The Pamir Plateau constitutes a region of the NW Tibetan Plateau, with the SW Tian Shan and Alai Valley to the north. The Pamir Plateau is a prominent feature separating the Tarim and Tajik basins to the east and west, respectively (Figure 1B). The Pamir Plateau contains several terranes that were accreted onto Eurasia from the Late Paleozoic to the Mesozoic (Schwab et al., 2004; Robinson et al., 2012; Angiolini et al., 2013). From north to south, this plateau can be divided into the North Pamir, the Central Pamir and the South Pamir, which probably correlate with the southern block of the Songpan-Ganzi terrane, the Qiangtang terrane and the Lhasa terrane (Schwab et al., 2004; Cowgill, 2009).

The North Pamir region, bordered by the Main Pamir Thrust (MPT) to the north and the Tanymas suture to the south, is composed of predominantly Paleozoic and Triassic sedimentary rocks, metamorphic sedimentary and metamorphic volcanic rocks, Permian sedimentary and metamorphic volcanic rocks and Triassic turbidites intruded by Triassic-Jurassic granitoid rocks (Schwab et al., 2004; Schmidt et al., 2011; Ding et al., 2013). The Central Pamir region is bordered by the Rushan-Pshart zone to the south, and consists of Paleozoic and Triassic-Jurassic (meta) sedimentary rocks. Magmatic rocks are mainly from the Cretaceous (80–70 Ma), Eocene (42–36 Ma) and Miocene (20–10 Ma) (Schwab et al., 2004; Rutte et al., 2017; Chapman et al., 2018b). The South Pamir is bordered by the Wakhan tirich boundary to the south, which is composed of Paleozoic or Triassic-Jurassic sedimentary rocks and metamorphic sedimentary rocks, as well as Cretaceous and Cenozoic magmatic and metamorphic rocks (Schwab et al., 2004; Blayney et al., 2016).

The SE Pamir Plateau connects with the West Kunlun and forms the Pamir-West Kunlun transition zone. This region is dominated by Precambrian and Paleozoic sedimentary and metasedimentary rocks, with abundant Paleozoic and Triassic intrusive rocks (Blayney et al., 2016). The West Kunlun is separated from the Tarim basin by the Kashi-Yecheng transfer system (KYTS) and the Tiklik fault to the north, and is divided into the northern and southern subterranes by the Tam Karaul thrust (Kudi suture zone). This region is mainly composed of Precambrian and Paleozoic metamorphic sedimentary rocks, and Paleozoic intrusive rocks (Cowgill et al., 2003; Liu et al., 2015; Schwab et al., 2004; Xiao et al., 2002, 2003). The active left-lateral Karakax fault separates the West Kunlun from the Songpan-Ganzi terrane, which consists predominantly of Triassic sedimentary rocks and Jurassic-Triassic intrusive rocks (Yin and Harrison, 2000). The Jinsha suture zone separates the Tianshuihai terrane from Songpan-Ganzi, and consists predominantly of Triassic-Cretaceous (meta) sedimentary rocks overlying the Paleozoic basement (Matte et al., 1996; Schwab et al., 2004; Robinson et al., 2012).

The Tarim Basin, located between the Tian Shan (to the north), the West Kunlun Mountains (to the south), the Pamir Plateau (to the west) and the Altyn Tagh Mountains (to the east), is a large rhomb-shaped geomorphic feature close to the NW Tibetan Plateau. Large amounts of erosion materials from the surrounding mountain belts have filled the Tarim Basin, creating >10 km thick Mesozoic and Cenozoic strata (Yang and Liu, 2002).

The Cenozoic strata have been divided into the Paleogene Kashi, the Miocene Wuqia groups, and the Atushi and Xiyu Formations. The Kashi group (E_k) is composed of the Aertashi (E₁ ^a), Qimugen (E_1-2 ^q), Kalatar (E₂ ^k), Wulagen (E₂ ^W) and Bashibulake (E_2-3 ^b) Formations in chronological order, which record a series of marine transgressions-regressions (Bosboom et al., 2014; Sun et al., 2016a).

Based on dating a volcanic ash bed as ∼11 Ma (Zheng et al., 2015a), the stratigraphic age has been reclassified, although whether it is a volcanic ash bed remains debated (Zheng et al., 2015b; Sun et al., 2015). Assigned ages are >∼41 Ma for the Aertashi to Wulagen Formations, ∼41 to 36.5 Ma for the Bashibulake Formation, ∼33.0 to 22.6 Ma for the Wuqia Group (including the Keziluoyi, Anjuan and Pakabulake Formations), ∼22.6 to 15 Ma for the Artux Formation, and <∼15 Ma for the Xiyu Formation (Zheng et al., 2015a).

The Keliyang section is located near the southern margin of the Tarim Basin. Drilling data show that the core of the Keliyang anticline is composed of Jurassic and Cretaceous sediments (Cheng et al., 2011). The Paleogene strata are overthrust by Mesozoic strata. The Mesozoic-Cenozoic strata of the whole anticline are relatively upright, due to passive uplift (Cheng et al., 2011). The Keliyang section (from 37°16′12.33″N, 77°51′42.51″E to 37°18′47.82″N, 77°51′36.08″E) has a thickness of about 5–6 km and is mainly composed of Jurassic, Cretaceous, Paleogene and Neogene strata, of which the Paleogene strata are the best exposed (Figure 2). All the Paleogene strata show structural inversion with a dip toward the SE at angles of 75–85°, and are mainly composed of delta, fluvial and lacustrine facies sediments (Figure 3).

We did not make a detailed grouping of strata, to avoid the strong debate on the chronologies of the Cenozoic strata. Based on the sedimentary environment and field observation, the Keliyang section can be roughly divided into three units of terrestrial facies, marine-terrestrial interbedded facies, and lacustrine-delta facies (Sun et al., 2016a). The lower part of this section consists of purplish-red gravelly sandstone, which is interpreted as the terrestrial facies sediments. The middle part consists of interbedded gray limestone, gray-green mudstone and gypsum, which is interpreted as marine-terrestrial interbedded facies. The upper part is dominated by orange-red sandstone and yellow sandstone, is which interpreted as lacustrine and delta facies. The last marine regression in the Tarim Basin occurred at ∼ 40 Ma (Bosboom et al., 2014; Sun et al., 2016a; Kaya et al., 2019). To gain sedimentary provenance information in this section, we collected four samples to represent before (KLY-1), during (KLY-3), and after (KLY-6 and KLY-7) Paratethys regression. These samples were analyzed using detrital apatite fission track and/or zircon U/Pb dating methods (Table 1).

We use previous high resolution magnetostratigraphy (Sun et al., 2016a) to quantify the depositional ages of KLY-3 and KLY-6 as ∼40 and ∼32 Ma, respectively (Table 1). According to the true thickness (between KLY-3 and KLY-1) and average sedimentary deposition rates (for the bottom segment of the magnetostratigraphic profile), KLY-1 is estimated to have been deposited at ∼56 Ma (Figure 3 and Figure 4). However, the lithology of KLY-1 is consistent with Upper Cretaceous gravel-bearing coarse sandstone strata in the SW Tarim Basin (Si et al., 2007; Sun et al., 2016b; Chen et al., 2018; Li et al., 2021), and the uncorrected magnetostratigraphic age may be affected by inaccuracies in the sedimentation rate or sedimentary discontinuities. On the other hand, detrital zircon U/Pb ages of KLY-3 and KLY-7 yielded the youngest single zircon ages of 43.7 Ma (n = 1) and 23.1 Ma (n = 1). The lag time between the youngest detrital zircon U/Pb age (43.7 Ma) and the magnetostratigraphic depositional age (40 Ma) is roughly 4 Ma, which is a reasonable time frame from crystallization to deposition. The KLY-7 sample was not included in previous magnetostratigraphic profiles, but the youngest detrital single zircon U/Pb age can be used to determine its depositional age (Dickinson and Gehrels, 2009). The youngest detrital zircon U/Pb age of KLY-7 was ∼23.1 Ma, so KLY-7 is assigned a depositional age younger than 23.1 Ma (Figure 4).

Detrital apatite fission-track (DAFT) measurements were carried out using the external detector method. Apatite was separated from rock samples by magnetic and gravity separation, embedded in epoxy resin, polished to expose internal crystal surfaces, and etched in 5 N HNO₃ at 20°C for 20 s to reveal spontaneous fission tracks. Internal surfaces of the crystals were then covered with low-uranium muscovite external detectors, packed together with CN5 standard dosimeter glasses, and irradiated. Induced tracks were revealed in the muscovite external detectors by etching in 40% HF at room temperature for 25 min. DAFT analyses were counted at ×1,250 dry (×100 objective). FT ages were calculated using the ξ-calibration method, with an overall weighted mean ξ of 272.78 ± 15.99 a/cm². DAFT analyses were performed at the Key Laboratory of Petroleum Resources, Northwest Institute of Eco-Environment and Resources at the Chinese Academy of Sciences.

After crushing, zircons were separated by standard heavy liquid and magnetic techniques. Zircon grains were randomly picked and mounted onto adhesive tape, enclosed in epoxy resin, and polished to about half their thickness. After being photographed under reflected and transmitted light, samples were prepared for cathodoluminescence (CL) imaging and U/Pb dating. Zircon U/Pb dating was performed using an Agilent 7500a ICP-MS equipped with a 2005M excimer ArF laser ablation system (GeolasPlus) at the China University of Geosciences (Wuhan). A laser spot diameter of 30 μm was used. The ICP-MS data calibration (10.7) program was used for data calibration.

The ²⁰⁷Pb/²⁰⁶Pb ages were used to date zircons older than 1,000 Ma, and ²⁰⁶Pb/²³⁸U ages were used to date zircons younger (Black et al., 2003) than 1,000 Ma, due to small amounts of ²⁰⁷Pb in young zircons limiting precise ²⁰⁷Pb/²⁰⁶Pb dating. To assess the similarity of samples and their potential sources, we use multi-dimensional scaling (MDS) (Vermeesch, 2013) to produce a map of points where similar samples cluster together, which quantifies the distance between the empirical cumulative distribution functions of two samples. The plot axes are non-quantitative.

KLY-1 contains 59 grains with relatively scattered ages ranging from 35.27 to 202.73 Ma, and a central age of 74.7 ± 3.3 Ma. KLY-6 has 33 grains with ages between 29.16 and 876.64 Ma, and a central age of 50 ± 4.4 Ma. KLY-7 has 24 grains with ages between 23.7 and 314.1 Ma, and a central age of 68.6 ± 9.6 Ma.

The DAFT ages of all samples failed the χ² test (P (χ²) <5%) (Table 2 and Figure 5). The result of KLY-1 can be statistically decomposed into three age components: 45.8 ± 4 Ma (21%), 69.1 ± 6.4 Ma (39%) and 96.8 ± 8.5 Ma (40%). The result of KLY-6 can be divided into two components with peak ages at 35.7 ± 2.7 Ma (64.5%) and 80.5 ± 7.1 Ma (35.3%). The result of KLY-7 can be divided into two components with peak ages of 25.9 ± 3.4 Ma (P1) and 80.6 ± 7.2 Ma (P2).

Detrital zircon U/Pb ages are shown in Supplementary Table S1 and Figure 6. A total of 116 concordant ages were obtained for KLY-1. The zircon U/Pb age spectrum shows age populations between 400 and 2,800 Ma, the majority around ∼400–500 Ma (∼81%) with a peak at ∼480 Ma. A smaller age population is observed at ∼600–900 Ma, with peaks at ∼640 Ma and ∼800 Ma. In addition, sporadic distributions are centered at ∼1800 and ∼2,600 Ma.

A total of 108 concordant ages were obtained for KLY-3, with the youngest age at 43.7 Ma. The zircon U/Pb age spectrum shows age populations between 44 and 3,200 Ma. Ages range between 0 and 100 Ma, with a peak at ∼40 Ma; ∼240–500 Ma with peaks at ∼280 Ma and ∼320 Ma; and ∼560–720 Ma with peaks at ∼580 and ∼640 Ma. A few additional ages are scattered at roughly 1,200 Ma, 1,600 Ma, 2,400 Ma and 3,200 Ma.

A total of 119 concordant ages were obtained for KLY-7, with the youngest age at 23.1 Ma. The zircon U/Pb age spectrum shows age populations between 23.1 and 2,800 Ma. These are concentrated between 0 and 80 Ma with peaks at ∼40 and ∼20 Ma; 200–300 Ma with a peak at ∼240 Ma; 400–500 Ma with a peak at ∼440 Ma; and 600–1,200 Ma with a peak at ∼800 Ma. A minor age population occurs between 1,200 and 2000 Ma, and there are some scattered ages between ∼2,200 and 2,800 Ma.

Geologists compare detrital zircon U/Pb ages between basin sediments and closed blocks to deduce potential provenance regions (Cawood et al., 2012; Carrapa et al., 2015; Sun et al., 2016b, 2020; Blayney et al., 2016; Chapman et al., 2019; Zhang et al., 2019). As discussed above, the main paleocurrents in the SW Tarim Basin both came from the south, with sources including the Pamir Plateau, the West Kunlun Mountains and the Songpan-Ganzi terranes. Furthermore, these different regions have different geochronological characteristics (Figure 8). The North Pamir region is dominated by the 200–300 Ma peak in detrital zircon U/Pb ages and does not show a 400–500 Ma peak (Schwab et al., 2004; Lukens et al., 2012; Carrapa et al., 2015). The Central Pamir region has a typical peak of ∼45 Ma and does not have a peak older than 400 Ma (Lukens et al., 2012). The South Pamir region detrital zircon U/Pb ages are mainly distributed around ∼100 Ma (Yin and Harrison, 2000; Lukens et al., 2012). The West Kunlun has two peaks, at ∼200–300 Ma and 400–500 Ma, and two less prominent peaks at ∼800 Ma and ∼1800 Ma (Robinson et al., 2004; Carrapa et al., 2014; Rittner et al., 2016). The Sonpan-Ganzi has four peaks, at ∼200–300 Ma, ∼400–500 Ma, ∼700–900 Ma and ∼1800–2000 Ma (Ding et al., 2013).

The distribution of zircon U-Pb ages of sample KLY-1 is more similar to those of the West Kunlun and the Songpan-Ganzi terranes than those of the Pamir Plateau (Figure 6). The two contemporaneous rivers in this region (the Tiznip and Karakax) provide more information on the source of KLY-1. The Tiznip River originates from the south of the West Kunlun and flows northwards into the Tarim Basin via the northern West Kunlun. The Karakax River originates from the Sonpan-Ganzi and flows northward into the Tarim Basin via the West Kunlun. The detrital zircon U/Pb ages of Tiznip River sediments have two peaks, at ∼200–300 and ∼400–500 Ma, while detrital zircon U/Pb ages of the Karakax River have a relatively shorter peak at 200–300 Ma (see further details in Blayney et al., 2016). Sample KLY-1 lacks ages of ∼200–300 Ma, and thus shows greater similarity to the Karakax River (Figure 6). The MDS analyses also show that KLY-1 is much closer to sample 1,363 from the Karakax River when compared to sample 1,344 from the Tiznip River (Figure 7). These observations suggest that sediments in KLY-1 may have had a multi-component source, from the Sonpan-Ganzi and the West Kunlun. Previously published results from other sections in the SW Tarim Basin before the Eocene show that detrital zircon U/Pb ages have two peaks, at ∼200–300 and ∼400–500 Ma, which should correspond to sources in the Sonpan-Ganzi, the West Kunlun and/or the North Pamir (Figure 8). In the Dashtijum section of the Tajik Basin, detrital zircon U/Pb ages of the Late Cretaceous sample have a main peak of ∼200–300 Ma, which likely corresponds to a source region in the North Pamir (Chapman et al., 2019) (Figure 8). The detrital zircon U/Pb ages of the sample DZ-04 in the Asku section are more similar to the Pamir region than to the Tian Shan (Figure 11 in Sun et al., 2020).

The obvious zircon U/Pb age peak at ∼45 Ma is similar to the Central Pamir rather than other terranes (Figure 8). Moreover, MDS analyses show that the detrital zircon U/Pb ages of KLY-3 and KLY-7 are more similar to the Central Pamir, than to the North or the South Pamir (Figure 7). The detrital zircon U/Pb age and εHf (t) analyses also provide additional evidence for the Central Pamir as the sediment source (Zhang et al., 2019). Sun et al. (2016b) suggested that the Early Eocene sedimentary change originated from the Kohistan-Ladakh arc. The detrital zircon U/Pb age and εHf (t) data from the Kohistan-Ladakh arc and the Central Pamir overlap to some extent, but further research indicated that the Early Eocene sediments originated from the Central Pamir rather than the Kohistan-Ladakh arc (Zhang et al., 2019). In addition to constraining the sedimentary source from the Central Pamir, the other sediment sources in the Keliyang section after the Eocene were likely located in the West Kunlun (Figure 6 and Figure 8). Previous research also confirms that other sources were derived from the West Kunlun and/or the North Pamir in the Tajik and SW Tarim basins (Chapman et al., 2019; Zhang et al., 2019).

Based on the evidence and discussion above, Eocene sediments with the detrital zircon U/Pb age peak at ∼45 Ma were derived from the Central Pamir. This result requires rapid uplift of the Central Pamir. Following this uplift, clasts were eroded from the new uplands of the Central Pamir.

These interpretations are supported by other evidence. Low-temperature thermochronology is widely used to reconstruct fast exhumation of mountains (Bernet et al., 2006). Unfortunately, no Eocene thermochronology results are available from within the Central Pamir, perhaps due to the overprinting of later tectonic events (Ducea et al., 2003), complete erosion (Zhang et al., 2019) or non-detection (Zhang et al., 2019). However, the Late Eocene (∼37 Ma) component of detrital apatite fission-track ages is detected in the Keliyang section of the SW Tarim Basin (this study) and in the Peshtova section of the Tajik Basin (Wang et al., 2019) (Table 2), where sediments were likely sourced from the Central Pamir. Moreover, abundant Eocene igneous rocks from 41 Ma to 36 Ma (with a peak age of 40 Ma) were reported within the Central Pamir, and were interpreted as a result of mantle drip or lithospheric delamination (Chapman et al., 2018b). Contemporaneous metamorphic peak ages are also detected within the Central Pamir domes (Smit et al., 2014; Stearns et al., 2015; Rutte et al., 2017; Chapman et al., 2018a). The prograde metamorphic monazite age obtained for lower-crustal xenoliths in the Miocene volcanic rocks is 50 Ma, which indicates that crustal thickening and plateau formation were already occurring during the Paleocene–Eocene in the Central Pamir (Ducea et al., 2003). Furthermore, the very short lag time between the youngest detrital apatite fission-track age component (∼36 Ma) and the deposition age (∼32 Ma) of the Keliyang section indicates that the Central Pamir experienced rapid uplift during the Eocene, caused by crustal thickening, which yielded large amounts of sediment to the Tarim and Tajik basins (Blayney et al., 2019; Kaya et al., 2019).

The Eocene rapid uplift event is also confirmed in other regions along the strike of the Central Pamir. Apatite fission-track and U-Th/He ages of the Qiangtang terrane are concentrated at around 40 Ma, suggesting that the Qiangtang terrane formed a plateau during the Eocene (Rohrmann et al., 2012). Paleoaltimetry shows that the Gonjo Basin within the Qiangtang terrane experienced rapid uplift during the Eocene (Xiong et al., 2020). Moreover, this Eocene rapid uplift event was also reconstructed in the North Pamir, based on bedrock apatite and zircon U/Th-He ages (Amidon and Hynek, 2010). In contrast, the West Kunlun experienced slow exhumation, with a paleo-elevation similar to that during the Mesozoic (Sobel and Dumitru, 1997; Cao et al., 2015; Blayney et al., 2016; Li et al., 2019). During the Eocene, the Central Pamir was uplifted to a paleo-elevation no lower than that of the present-day West Kunlun, and provided sediments to the SW Tarim Basin.

We provide an evolution model for the Pamir-West Kunlun area from the Late Cretaceous to the Late Eocene. During the Late Cretaceous (Figure 9A), the West Kunlun, with moderate paleo-elevation, was the main sedimentary source to the SW Tarim Basin, and the North Pamir was the main source to the Tajik Basin. The Central Pamir had not formed, and was connected with the Paratethys Ocean. During the Late Eocene (Figure 9B), the northward indentation of the Pamir region caused deformation at the western and eastern margins, and the Central Pamir region experienced rapid uplift to reach a paleo-elevation higher than the modern West Kunlun. Sedimentary materials from the newly-formed Central Pamir were transported by rivers into the Tarim and Tajik basins.