In this study, the early Miocene samples (dated to 18.3 Ma and 18.8 Ma) from the HLG section contained sufficient fractional concentrations to calculate the brGDGT-based paleotemperature proxy. Sedimentary facies analyses revealed that the two samples were composed of lacustrine sediments (Figure 2C). We therefore chose lake brGDGT-temperature calibration equations to calculate the paleotemperature (Tierney et al., 2010; Pearson et al., 2011; Sun et al., 2011; Loomis et al., 2012; Günther et al., 2014; Wang et al., 2016).

The MAAT value reconstructed by the three equations of samples mainly from Africa is obviously too high (Figure 4; Supplementary Table S3; Tierney et al., 2010; Pearson et al., 2011; Loomis et al., 2012). To be specific, the MAAT of 18.3 Ma was 40.4°C, 45.0°C, and 43.5°C, respectively, while that at 18.8 Ma was 31.7°C, 38.0°C, and 36.4°C, accordingly. Compared with the equations whose samples are mainly from Africa, the other three equations whose samples are mainly collected from China (Figure 4 and Supplementary Table S3; Sun et al., 2011; Günther et al., 2014; Wang et al., 2016) seem to be more reliable. Then, the mean MAAT for 18.3 Ma and 18.8 Ma was taken as the possible paleotemperature in this area at 18.0 Ma (Figure 5), whereas at 18.0 Ma, the MAAT values calculated using these three formulae were 31.3 ± 4.3°C, −0.3 ± 1.2°C, and 25.6 ± 1.5°C, separately. Among the three equations, it is believed that the reconstructed-MAAT of Sun et al. (2011) and Wang et al. (2016) is the most reliable for the following reasons. Firstly, Sun et al. (2011) and Wang et al. (2016) have a larger number of samples (n = 100 and n = 27), including a wide range of lakes featured with various climatic conditions, lake size, and depths. Secondly, the reconstructed-MAAT values of Sun et al. (2011) and Wang et al. (2016) are closer. Thirdly, the MAAT value of 18.0 Ma calculated by Günther et al. (2014) is much lower than the late Miocene long-chain alkenone temperature (18.9 ± 0.5°C) and lacustrine carbonate oxygen isotope temperature (19.0 ± 0.5°C) in the Dahonggou (DHG) section (Figures 1B, 5, Sun et al., 2012) and Xitieshan area of the QB (Figure 5, Wu et al., 2007), which is inconsistent with the global cooling trend during Middle Miocene to late Miocene (Figure 5, Wang et al., 2003). We therefore felt it fair to assume that 31.3 ± 4.3°C and 25.6 ± 1.5°C could be taken as reliable paleotemperature values. In order to reduce any errors in the paleotemperature calculations further, our final 18.0 Ma temperature took the mean value of 31.3 ± 4.3°C and 25.6 ± 1.5°C to be 28.4°C, with the error being calculated as half of the difference between 31.3°C and 25.6°C (i.e., 2.9°C) (Figure 5).

The climatic environment of the QB during the Miocene was quite different from that of today. A considerable corpus of fossils, sedimentary facies, and paleoclimatic evidence has pointed to this region experiencing a warm period at this time. For instance, pollen from the genus fupingensis, representative of the humid environment typical of a subtropical, warm/temperate zone, was found in the Miocene sediments of the Naoge section (Miao et al., 2016) in the QB. Large mammalian fossils from the QB such as rhinoceros and elephant can be dated to the middle to late Miocene (Wang et al., 2011). The discovery of Castor fiber fossils in the DHG section would suggest that the area was warm and humid around 13.0 Ma (Li and Wang, 2015). The discovery of a large number of Chalicothere fossils from the early Pliocene would also intimate that the QB was still a humid and treeless environment during that period (Chen et al., 2015). Additionally, during the early and middle Miocene in the QB, the development of algal limestone and the wide distribution of homogeneous fine-grained layers would indicate that a megalake dominated the Basin (Shou et al., 2003; Miao et al., 2011; Zhuang et al., 2011; Liu et al., 2014; Chang et al., 2015; Guo et al., 2018). Moreover, the hydrocarbon isotopes of leaf wax identified in the Honggouzi section in the western QB appear to show that the climate in this area was relatively warm and wet in the middle Miocene (Wu et al., 2019).

We also used the brGDGT-temperature calculation method employed in this study to recalculate the paleotemperatures of lacustrine sediments during the 9.5–5.1 Ma (mean age 7.3 Ma) period from the Huaitoutala (HT) section (Figure 1B; Zhuang et al., 2019). These calculations rendered a mean value of 17.8 ± 2.9°C. As shown in Figure 6A, the MAAT of the QB decreased from 28.4°C at 18.0 Ma to 17.8°C at 7.3 Ma (△T = 10.6°C), consistent with the significant late Miocene cooling trend recorded by the oxygen isotope record in deep-sea sediments (Zachos et al., 2008; Figure 6B).

In conclusion, numerous fossils, sedimentary facies, and paleoclimatic archives in the QB demonstrate the reliability of this study’s calculated paleotemperatures.

The temperature lapse rate (TLR) is an important indicator of altitude and gradient. The TLR for the TP derived from long-term, near-surface air temperatures recorded at meteorological stations ranges from 4.8°C/km to 6.6°C/km (Guo et al., 2015). Due to the influence of many less well-constrained parameters such as underlying surface conditions, elevation, atmospheric circulation, and humidity (Spicer and Yang, 2010), any assumption of any paleo-TLR remains uncertain. Therefore, 6.6°C/km was regarded as the upper limit of the TLR in this study in order that the lower limit of any change in the elevation of the QB since the early Miocene might be calculated; 4.8°C/km was taken as the lower limit of the TLR so as to estimate the upper limit of any change in elevation for the same period.

In addition to determining the TLR, another important step in GDGT-paleoelevation calculation is to find the reference point of the temperature change at “Zero Altitude.” After a large number of screenings, we found three reference points that were close to the latitude and age of this study site, with the temperature proxy being either long chain alkenones or GDGTs. The first reference point was ODP Site 1,010 (30°N, 118°W) in the Pacific Ocean (LaRiviere et al., 2012), which is relatively close to the QB, though the age of its base is 13.4 Ma, quite different from that of this study. The U_k ³⁷-temperature change since 13.4 Ma at this site was 14.0°C. The second reference point was ODP Site 608 (42.8°N, 23.1°W), relatively far away from the QB, but close in age to this study (with a range of 24–0 Ma). Since 18.8 Ma, the GDGT-calculated temperature at this site has changed by 13.6°C (Super et al., 2018, 2020). Reference point three was located in the coastal lowlands of Northwest Europe (51.5°N, 6°E). Although the site is far away from the QB, its basal age is 16.5 Ma, consistent with that of this study. It is also a terrestrial environment like the QB, meaning any temperature change should be closer (Donders et al., 2009) to that identified in this study. From 16.5 Ma to now, the GDGT-based temperature change at this site was 13.9°C, very similar to 14.0°C at ODP Site 1,010 and 13.6°C at ODP Site 608. We therefore felt it appropriate to select the Northwest Europe low elevation site in as a reference point.

However, nothing is known about the latitudinal change between the low elevation point in Northwest Europe and the HLG section since the early Miocene, which will inevitably cause uncertainty in the temperature change (△t). Besides, the difference of △t between the modern “Zero Altitude” point and the HLG section was calculated. The △t of the Brussels Meteorological Station (50.9°N, 4.5°E, 55 m asl) from 1973 to 2020 that is closest to the low elevation point of Northwest Europe is 3.1°C (https://en.tutiempo.net/climate/ws-64510.html). Additionally, the △t for the Lenghu Meteorological Station (38.8°N, 93.4°E, 2,771 m asl) nearest to the HLG section from 1973 to 2020 was calculated, and it is 2.5°C (https://en.tutiempo.net/climate/ws-526020.html). During 1973–2020, the difference in △t between the “Zero Altitude” point and the HLG section was 0.6°C. Furthermore, as the latitude pattern between the Northwest Europe and the QB in the early Miocene cannot be obtained, it is assumed that the value, twice of 0.6°C (1.2°C), is the error representing the latitudinal difference between the “Zero Altitude” point and the HLG section since the early Miocene.

In order to calculate the paleoelevation of the QB in the early Miocene, firstly, the MAAT of lake surface sediments in Lake Sugan, Hurleg, and Tuosu of the QB (raw data from Wang et al., 2016) was recalculated (Figure 1B), by using the conversion equations of brGDGT-temperature from Sun et al. (2011) and Wang et al. (2016). Besides, their mean value was 6.1°C as the MAAT in the modern QB. From the early Miocene to the present day, the variation in the low elevation MAAT value for Northwest Europe is 13.9°C. The modern MAAT of the QB is 6.1°C. Therefore, if the temperature change caused by uplift in the QB between the early Miocene and the present day were to be taken as ΔT = 28.4°C − 13.9°C − 6.1°C = 8.4°C, and the upper limit of the TLR as 6.6°C/km, then the figure of 8.4°C would be equivalent to an uplift of ∼1,273 m. When the TLR is 4.8°C/km, the 8.4°C would be equivalent to an uplift of 1,750 m in the study area since the early Miocene. We took a mean value of 1,512 m from the lower limit of 1,273 m and the upper limit of 1,750 m as the change in elevation of the QB since the early Miocene. The modern mean elevation of the QB is ∼3,000 m asl, meaning that the paleoelevation of the study area would have been < ∼1,488 m asl during the early Miocene (Figure 4C). The uncertainty in this modeling was estimated to be 738 m, derived principally from the uncertainty within the calibrations of Sun et al. (2011) and Wang et al. (2016) being 2.9°C, and the difference in temperature change (△t) between the low elevation point in Northwest Europe and the HLG section caused by latitude variation is 1.2°C.

As mentioned above, the MAAT in the QB from 18.0 Ma to 7.3 Ma decreased by 10.6°C. During this period, the temperature change in the lowlands of Northwest Europe was 5.6°C (Donders et al., 2009). When the TLR is taken as 6.6°C/km, the change in elevation would be 758 m. When the TLR is 4.8°C/km, the elevational change would be 1,042 m. We took a mean value of 900 m from the upper limit and the lower limit as the variation in the paleoelevation of the QB from 18.0 Ma to 7.3 Ma. If the paleoelevation of the QB at 18.0 Ma is taken to be ∼1,488 m asl, its paleoelevation during the late Miocene would have been ∼2,388 ± 738 m asl (Figure 6C).

Using our reconstructed paleoelevations for the QB during the early Miocene and other previously published data (Meng et al., 2001; Zhuang et al., 2019), we reconstructed the Cenozoic altitudinal history of the QB (Figure 6C). This would suggest that at the initiation of the QB in the Eocene, the Basin’s surface was close to the sea level because the QB was connected with the Tarim Basin (Meng et al., 2001). Our brGDGT-based paleoaltimetry showed that the QB was still a low-elevation region (∼1,488 m asl) until ∼18.0 Ma. Even at 7.3 Ma, the brGDGT-based recalculated paleoelevation (∼2,388 m asl) of the QB shows that the Basin’s modern elevation had not yet been obtained. Subsequently, the QB was uplifted by ∼600 m, reaching the present elevation of 3,000 m asl.

However, recent fossilized plant finds in the DHG section have suggested a much earlier date at which the QB attained its present height, i.e., during the Oligocene (Figures 1B, 6C; Song et al., 2020). The plant fossil assemblages belong to a temperate-deciduous-dominated woodland environment and were buried in fluvial sandstone beds. We assumed from this finding that these plants grew on the slopes of the Qilian Mountains and were transported by river action into the Basin, most probably during flooding. This assumption can be based on the analyses of pollen spores from the QB, which indicate that, during the Oligocene, arid-tolerant species of the genera Ephedra, Chenopodiaceae, and Artemisia dominated the Basin’s surface (Exploration and Development Research Institute of Qinghai Petroleum Administration Bureau., and Nanjing Institute of Geology and Paleontology, Chinese Academy of Sciences, 1985), while deciduous Ulmus tree species and fewer numbers of Picea and Pinaceae coniferous trees grew in the Qilian Mountains (Sun et al., 1980). This forest composition would suggest an elevational range of ∼2,000–3,500 m asl when compared with the nearest modern species (Mosbrugger et al., 2018). The elevations suggested by these tree fossils thus most likely reflect the paleoelevations of the Qilian Mountains, rather than the QB. Furthermore, leaf fossil assemblages dominated by Populus and Podocarpium were found in the DHG Gulley (Song et al., 2020). Of these, the same modern Populus indicates a temperate, semi-humid climate and grows mostly along riverbanks in the Lanzhou Basin and at an elevation of ∼1,400 m asl (Sun et al., 2004). This would imply that the fossilized Populus found in the DHG Gulley in the QB most likely grew in valleys during the Oligocene at an elevation of ∼1,400 m asl; this would not support an elevation of ∼3,000 m asl for the QB at that time. Moreover, fossilized plant leaves and fruits of genera Ailanthus found in the Huatugou area in the northwestern QB and dating from the early Oligocene lend further support to our inference (Yang et al., 2020), indicating that the northwestern QB was most likely at low elevations of 500–2,000 m asl during the Oligocene. The recent discovery of fossilized Paleoschizothorax vertebrates (Cyprinidae, a genus of freshwater fish living in warm water environments and at low elevations of ∼500–1,800 m asl in southwestern China) in the Oligocene stratigraphy in the same area would appear to further support the inference of a low-elevation QB during the Oligocene (Yang et al., 2018). The pharyngeal bones of the Schizothorax subfamily or Barbinidae found in the late Miocene strata of the HT area exist purely in the original Schizothorax subfamily and Barbinidae fish, and are only distributed in areas with an elevation of 1,250–2,500 m asl or lower (Chang and Miao, 2016), indicating that the QB did not reach its modern height until the late Miocene, consistent with this study.

The above topographical histories of the QB suggest that a slow rise of the NE TP, and a lowland environment, dominated the QB for much of the Eocene and Miocene, and that the rapid uplift of the Plateau’s surface to the present-day elevation of ∼3,000 m asl occurred from the late Miocene onwards (Figure 6C).

The reconstructed Cenozoic tectonic history of the QB shows a highly variable crustal shortening rate (Figure 6D; Bao et al., 2017), being high from 43.8 Ma to 35.5 Ma and from 15.3 Ma to 0 Ma, and low from 53.5 Ma to 43.8 Ma and from 35.5 Ma to 15.3 Ma. The Miocene represents the dividing lines between these high and low rates, with ∼50% of the Basin’s total shortening occurring by the onset of the Miocene.

The reconstructed Cenozoic paleoelevation evolution of the QB can also be confirmed by overwhelming geological evidence. For instance, a low topography in the QB during the Paleogene is indicated by the age of the last-remaining marine units in the area (Meng et al., 2001), which was connected with the southwestern Tarim Basin when the region was occupied by the Paratethys Sea between the Cretaceous and the early Paleogene (e.g., Popov et al., 2004). Additionally, Yin et al. (2008) suggested that the Paleogene (65–24 Ma) Qaidam and Hoh Xil basins on both sides of the Eastern Kunlun Mountains may have been parts of a single topographic depression called the “Paleo-Qaidam Basin,” which would imply that the QB experienced a process of slow uplift during the Paleogene. During the period from the early to middle Miocene, the sedimentology, paleomagnetism, thermochronology, and structural geology (Jolivet et al., 2003; Fang et al., 2007; Clark et al., 2010; Li B. et al., 2017) all point to a rapid denudation and uplift of the QB and its surrounding mountain ranges. This body of evidence is consistent with this study’s brGDGT-based paleoelevation reconstruction. A significant number of stratigraphic, petrographic, sedimentological, thermochronological, and climate records have also demonstrated that since the late Miocene, the Qaidam region and even the entire NE TP have undergone a period of intensive uplift, laying the foundations for the region’s present-day geomorphic patterns and elevation, and causing huge climatic and environmental change (Fang et al., 2005; Li et al., 2014).

In summary, all the aforementioned geological evidence agrees well with our reconstructed Cenozoic paleoelevation history of the QB on the NE TP, a reconstruction that supports a rapid late Miocene uplift and growth of the NE TP. However, due to the difficulty in preservation of the GDGTs, the interpretation and resulting conclusion in this study is based primarily on limited datasets. Further studies with more samples from a broad region can subsequently verify and reinforce the results from this study.