000 m at 28 Ma (Xu et al. 2013). It is not possible to say when this elevation was achieved, but the uplands were shedding sediment northwards in the Paleogene as evidenced by sediments in the Hoh–Xil Basin (Wang et al. 2008; Dai et al. 2012), and suggest that large parts of the Qiangtang terrane were already elevated in the Eocene. Very recently, modelling has been used to verify assumptions underlying a height estimate for radiometrically dated sediments within the Gonjo Basin (Fig. 1), which showed a rapid rise of the eastern Qiangtang uplands to ~ 3.8 km by ~ 44–40 Ma (Xiong et al. 2020).

Central Tibet in the Paleogene

Popular orographic models of Tibet typically envisage a palaeo-plateau from the Gangdese northwards across the BNSZ (England and Houseman 1988; Molnar 1989; Rowley and Currie 2006; DeCelles et al. 2007). This is particularly so for the Proto-Tibetan Plateau model that specifically envisages a ‘high and dry’ central core of Tibet throughout the Paleogene (DeCelles et al. 2007; Wang et al. 2014). To test these models, it is therefore essential to measure the surface height of the BNSZ over time, but when this has been attempted, different techniques have given markedly different results (Deng and Ding 2015).

Isotope palaeoaltimetry in Central Tibet

Until recently, estimating Cenozoic height changes across central Tibet mostly relied on the use of stable isotopes, with all the additional uncertainties that arise away from the windward slope of a single mountain front. Nevertheless, the ‘educated guesses’ (Hoke 2018, p. 96) used to minimise these and other uncertainties have routinely suggested late Paleogene BNSZ elevations similar to those of present (> 4 km) (Rowley and Currie 2006; DeCelles et al. 2007).

There are several depositional basins preserved along the BNSZ and two, the Nima and Lunpola (Fig. 1), are of particular importance because they yield surface height estimates based on both isotopes and fossils. The successions within these basins are divisible into a predominantly fluvial Paleocene–Eocene Niubao Formation and an overlying 1000-m-thick, predominantly lacustrine, Oligocene–Miocene (25.5–20 Ma) Dingqing Formation (Du et al. 2004; Ma et al. 2015, 2017; Han et al. 2019), with limited age control constrained by radiometrics (He et al. 2012), magnetostratigraphy (Sun et al. 2014a, b), cyclostratigraphy (Ma et al. 2017) and palynology (Wang et al. 1975; Sun et al. 2014a, b). Oxygen isotopes from carbonate palaeosol nodules in the upper Niubao Formation, Lunpola Basin (Fig. 1), yield Late Eocene elevations of 4850 m + 380/− 460 m, while underlying lacustrine limestones and marls within the same formation yield an elevation of 4050 m + 510/− 620 m (Rowley and Currie 2006). Thin-bedded lacustrine marls and limestones of the overlying Oligocene and Miocene Dingqing Formation, still within the Lunpola Basin, also give high mid-Miocene elevations of 4260 m + 475/− 575 m (Rowley and Currie 2006). In the nearby Nima Basin (Fig. 1), radiometrically dated soil carbonates give a palaeoelevation estimate of 4.5–5 km at 26 Ma (late Oligocene) (DeCelles et al. 2007). Deuterium/hydrogen (D/H) ratios in n-alkanes derived from leaf waxes also give similar results (3600–4100 m for the Niubao Formation and 4500–4900 m for the Dingqing Formation) (Polissar et al. 2009).

Palaeontological evidence for surface height

Fossil finds from the BNSZ basins tell a totally different story and attest to low elevation humid conditions. The BNSZ basins host numerous faunal (insect, fish, mammal), floral (megafossil and pollen/spores) and molecular (biomarker) fossils within a 4-km-thick Cenozoic succession of palaeosols, fluvial, fluvio-deltaic and lacustrine units, some indicative of freshwater and some of saline conditions (Ma et al. 2015, 2017).

Insects

The fine-grained laminated grey lacustrine units of the Dingqing Formation in the Lunpola Basin (Fig. 1) indicate a long-lived stable lake environment ideal for preserving delicate structures such as those seen in insects, and the find of a water strider, Aquarius lunpolaensis (Fig. 5a, b), suggests a comparatively low elevation at ~ 25 Ma (Cai et al. 2019). The closest modern relatives are today found at low elevations across parts of Europe, Scandinavia and north Africa (Damgaard 2005).

Fig. 5

Examples of fossil water striders and a reconstruction of Limnobiophyllum inhabiting the Dingqing Formation lacustrine environment, central Tibet. a, b Aquarius lunpolaensis, c reconstruction by Fei-Xiang Wu of the floating aquatic plant Limnobiophyllum pedunculatum

Fish

In the Nima basin (Fig. 1), numerous fish fossils include Tchunglinius tchangii (Wang and Wu 2015) dating from 26 to 23.5 Ma (late Oligocene), which is a species also indicative of low elevations. It belongs to the subfamily Cyprininae of the Cyprinidae, which are widely distributed across Asia, Africa, and central and southern Europe. Low-elevation/warm-climate species have fewer vertebrae than those that live in high, cold environments such as the endemic ‘snow carps’ (Schizothoracinae, Cyprinidae) of Tibet today. With T. tchangii having only 33 vertebrae, compared with the 46–48 of the modern high-elevation snow carps, Wang and Wu (2015) suggested that in the late Oligocene the Nima Basin must have been at a low elevation. However, factors other than those related to altitude can affect vertebrae number, and several North American cyprinids have just 34 (Coburn 1986) and, lacking a physiological explanation linking vertebrae number and elevation, the use of such a metric as a palaeoaltimeter has been questioned (Renner 2016).

The Nima and Lunpola Basins (Fig. 1) also yield the earliest (by at least 20 Ma) example of a climbing perch. Eoanabas thibetana (Anabantidae) (Wu et al. 2017) (Fig. 6) from the Chattian (late Oligocene), or possibly earlier, lower Dingqing Formation likewise suggests a low palaeoelevation. Climbing perches are today restricted to tropical lowlands < 1200 m a.s.l. in South Asia and sub-Saharan Africa, and live in temperatures of 18–30 °C (www.fishbase.org) (Skelton 2001). Normally living in low oxygen conditions, climbing perches have developed the capacity to breathe air directly, but as a consequence their ability to take in oxygen only from the water is restricted. This means that they cannot survive where lake surfaces become frozen, so are confined to low elevations, and certainly well below the 4 km or more suggested by the oxygen isotope palaeoaltimetry. Because of this requirement for ice-free lake surfaces, the biomarker interpretation of a cold dry climate in the Lunpola in the lower part of the Dingqing Formation (Cao et al. 2020) is doubtful.

Fig. 6

Specimens of the fossil climbing perch, Eoanabas thibetana. This species is found in the Oligocene lacustrine lower Dingqing Formation sediments of the south Nima and Lunpola Basins (Wu et al. 2017), and occurs at the same site that yields the palms shown in Fig. 7. Such specimens preserve the specialised air-breathing apparatus seen in modern climbing perches. Drawing by Fei-Xiang Wu. Scale bar represents 5 mm

In central Tibet, lowland thermophillous fishes such as E. thibetana and T. tchangii do not persist into the Miocene, but are replaced higher in the Dingqing Formation by Plesioschizothorax microcephalus, a primitive representative of the snow carps that persist today on the Tibetan Plateau. Comparison with living snow carps suggest that P. microcephalus was likely to have lived at ~ 3000 m after adjusting for secular climate change (Chang et al. 2010), but again this estimate could be challenged as with T. tchangii.

Mammals

The Dingqing Formation fauna has also yielded an early Miocene rhinocerid almost identical to Plesiaceratherium gracile from the lowland early Miocene Shanwang fauna in Shandong, eastern China, which according to Deng et al. (2012) limits the elevation to below ~ 3000 m.

Palynology

The above elevation estimates used free air thermal lapse rates, qualitatively by implication, to assess palaeoelevation, but in the case of fossil pollen, free air lapse rates have been used quantitatively. The pollen of the Dingqing Formation (Wang et al. 1975; Sun et al. 2014a) represents a peculiar mixture of both cool temperate forms together with more thermophillous taxa typical of subtropical conditions. The succession spanning 25.5 to 19.8 Ma is divisible into three palynological zones, with the wind-dispersed bisaccate conifer Pinuspollenites dominant in all three zones. Wind-dispersed pollen representing other cool temperate conifers such as Piceapollenites (indicative of spruce trees) and Abiespollenites (firs) co-occur with temperate broadleaved trees represented by Betulaepollenites (birch), Ulmipollenites (elms), Alnipollenites (Alder) and Quercoidites (oaks). However, predominantly tropical to subtropical evergreen broadleaved species such as Castanopsispollenites and Meliaceoidites also occur in the assemblages, and such a mix of taxa with widely diverse thermal tolerances cannot reflect co-occurrence in the source vegetation. Such an ecologically improbable amalgam was most likely formed during transport from disparate growth sites into the lake sediments.

To estimate palaeoelevation, Sun et al. (2014a) used the total palynological assemblage interpreted through the coexistence approach (Mosbrugger and Utescher 1997; Utescher et al. 2014) to obtain a mean annual temperature, together with an Eocene free air lapse rate (Song et al. 2010). The outcome was a palaeoelevation estimate of 3190 ± 100 m, but by aggregating the whole assemblage, this number represents an elevation somewhere within the altitudinal range occupied by all the vegetation communities contributing to the assemblage; both those high on the mountain sides (cool temperate vegetation) and those growing around the margins of the palaeo-lake occupy the floor of the basin/valley system (the sub-tropical community) (Su et al. 2019). As such, this surface height estimate for the lake margins is likely to be too high, even though it is lower than the isotopes indicate. There is no reliable and objective way to filter the various species contributions to correct for community mixing, but preserved alongside the pollen are plant megafossils, which because of their inherent susceptibility to degradation during pre-burial transport are more likely to reflect vegetation close to their lacustrine site of deposition, even in mountainous regions (Spicer and Wolfe 1987).

Plant megafossils

Plant megafossils from the lower (late Oligocene) Dingqing Formation, which are associated with the climbing perch, include palm leaves named Sabalites tibetensis (Su et al. 2019), leaves of the ‘golden rain tree’ Koelreuteria lunpolaensis and K. miointegrifolia (Jiang et al. 2019), a Pistacia leaflet, Cedrelospermum tibeticum (Jia et al. 2019), a leaf of Araliaceae, a palmately compound leaf of Handeliodendron sp., with six leaflets, leaves of the water plant Limno-biophyllum pedunculatum (Fig. 5c) (Low et al. 2019), leaves of Exbucklandia sp., Magnoliales, Salix sp., several unidentified toothed and untoothed woody dicot leaves, and the lake margin monocot Typha (Wu et al. 2017) (Fig. 7). The flora as a whole consists of intact leaves and leaflet clusters and shows no sign of long-distance transport, and so represents vegetation growing spatially and altitudinally very close to the lake margin. Qualitatively these taxa represent sub-tropical to warm temperate conditions consistent with those indicated by the associated fish and insect fauna.

Fig. 7

Illustrations of representative plant fossils of the Dingqing flora, modified from Wu et al. (2017) and Su et al. (2019). a A frond of Sabalites tibetensis T. Su et Z.K. Zhou (Su et al. 2019); b compound leaf of Araliaceae; c Koelreuteria sp. (golden rain tree); d Pistacia leaflet; e winged fruit of Cedrelospermum sp.; f unidentified toothed leaf; g unidentified entire-margined leaf; h unidentified toothed leaf; i infructescence of probable Araceae; j leaf of Magnoliales; k unidentified entire-margined leaf; l Salix sp.; m unidentified entire-margined leaf; n fragment of Typha leaf

Until further collecting is undertaken, the Dingqing leaf flora is not diverse enough to conduct a leaf physiognomic analysis (CLAMP—http://clamp.ibcas.ac.cn). However, of particular interest are the palm remains because palms are intrinsically cold intolerant (Reichgelt et al. 2018), and if the relevant cold month mean terrestrial thermal lapse rate were known, the palms could provide a maximum possible elevation for the basin floor (Su et al. 2019).

Large (~ 1 m long) fronds of the coryphoid palm Sabalites tibetensis T. Su et Z.K. Zhou (Fig. 7a) have been recovered from the grey finely laminated sediments of the lower Dingqing Formation (Su et al. 2019). By using 13 possible topographic configurations for central Tibet, a General Circulation Model with Chattian boundary conditions showed that palm winter survivability was only possible within a valley, the bottom of which could have been no more than 2.3 km above sea level. This is almost 3 km lower than some stable isotope determinations (Su et al. 2019). Such a climate model–mediated approach to quantitative palaeoaltimetry avoided the use of an inappropriate mean annual free air lapse rate, using instead the more precise cold month mean temperature terrestrial lapse rate and, by integrating a model with Chattian boundary conditions, secular climate change was automatically compensated for.

Hosting several plant-bearing layers, the underlying middle Eocene (Lutetian) Niubao Formation in the Bangor Basin (Fig. 1), near the village of Jianglang, contains a diversity of remains including leaves, fruits and seeds, such as Ailanthus maximus, Asclepiadospermum marginatum, A. ellipticum and Lagokarpos tibetensis (Liu et al. 2019; Tang et al. 2019; Del Rio et al. 2020). These represent low-elevation humid sub-tropical vegetation with floristic links to the Eocene Green River flora of western USA, and the middle Eocene Messel flora in Germany, as well as other similarly dated assemblages in Africa and the UK (Del Rio et al. 2020), and suggests that a deep central valley existed between the Gangdese and Qiangtang highlands throughout much of the Paleogene (Fig. 8). Figure 9 shows examples of Largokarpos from North America, Germany and the Jianglang flora (Tang et al. 2019).

Fig. 8

Cartoon showing southern and central Tibet in the Paleogene. Lowland vegetation was characterised by sub-tropical taxa, including palms as shown in the inset reconstruction, along with a diverse fauna that included the climbing perch Eoanabas thibetana. Higher on the mountainsides, cool temperate coniferous vegetation predominated. Lake level was below 2 km, while the Gangdese and Qiangtang uplands exceeded 4.5 km. The blue graded arrow indicates summer airflow becoming drier as it passes over the Gangdese highlands. Reconstruction by Fei-Xiang Wu

Fig. 9

Examples of Lagokarpos from the Eocene of a North America; b Messel, Germany; c Jianglang flora, central Tibet; d reconstruction of Largokarpos. Scale bar represents 5 mm

Lowland thermophilic humid forests in the Paleogene of central Tibet are also evidenced by finds of dipterocarp amber (Wang et al. 2018a), apparently reworked from the Niubao into the Dingqing Formation. Dipterocarps are tropical lowland forest dominants across S. Asia today (Ashton 1982, 2003; Corlett 2019; Morley 2000), but at the time of writing no definitive dipterocarp megafossils have been found, so the existence of dipterocarps in the Paleogene of central Tibet needs to be further investigated.

Reconciling stable isotope and palaeontological palaeoaltimetry

Stable isotope and palaeontological proxies give highly divergent results in central Tibet, so how can this be reconciled? The first question to answer is: are these two proxies inherently different in terms of calibration? Unfortunately, multiproxy cross-validation is still rare, but one location that has been intensively studied is the mid-Miocene (15 Ma) Namling–Oiyug Basin (Fig. 1, Table 1) within the Gangdese highland. Here, both well-dated leaf fossils and lacustrine/palaeosol carbonates have been examined from a palaeoaltimetric perspective using foliar physiognomy, oxygen and D/H ratio isotope systems.

Table 1 Location and climate attributes for plant fossil assemblages in the Tibetan region (Fig. 1) as derived using CLAMP with a PhysgAsia 2 modern vegetation and WorldClim2 climate calibration

The first quantitative height determination in the basin used an early version of CLAMP to derive palaeo-moist enthalpy from the Namling flora; a leaf assemblage from predominantly lacustrine sediments exposed within a single inclined horizon spanning modern elevation of 4300–4700 m (Spicer et al. 2003). Compositionally, this flora represents cool temperate vegetation with strong affinities to taxa found today in Yunnan, Sichuan, the Himalaya and northern China (Guo et al. 2019), and is indicative of high elevations. Because suitably diverse proximal contemporaneous sea-level floras were not available at the time of the original study, a constrained climate model was used to provide the required moist enthalpy datum from which a palaeoelevation of 4689 ± 895 m was obtained. Subsequently, this was revised to 5260–5540 m using a calibration more suited to Asia, and a near sea level datum from a similar-aged flora (Kameng River, Fig. 1 and Table 1) in the Siwaliks, NE India, but climate model mediation was still used to correct for palaeolatitude (Khan et al. 2014). This high surface height measurement is indistinguishable from an elevation of 5200 + 1370/− 606 m obtained using oxygen isotopes in carbonates at the same location (Currie et al. 2005) and the Rayleigh fractionation model. It is also the same as a value of 5100 +1300/− 1900 m subsequently obtained using hydrogen isotopes from leaf waxes (Currie et al. 2016). This congruency shows that all three methods are similarly calibrated, so the disparities in central Tibet must be due to other factors.

Figure 10 illustrates why stable isotopes may give a falsely high estimate of elevation across a valley system, in this case along the Paleogene BNSZ, which then inevitably appears as a plateau. In the case of Tibet, a predominant Indian Ocean source to the south leads to moisture-laden winds encountering the E–W Gangdese mountain system. As moist air is forced upwards and cools, heavy isotopes preferentially rain out, leaving only moisture enriched with light isotopes to crest the mountain tops and move downslope (Deng and Jia 2018), adiabatically warming, into the valley to the north. Water entering the valley lake system by mountain streams on the leeward side of the mountains would also be depleted in the heavy isotopes. In winter, drier air moving southwards from a deep Asian interior (Siberian) low, and passing over the Qiangtang mountains, will also have a depleted heavy isotope content. This means that overall, the valley system will be isotopically light and therefore yield a ‘phantom’ high elevation more indicative of the height of the mountain crests bounding the valley than the valley lowlands. This effect is also shown by isotope-enabled computer models (Fig. 11). In model simulations where the western end of the BNSZ valley is open, a small amount of moisture penetrates on westerlies from the retreating Paratethys Ocean, but the proportion of heavy isotopes this introduces is insufficient to lower the apparent ‘phantom’ high plateau because the dominant moisture supply is via summer southerlies.

Fig. 10

Diagrammatic explanation of why isotope analysis results in a phantom elevated surface height for the lowlands of the Paleogene central Tibetan valley. In summer, moist air from the south rises against the Gangdese highlands preferentially precipitating out heavy isotopes leaving only isotopically light water to crest the mountain ridge and enter the semi-enclosed valley hydrological system. In winter, drier air, which is already isotopically light because of continental interior source, is derived predominantly from the north and has to pass over the Qiangtang highlands before it too enters the internally drained valley. Consequently, only isotopically light water is preserved in the valley as organics or carbonates, resulting in a surface height estimate reflecting of the height of the mountain crests bounding the valley

Fig. 11

Maps showing the results of an isotope-enabled experiment we have conducted using Lutetian boundary conditions, illustrating the lack of a clear isotopic signature when a valley is present in central Tibet. a South Asia orography with a deep central valley positioned at 35°N, b similar to (a) but with a high plateau, c stable oxygen isotope map with a deep valley as in (a), d stable isotope map for the plateau scenario as in (b). The valley is not discernible in (c) where the isotopic map is similar to that for a plateau (d)

Northern Tibet

Stable isotope palaeoaltimetry studies seem to suggest that the Qaidam Basin, currently at an elevation of between 2.8 and 3.2 km, was apparently low in the Paleogene, and only became elevated in the Miocene (Kent-Corson et al. 2009; Li et al. 2016). However, there is ample evidence that parts of northern Tibet rose much earlier, e.g. areas near the Xining (Dupont-Nivet et al. 2007, 2008; Garzione 2008), Qaidam and Tarim Basins (Graham et al. 2005; Bershaw et al. 2012; Kent-Corson et al. 2009). A new discovery of a cool temperate, predominantly deciduous, early Oligocene flora (Fig. 12) in the northern Qaidam Basin also supports that view (Song et al. 2020), as well as some leaf wax hydrogen isotope results from the radiometrically dated sediments on the Hoh–Xil Basin (Lin et al. 2020). However, until the Hoh–Xil Basin results are evaluated using isotope-enabled climate models, they must be regarded as speculative because of unknowns regarding the source waters, air parcel trajectories and the ‘continental effect’.

Fig. 12

a Stratigraphic column of the section that preserves the plant fossils illustrated in (b)–(z), modified from Song et al. (2020). This is located in the lower member of Shangganchaigou Formation of the Dahonggou section, northern Qaidam Basin. Shown is the magnetostratigraphic correlation, the chronology and the lithology of the studied section, as well as the chronologic and stratigraphic position of the plant fossil locality. b–z Early Oligocene Qaidam plant fossils, modified from Song et al. (2020). All scale bars = 1 cm. b Leafy shoot of Glyptostrobus; c fruit of Cyclocarya; d single-seeded pod of Podocarpium; e fruit of Populus; f–z representative specimens of the 21 morphotypes of woody dicot leaves

The age of this new Qaidam flora (Fig. 1) is obviously crucial in any discussion of the timing of uplift in this region. The flora comes from the lower member of the ~ 1500-m-thick Shangganchaigou Formation and is dated as ∼ 30.8 Ma (Rupelian) based on a high-resolution magnetostratigraphic study (Ji et al. 2017) of a long continuous section. The magnetostratigraphy spans what is known as the Dahonggou section, and is constrained by a detailed study of the ostracod assemblage succession in the section: a distinct shift in ostracod assemblages from freshwater Ilyocypris to brackish-water Cyprideis, as well as mammal fossils (Trilophodon). This section exposes, from oldest to youngest, the Lulehe Formation (Fm.), the base of which is dated as ~ 52 Ma, the Xiaganchaigou Fm., the Shangganchaigou Fm., the Xiayoushashan Fm., the Shangyoushashan Fm. and the Shizigou Fm., the top of which is dated as ~ 7 Ma (Ji et al. 2017). This dating framework is consistent with that of the nearby Hongliugou section (Fang et al. 2019) in which the age of the Shangganchaigou Formation is also well constrained by mammal fossils and magnetostratigraphy, while the Eocene age of the Lulehe Formation matches the cooling ages, as indicated by independent thermochronological dating in the surrounding mountains (Zhuang et al. 2018; Cheng et al. 2019). A younger age for the Lulehe Formation proposed by Wang et al. (2017b) was not adopted by Song et al. (2020) because the chronology of Wang et al. (2017b) is in conflict with other regional chronostratigraphic schemes, and Cenozoic source-to-sink provenance studies for the basin (for a detailed discussion see Song et al. 2019, 2020).

Using a Rupelian (33.9–27.82 Ma) climate model to determine sea level moist enthalpy, coupled with a CLAMP-derived moist enthalpy measurement from the Qaidam flora (Table 1), resulted in a palaeoelevation of 3.2 ± 1.4 km for the lakeside vegetation, which is indistinguishable from that of the present elevation. It follows that either there was an Eocene uplift of the region or an upland derived from pre-Cenozoic terrane collisions already existed in the Eocene. An Eocene rise is consistent with thermochronological studies that indicate an Eocene exhumation/cooling in the nearby Beishan mountains (Cheng et al. 2016; Qi et al. 2016; Zhuang et al. 2018), as well as numerous Eocene cooling ages derived from detrital thermochronological studies from the Qaidam Basin (Guan and Jian 2013; Wang et al. 2015; Du et al. 2018; Jian et al. 2018), Subei Basin (Li et al. 2017a), Jiuquan Basin (He et al. 2017) and Lanzhou Basin (Wang et al. 2017c). Irrespective of whether the area rose in the Eocene, or much earlier, this high landscape in northern Tibet preceded the rise of the Himalaya. Taken together with the Paleogene rise of other areas of northern Tibet (Dupont-Nivet et al. 2008), it suggests significant ‘far field’ deformation and uplift in northern Tibet, seemingly simultaneously with eastward extrusion of Tibet and the building of the Hengduan mountains.

Eastern Tibet and the Hengduan Mountains

The modern SE Tibetan Plateau consists of several continental fragments. These include parts of the Lhasa, Qiangtang, Songpan-Ganzi and Yangzte terranes, the Yidun arc, and the Tengchong, Baoshan and Lanaping–Simao terranes (Burchfiel and Chen 2012). A significant proportion of N-S shortening under compression from the India collision was likely to have been accommodated by extrusion of some of these fragments in a semi-rigid manner to the east and south-east (Leloup et al. 1995; Tapponnier et al. 2001; Wang et al. 2016; Li et al. 2017b) along several thousands of kilometres of strike–slip faults within the SE margin of Tibet, including the Ailaoshan–Red River Shear Zone, Chongshan Shear Zone and Gaoligong Shear Zone (Searle 2006; Li et al. 2017b). This could have occurred early in the India-Asian collisional process (Spurlin et al. 2005; Ding et al. 2007), but the low-relief, high-elevation topography has also been used as evidence for ductile flow of the lower crust into the SE margin of the Tibetan Plateau in the Miocene (Royden et al. 1997; Clark and Royden 2000). However, this interpretation depends crucially on a reliable regional dating framework as we discuss below.

A modification of the brittle model envisages extrusion, translation to the south and rotation of the Indo-China block. Based on palaeomagnetic analysis, Tong et al. (2017) argue that since the late Eocene the Lhasa and Qiangtang terranes have not experienced enough crustal shortening to have driven the southward extrusion of SE Tibet. Instead, they suggest the displacement of the Indo-China block, which was located to the north of the Qiangtang terrane until ~ 43 Ma (Fig. 13). The displacement would have involved as much as 600 km of extrusion, and rotation and deformation of the Indo-China block (Li et al. 2017a, b).

Fig. 13

Extrusion of the Indo-China block based on Tong et al. (2017). Palaeolatitude is only approximate

If the crustal processes driving the uplift of SE Tibet are unclear, so is the timing of its rise. Based on low-temperature thermochronological studies, a multi-phased rapid uplift of SE Tibet, starting as early as the Late Cretaceous (Wang et al. 2012, 2018a, b; Liu-Zheng et al. 2018; Cao et al. 2019), has been suggested. However, until recently the timing most widely favoured has been a predominantly Miocene uplift of the region based on river incision measurements (Clark and Royden 2000; Clark et al. 2004). Raising of land surfaces results in river incision and re-organisation of drainages, and rapid incision of major river systems across eastern Tibet and the Hengduan mountains is estimated to have taken place between 15 and 10 Ma, a phenomenon used to support the argument for lower crustal flow (Clark et al. 2005; Ouimet et al. 2010). However, an intensification of monsoon rainfall at around this time (Farnsworth et al. 2019) could also have increased incision rates (Nie et al. 2018), and so incision alone is not definitive evidence of uplift. Nevertheless, Eocene to Miocene provenance shifts also took place (e.g. Clark et al. 2004; Clift et al. 2006, 2020; Yan et al. 2012; Wissink et al. 2016), but interpreting these data correctly depends on a reliable regional dating framework.

Dating is everything

Until recently, the baseline dating reference for the region lacked absolute (radiometric) age constraints, so it relied in large part on biostratigraphic correlations, which carry a large element of circularity in regard to climatic and surface height studies. Because many of the floral and faunal assemblages appeared similar to those of today, most of them came to be regarded as Miocene, with an inevitable impact on regional palaeoaltimetric work, not just that relating to drainage patterns but also to stable isotope palaeoaltimetry (Hoke et al. 2014; Li et al. 2015).

A study by Gourbet et al. (2017) provided the first clue that regional dating required revision. Using radiometric constraints, they showed that within the Jianchuan Basin (Fig. 1), supposedly Oligocene to Pliocene sediments (the Jinsichang, Shuanghe and Jianchuan formations) were in fact all late Eocene in age. Earlier palaeoelevation estimates that assumed a Miocene age were revised from 2.6 km (Hoke et al. 2014) down to 1.2 ± 1.2 km (Gourbet et al. 2017), while subsequent more detailed re-analysis using both isotopic and pollen analysis revised the elevation a third time to be 1.3–2.6 km in the Eocene. Although lower than now, this estimate suggests some uplift long before the Miocene.

The discovery of a primary (i.e. not detrital) ash bed in the Lühe Basin (Fig. 1) with a zircon U–Pb age of 33 ± 1 Ma, and additional overlying ash beds dated as ~ 32 ± 1 Ma (Linnemann et al. 2018), has led to a major re-appraisal of the age of regional floras. These dates showed beyond doubt that the associated ‘Miocene’ flora was some 20 million years older than previously thought based on the modern aspect of the component leaves, seeds and pollen (Zhang et al. 2007; Xu et al. 2008; Huang et al. 2016; Tang et al. 2020). Further finds of ashes near the base of the nearby Luhe coal mine have yielded ⁴⁰Ar/³⁹Ar ages for biotites and feldspars of ~ 33 Ma (Li et al. 2020), and thus match the U–Pb dates obtained by Linnemann et al. (2018) and those of detrital zircons found in fluvial sands higher within the mine section (Wissink et al. 2016). Coupled with palaeomagnetostratigraphy, these ages suggest the Lühe Basin evolved between ~ 35 and 26.5 Ma, and constrain the onset of movement of the basin-bounding Chuxiong fault to 35 Ma (Li et al. 2020). This demonstrates that the onset of fault movement was contemporaneous with initiation of the Ailaoshan–Red River fault system (Tapponnier et al. 1990; Leloup et al. 1995; Cao et al. 2011) and suggests a common, region-wide, cause.

Radiometric re-dating of ‘Miocene’ floras has also led to a re-evaluation of topographic development further north on the eastern margin of Tibet. The Mankang (Markam) Basin (Fig. 1) previously thought to have formed in the Miocene (Tao and Du 1987; B.G.M.R. 1991; Su et al. 2014) is now known to host sediments and fossil plant assemblages that span the Eocene–Oligocene (E-O) boundary (Su et al. 2018). A lower assemblage here called Markam 3, bounded by volcaniclastics that yield ⁴⁰Ar/³⁹Ar ages of 35.5 ± 0.3 below and 34.61 ± 0.8 Ma above (latest Eocene), represents evergreen and deciduous mixed sub-tropical to warm temperate (Table 1) vegetation. An overlying assemblage (Markam 1) is also bounded by volcaniclastics dated below as 34.7 ± 0.5 and 33.4 ± 0.5 Ma above, placing it at or just above, the E-O boundary and is composed of markedly smaller leaves indicative of cooler temperate (Table 1) vegetation. CLAMP-derived moist enthalpy from these and sea-level floras in India (Fig. 1 and Table 1) yielded an elevation of 2.9 ± 0.9 km for the latest Eocene assemblage and 3.9 ± 0.9 km for the earliest Oligocene assemblages, matching the modern elevation of 3910 m. The ~ 1 km difference between the two assemblages may be genuine or may in part reflect secular cooling across the E-O transition, but such a rapid rise would be consistent with widespread crustal shortening and igneous activity within the Qiangtang terrane at that time (Horton et al. 2002; Kapp et al. 2005).

Tectonism and motion to the SE is ongoing in the Hengduan mountains and Yunnan (Fig. 5), but if near-modern relief across parts of SE Tibet and NW Yunnan was achieved by the end of the Eocene, when did most of the regional uplift occur? Palaeomagnetic studies of the Gonjo and Ranmugou formations on the eastern Qiangtang terrane (Tong et al. 2017) reveal that both the Lhasa and Qiangtang terranes underwent N-S crustal shortening before 54–43 Ma (early Eocene), while a recent clumped isotope study of radiometrically dated carbonates from the Ranmugou Formation in the Gonjo Basin, ~ 100 km north of Markam (Fig. 1), showed that the area rose from ~ 700 m in the early Eocene (54–50 Ma) to ~ 3800 m by the middle Eocene (44–40 Ma) (Xiong et al. 2020). Unlike earlier studies, Xiong et al. (2020) used an isotope-enabled climate model to verify air parcel trajectories, isotopic lapse rates and assumptions related to the continental effect. This deformation/elevation was followed by ∼ 1300 ± 410 km of crustal shortening in the northern Qiangtang terrane after 35.4 ± 2.4 Ma (Tong et al. 2017) suggesting parts of northern Tibet were being uplifted at this time, perhaps affecting the Qaidam Basin where near current elevations were achieved by ~ 31 Ma (Song et al. 2020).

The Himalaya

The Himalaya are the most recent component of the Tibetan region to develop, and although their origin is in the early Eocene, they are essentially a Neogene construct (Ding et al. 2017). An early attempt to quantify the rise of the Himalaya was made nearly 50 years ago after the discovery of Quercus (Quercus sect. Heterobalanus) leaf fossils at a reported modern elevation of 5.6 km on the 8027-m-high Himalayan peak known as Shishapangma (also known as Gosainthān) (Fig. 1) (Xu et al. 1973). Because the fossils were assigned a Pliocene age, it seemed to suggest a very recent Himalayan uplift, but these remains were found only as loose blocks (‘float’) and their precise origin and exact age remain unknown.

More secure evidence for the growth of the Himalaya was offered by Ding et al. (2017) and Xu et al. (2018), incorporating isotope palaeoaltimetry of Qoumolungma (Mount Everest) (Gébelin et al. 2013) (Fig. 1). They suggest that uplift began in the early Eocene, but it was not until ~ 26 Ma when the rate of convergence between India and Asia slowed (Meng et al. 2017), that the locus of deformation produced by the India–Eurasia collision moved from Tibet to south of the Gangdese and a rapid acceleration in Himalayan uplift (Ding et al. 2017). This phase of rapid Himalayan development coincided with a slowdown in India’s northward motion (Molnar and Stock 2009) (Fig. 4), suggesting greater resistance to further movement was being exerted by the compression of Tibet, which was accompanied by enhanced eastward extrusion. To quantify the rates of Himalayan rise, Ding et al. (2017) and Xu et al. (2018) used a combination of moist enthalpy derived from radiometrically dated plant fossils and oxygen isotope analysis (Fig. 4).

Moist enthalpy estimates derived from tropical early Eocene (~ 56 Ma) plants of the Liuqu flora (Fang et al. 2005) (Fig. 1 and Table 1), just south of the YTSZ, show that at that time the land surface had risen to an elevation of ~ 1 km. Locally marine units persisted until ~ 58.5 to ~ 55 Ma (Ding et al. 2005) or even as late as ~ 50 Ma (Najman et al. 2010). By 24–21 Ma, a CLAMP analysis of the nearby warm temperate Qiabulin flora (Fig. 1 and Table 1) showed a rise to ~ 2.3 km; an elevation matching that derived from oxygen isotopes (2.5 km). Analysis of oxygen isotopes from a horizon higher in the Qiabulin section, and radiometrically constrained to be 21–19 Ma, yields an elevation of ~ 4 km, and by ~ 15 Ma the Qomolangma (Everest) region had risen to at least 5 km (Gébelin et al. 2013), reaching the elevation of the Gangdese. It was only after the middle Miocene that the Himalaya rose above the Gangdese and imposed their own influence on atmospheric circulation by virtue of friction from both their height and width, and formed a strong rain shadow effect across Tibet (Ding et al. 2017) (Fig. 14). This led to an inevitable restructuring of air parcel trajectories and atmospheric dynamics across the Tibetan region and beyond.

Fig. 14

Cartoon showing the rise of the Himalaya and summer rain-shadow effect across an elevated (> 4.5 km) Neogene Tibet

Concluding remarks—the growth of Tibet, monsoons and biodiversity

As regards climate and biotic evolution, the most important aspect of the assembly of the Tibetan region has been its complex and dynamic topographic history. By combining isotope and palaeontological palaeoaltimetry with climate modelling, it has been possible to characterise south and central Paleogene Tibet as consisting of two E-W-orientated mountain ranges, the Gangdese and Qiangtang (Tangula) uplands, bounding a wide lowland centred on the Bangong–Nujiang Suture Zone (Fig. 8). Until sometime in the Eocene, this valley drained to the south through the Gangdese (Laskowski et al. 2019), but this ceased as the Himalaya grew and the valley became largely, if not exclusively, internally drained and supported closed lake systems (Ma et al. 2015, 2017). Over time, compression narrowed this valley and marked uplift began at 23.7 Ma (Ma et al. 2017). This compression, together with the internal drainage (Han et al. 2019; Su et al. 2019) and ‘bathtub’ sediment fill, raised the surface of the valley floor to approach the present central plateau elevation of ~ 4800 m in the late Neogene. Some contribution from local lower lithosphere thinning or crustal flow (Han et al. 2019) cannot be ruled out, but this requires further investigation.

This compression in the Eocene, and extending into the early Oligocene, also uplifted parts of northern Tibet, led to the elevation of eastern Tibet and the Hengduan mountains, and initiated the rise of the Himalaya. At the end of the Oligocene, deformation across Tibet slowed and switched to building the Himalaya, which showed its fastest rise after 23 Ma coincidental with a slowdown in the rate of India’s northward progression. However, only after the mid-Miocene did the Himalaya substantially exceed the height of the Gangdese to exert a strong drying over the forming plateau, and likely altering the South Asia monsoon system (Boos and Kuang 2010; Molnar et al. 2010) to something resembling its modern state.

There is ample evidence, both qualitative and quantitative, that monsoon systems affecting Asia long pre-date the Miocene (Huber and Goldner 2012; Srivastava et al. 2012; Shukla et al. 2014; Licht et al. 2014; Spicer et al. 2016, 2017) and even the Cenozoic (Armstrong et al. 2016; Farnsworth et al. 2019). If the links between topography and climate are as close as they are argued to be (Boos and Kuang 2010; Molnar et al. 2010), and barring any threshold effects, the incremental development of the Tibetan region should be matched by an incremental evolution of the Asian monsoon systems. It then follows that because those systems display considerable spatial variation in how they are characterised, no single proxy data point can be used reliably to infer the existence, or otherwise, of large-scale monsoon circulation. That aside, it is clear that in the Eocene there was intensification of the monsoon in East Asia (Quan et al. 2011), and southern Asia experienced a monsoonal climate (Shukla et al. 2014), but leaf adaptations suggest that the monsoons across southern Asia had more in common with today’s Indonesia–Australia monsoon than the modern South Asia monsoon or East Asia Monsoon (Spicer et al. 2016, 2017). Unlike the Indonesia–Australia monsoon, which is largely a function only of seasonal migrations of the Inter-tropical Convergence Zone, the East Asia and South Asia monsoons appear strongly influenced by topography (Boos and Kuang 2010; Molnar et al. 2010). Both palaeogeography and the topography of the Tibetan region appear to play a greater role in monsoon dynamics than atmospheric carbon dioxide (Farnsworth et al. 2019), so future modelling exercises will need to move away from treating the Tibetan region in a simplistic monotonic fashion, but instead attempt to replicate past orography in as realistic a way as possible. If this does not happen, model-related artefacts will become baked into perceptions of the interplay between orography, climate and biotic systems, and potentially mislead future research directions, just as simplistic thinking has done for the orographic evolution of the Tibetan region.

The interactions between topography, climate and biodiversity in the Tibetan region are complex, but a dynamic high-relief orography offers high levels of niche heterogeneity favouring multiple speciation events (e.g. Spicer 2017, Spicer et al. 2020a; Rahbeck et al. 2019). Numerous molecular phylogenetic analyses suggest intense speciation in the Miocene and many authors have attempted to make links between this diversification and the perceived ‘uplift of the Tibetan Plateau’ (Renner 2016; Lu et al. 2018). The evidence reviewed here shows that such an uplift of the plateau never happened, instead a high low-relief surface (a plateau) developed as a result of tectonic processes spanning at least 200 million years through a combination of uplift, erosion and sediment-infilling of basins. However, the Miocene is clearly a time by when the Hengduan Mountains, Himalaya and an extensive upland region across Tibet had collectively become established, with associated changes in atmospheric dynamics towards monsoon systems similar to those of the present day. After 15 Ma, the Miocene is also a time of global cooling, producing stronger altitudinal thermal gradients and offering new ecological niches for novel plant assemblies such as that of the modern Alpine biome (Ding et al. 2020). It is this development of the modern topography and climate across the Tibetan region that are likely responsible for the Miocene diversification apparent in so many Asian molecular phylogenies (Renner 2016; Lu et al. 2018).

The humid sub-tropical forests of the middle Eocene central valley of Tibet exhibit close floristic links to widely separated parts of the Northern Hemisphere (e.g. Europe, western North America, Africa) (Tang et al. 2019; Del Rio et al. 2020) showing that the valley was not isolated. However, the paucity of taxa in common with India (Liu et al. 2019) shows the strength of the biogeographic barrier posed by the Gangdese mountains compared with the other highlands bounding the valley. From the Late Cretaceous and into the early Paleogene, only one significant river system is known to have cut through the Gangdese Arc uplands (Laskowski et al. 2019). Sedimentologically, the valley system seems internally drained, at least from the late Paleogene onwards (Ma et al. 2015; Han et al. 2019), with sediment build-up eventually contributing to raising the height of the valley floor and creating a plateau late in the Neogene (Su et al. 2019). The bounding mountains seemed to have been high enough to produce a predominance of light stable isotopes in the semi-closed hydrological system within the valley. However, even if we ignore avian propagule dispersal, just a few narrow mountain passes would have been sufficient to afford biotic exchange.

Future work

Treating Tibet as a monolithic entity rising as a plateau, whether in geodynamic or climate models, confounds our ability to understand the complex interactions between orography, atmospheric dynamics and Asian biodiversity, and ignores major geological properties inherited from Tibet’s Mesozoic piecemeal assembly. Tibet evolved by multiple terrane accretion throughout the Mesozoic producing a landscape with complex relief by the time India joined Eurasia early in the Cenozoic.

Understanding the growth of Tibet, the Himalaya and the Hengduan mountains cannot be undertaken without the use of reliable palaeoaltimetric data. Future work needs to move away from simplistic slab-like treatments of Tibet by integrating into Earth system models more realistic topographic reconstructions derived by combining the various characteristics of different palaeoaltimetric proxies, particularly the rich palaeontological discoveries that are taking place. Through the use of isotope-enabled climate model mediation, new fossil finds and, crucially, radiometric dating, a new synthetic view of Paleogene Tibet is emerging. More detail will undoubtedly unfold as new, precisely dated, fossil finds and stable isotopes are analysed because such data are key components of understanding this landscape evolution, both on the present plateau and across SW China. The interactions between the geometry of Tibet, climate and vegetation are highly complex. To investigate and understand them requires suitably configured models that combine ocean and atmosphere dynamics, realistic palaeogeographies, isotope systems and vegetation/biodiversity feedbacks.