Towards a clarification of the provenance of Cenozoic sediments in the northern Qaidam Basin Open Access

In Tibet, several ranges with upthrusted basement (Gangdese, Tanggula, Kunlun, Altyn Tagh, and Qilian Shan) are separated by numerous large and small Tertiary basins trending approximately E-W to NW-SE (Liu, 1988; Pan et al., 2004; Tapponnier et al., 2001; Wang et al., 2013; Yin and Harrison, 2000; Yin, 2010). The sedimentary archives of these intermontane basins can be used to constrain the processes and mechanisms of the evolution of Tibet (Metivier et al., 1998; Yin et al., 2002). Various indicators in sedimentary basins in southern Tibet, such as the termination of marine facies sediments, the initial deposition of material derived from the Asian plate on the Indian margin, and the earliest strata to include mixed India- and Asia-sourced detritus, have been used to provide minimum constraints on the timing of the initial India-Asia collision (Garzanti et al., 1987, 1996; Najman, 2006). Sedimentologic, geochronologic, and structural analyses of Cretaceous–Tertiary basins in central Tibet have indicated multiple-phase contractional deformation since the India-Asia collision (Horton et al., 2002; Kapp et al., 2007; Li et al., 2017; Spurlin et al., 2005; Staisch et al., 2014). Further northwards, in the Qaidam Basin, there are exceptionally thick (average of 6–8 km) sequences of Cenozoic siliciclastic sediments (Rieser et al., 2006a), which have been extensively used to reconstruct depositional systems (Bush et al., 2016; Jian et al., 2013; Meng and Fang, 2008; Rieser et al., 2005; Wang et al., 2006; Zhu et al., 2006; Zhuang et al., 2011), erosional unroofing patterns (X. Cheng et al., 2016; Rieser et al., 2006a, 2006b; Wang et al., 2017; Y. Wang et al., 2015), and kinematic histories of fold-and-thrust belts and strike-slip faults along the northern, western, and southern margins of the intermontane basin (F. Cheng et al., 2016b; Fu et al., 2015; Mao et al., 2016; Meng et al., 2001; Ritts et al., 2004, 2008; Wu et al., 2012; Yin et al., 2007a, 2008a, 2008b; Yue and Liou, 1999; Yue et al., 2001; Zhang et al., 2016, 2018).

Recent integrated analyses of seismic reflection profiles and thickness distributions of Cenozoic strata across the Qaidam Basin indicated that the basin structure is predominantly a broad Cenozoic synclinorium, and hence the main depocenter lies persistently along the northwest-trending central axis of the basin (Yin et al., 2008a; Zhu et al., 2006). The spatial distribution of Cenozoic strata in the Qaidam Basin indicates that the sediments are supplied by an endorheic drainage system with sources in several ranges: in the northern basin, mainly the southern Qilian Shan (Fig. 1A); in the western basin, mainly the Altyn Tagh Range; in the southwestern basin, mainly the Altyn Tagh Range and Qimen Tagh; and in the southern basin, mainly the East Kunlun Shan. This sediment dispersal model has been widely substantiated by paleocurrent analyses within the basin (Heermance et al., 2013; Ji et al., 2017; Jian et al., 2018; Meng and Fang, 2008; Song et al., 2013; Wu et al., 2012; Zhuang et al., 2011). However, two recent provenance studies, together with detrital zircon U-Pb and apatite fission-track dating, have challenged the widely held view by suggesting that several major paleorivers flowed across the basin axis. Specifically, Bush et al. (2016) identified E-directed paleocurrent orientations for the Lulehe Formation and the lower part of the lower Ganchaigou Formation from the Dahonggou section in the northern Qaidam Basin and suggested that the distant westernmost East Kunlun Shan (Qimen Tagh) was the principal sediment source feeding the E-directed fluvial system (Fig. 1A). Wang et al. (2017) measured NE-directed paleocurrent orientations for the Lulehe Formation and upper and lower Ganchaigou Formations from the same section and suggested that the East Kunlun Shan was the dominant provenance source (Fig. 1A), which was subsequently replaced by the southern Qilian Shan. The controversy centers on a dispute over the provenance of the Lulehe Formation strata in the northern Qaidam Basin. It is noteworthy that these two long sediment dispersal pathways require linear distances of ∼200 km and ∼350 km, respectively, between source and sink (Fig. 1A). These values represent only lower limits, since Cenozoic shortening strain across the Qaidam Basin has not yet been considered; notably, the magnitude of Cenozoic shortening strain increases systematically westward from ∼11% in the center to ∼35% in the west (Yin et al., 2008b).

It is very important to validate these discrepant provenance models, because they indicate significant topographic relief and mountain building for different orogenic belts. Most studies favoring roughly S-directed paleocurrents in the northern Qaidam Basin have linked the synorogenic coarse-grained Lulehe Formation with rapid uplift of the southern Qilian Shan at 65–50 Ma (Ji et al., 2017; Yin et al., 2008a; Zhuang et al., 2011). However, Bush et al. (2016) attributed the Lulehe Formation and the lower part of the lower Ganchaigou Formation to uplift-induced exhumation of the Qimen Tagh, whereas Wang et al. (2017) concluded that the East Kunlun Shan emerged as high relief after the deposition of the Lulehe Formation, supplying large amounts of detritus to the northern margin of the Qaidam Basin. The high-topography hypothesis for the East Kunlun Shan and Qimen Tagh during the sedimentation of the Lulehe Formation is, however, inconsistent with recent paleontological studies, which show that the Paleogene and early Neogene strata of the western Qaidam Basin contain many reefal deposits, representing lacustrine sedimentary facies (Zhong et al., 2004). The low-gradient depositional environment indicates no substantial surface uplift of the surrounding mountains.

Because the northern Qaidam Basin is a composite structure including diverse rock types produced in various ways by diverse geodynamic processes (Yang, 2002; Yin et al., 2007b), its orogenic detritus represents a variety of signatures. Unraveling the provenance of clastic wedges accumulated in the northern Qaidam Basin therefore requires sophisticated and integrated investigations. Numerous methods have been used to investigate the potential source areas of Cenozoic sedimentary rocks in the northern Qaidam Basin, including counts of conglomerate clasts (Zhuang et al., 2011), sandstone modal analysis (Bush et al., 2016; Jian et al., 2013; Rieser et al., 2005), U-Pb detrital geochronology (Bush et al., 2016; Wang et al., 2017), paleocurrent analysis (Bush et al., 2016; Ji et al., 2017; Jian et al., 2018; Song et al., 2013; Wang et al., 2017; Zhuang et al., 2011), heavy mineral analysis (Bush et al., 2016; Jian et al., 2013), ⁴⁰Ar/³⁹Ar dating of detrital white mica (Rieser et al., 2006a), and detrital apatite fission-track dating (Wang et al., 2017). Most of these methods, however, only provide indirect inferences for provenance reconstruction. For instance, Cenozoic sandstone samples within the northern Qaidam Basin exhibit a heavy mineral assemblage associated with magmatic and metamorphic basements (e.g., pyroxene, hornblende, epidote, and garnet), potentially derived from both the Kunlun Shan and Qimen Tagh and the northern Qaidam Basin margin and the southern Qilian Shan (Bush et al., 2016; Jian et al., 2013). Thus, heavy mineral analysis is not an unequivocal indicator of provenance in the northern Qaidam Basin. In addition, neither the framework grain composition of sandstones nor the geochemical composition of mudstones exhibits significant variation within the Cenozoic formations of the northern Qaidam Basin (Bush et al., 2016; Jian et al., 2013; Rieser et al., 2005). Thus, they do not enable identification of a specific source terrane. These issues are partly attributed to the orogenic recycling effect of older sedimentary units, which complicates potential provenance analyses. However, this effect has been ignored in recent studies (e.g., Wang et al., 2017).

Paleocurrent measurements can provide direct information about the orientation of the sedimentary systems and the flow directions of rivers. As a ubiquitous, chemically and physically resistant detrital mineral derived from multiple rock types, zircon is ideal for geochronological studies (Carter and Moss, 1999), and U-Pb zircon dating has been extensively used to discriminate potential sediment sources in the Tibetan Plateau (DeCelles et al., 2014; Gehrels et al., 2011; Wang et al., 2011; Wu et al., 2014; Zhuang et al., 2015). In this study, we obtained sedimentological and sedimentary petrological data from the Lulehe Formation in the Dahonggou locality, and we measured paleocurrent orientations for the Lulehe Formation at four localities; we also conducted detrital zircon U-Pb geochronologic analyses of Mesozoic sandstones from five localities in the northern Qaidam Basin. Combined with existing paleocurrent and seismic reflection data, our results strongly indicate that the provenance of the Lulehe Formation is the northern Qaidam Basin margin and the southern Qilian Shan and that the two distant sediment source models are not supported by geological observations.

Northern Tibet is characterized by several of the most prominent tectono-geomorphological features developed during the Cenozoic, including the Qilian Shan, the Altyn Tagh Range, the East Kunlun Shan, and the Qaidam Basin (Fig. 1A). These mountains have an average elevation of ∼5000 m, with peaks exceeding 7000 m, whereas the intervening Qaidam Basin has an average elevation of ∼2800 m, with little internal relief (Yin et al., 2007a). The transition zones between these mountains and the Qaidam Basin yield the most extensive topographic front (>500 km long) and the greatest relief (>2.5 km on average) within the plateau. The ∼120,000 km², rhombus-shaped Qaidam Basin is tectonically bounded by the Qilian Shan–Nan Shan and northern Qaidam thrust belts to the north, the sinistral Altyn Tagh fault to the west, the Eastern Kunlun thrust belt to the south, and the dextral Wenquan fault to the east. The trend of the northern Qaidam Basin is parallel to the northern Qaidam thrust belts and the regional trend of the southern Qilian Shan. Several secondary mountain ranges are located in the northern Qaidam Basin, including, from west to east, the Saishenteng Shan, Lvliang Shan, Xitie Shan, Aimunik Shan, and Olongbluk Shan.

Tectono-stratigraphically, the Qilian Shan–Nan Shan region consists of three belts, composed mainly of Proterozoic shallow-marine strata in the central part sandwiched between the North Qilian complex (mainly Lower Paleozoic mélange, turbidites, and arc-type volcanic rocks) and the South Qilian metamorphic belt (Upper Proterozoic–Lower Paleozoic metamorphic rocks; Gehrels et al., 2003a, 2003b; Yin et al., 2007b). The crystallization ages of the late Proterozoic–Mesozoic plutons are confined to two groups in the Qilian Shan and Nan Shan (Fig. 2A; Cowgill et al., 2003; Gehrels et al., 2003a, 2003b): (1) 960–920 Ma small and isolated plutons, and (2) 520–400 Ma plutons from across the entire Qilian Shan and Nan Shan. Early Paleozoic arc magmatism was contemporaneous and spatially overlapped with ultrahigh-pressure (UHP) metamorphism in the northern Qaidam Basin (Yang, 2002). The bedrock of the Eastern Kunlun Range can be subdivided into three types (Yin et al., 2007a): (1) Precambrian gneiss, Neoproterozoic metasediments, and Devonian–Carboniferous marine strata; (2) widespread Paleozoic and early Mesozoic igneous rocks (Figs. 2A and 2B); and (3) Jurassic to Cenozoic terrigenous clastics. Two main phases of plutonism are constrained to the Ordovician–Silurian and Permian–Triassic (Liu, 1988; Wu et al., 2016). The Altyn Tagh Range exposes Paleoproterozoic migmatite, gneiss, and schist, and Mesoproterozoic carbonates intruded by Proterozoic–Mesozoic granitoids (Chen et al., 2012; Cheng et al., 2017; XBGMR, 1993). The bedrock of the Qaidam Basin is composed of Precambrian–Silurian metamorphic rocks (Huang et al., 1996); in addition, one phase of prominent arc magmatism related to the subduction of the Paleo-Tethys Ocean lasted from 290–280 Ma to 215 Ma across the Kunlun-Qaidam terrane (Fig. 2C; Chen et al., 2012; Wu et al., 2016). Although the Permian–Triassic magmatism is widely exposed along the Kunlun Range, it only occurs as isolated plutons in the northern Qaidam Basin (Fig. 2B; Chen et al., 2012; Cheng et al., 2017; Wu et al., 2016). U-Pb zircon dating of granitoid samples from drill cores also testifies to the existence of a Permian–Triassic Neo-Kunlun arc in the northern and southern Qaidam Basin (Fig. 2B; Cheng et al., 2017). The earliest phase (2.5–2.3 Ga) of magmatism was identified in the Quanji Massif, in the northern Qaidam Basin (C. Wang et al., 2015).

Besides the early Paleozoic UHP metamorphic gneiss in the Lvliang Shan, Xitie Shan, and Dulan areas along a 450-km-long belt in the northern Qaidam Basin (Yang, 2002; Yin et al., 2007b), Jurassic and Cretaceous sedimentary rocks crop out discontinuously along the margin of the northern Qaidam Basin from Lenghu to Dameigou (Figs. 1 and 2; Ritts and Biffi, 2001; Yu et al., 2017). The ages of Mesozoic strata have been determined predominantly by paleontological studies and lithostratigraphic correlation (Ritts and Biffi, 2001, and references therein). Jurassic and Cretaceous strata are entirely nonmarine, nonvolcanogenic sedimentary rocks, and their lithology grades from boulder conglomerates to laminated shales (Ritts and Biffi, 2001). The depositional environments of these sedimentary rocks ranged from alluvial fans through coarse- and fine-grained fluvial systems, to deep, open lakes (Ritts and Biffi, 2001). Detrital zircon age data indicate that the Jurassic sedimentary rocks have two major age peaks of ca. 250 Ma and ca. 2400 Ma, and two minor peaks of ca. 450 Ma and ca. 850 Ma (Yu et al., 2017). This age distribution pattern is similar to that of the East Kunlun Range, which may complicate potential provenance analyses of Cenozoic strata in the northern Qaidam Basin. The inferred tectonic setting of Mesozoic rocks on the margin of northern Qaidam Basin remains disputed, with most researchers arguing for a predominantly extensional setting (Chen et al., 2003; Wu et al., 2011; Xia et al., 2001; Yu et al., 2017), and a few for a compressional setting (e.g., Ritts and Biffi, 2001). The provenance of these strata is also highly controversial, with one view arguing for the Qilian Shan (Ritts and Biffi, 2001) and the other for the East Kunlun Shan (Yu et al., 2017).

Cenozoic deposits with a maximum thickness of 12,000 m and an average thickness of 6–8 km are widely distributed across the Qaidam Basin. Seismic reflection data indicate that the Qaidam Basin is a broad Cenozoic synclinorium with its depocenter located persistently along the northwest-trending central axis of the basin (Meng and Fang, 2008; Yin et al., 2008a; Zhu et al., 2006). The depocenter has shifted progressively southeastward during the Cenozoic, from an initial position in the northwestern basin to its current depocenter at Dabuxun Lake near the basin center (Yin et al., 2008a; Zhu et al., 2006). A composite structural model was proposed to account for the spatio-temporal evolution of the sedimentary architecture in the Qaidam Basin, including south-directed thrusts carrying the low-elevation Qaidam Basin over the high-elevation Eastern Kunlun Range along the southern margin of the basin and crustal-scale triangular zones tapering from the northern margin toward the basin interior (Yin et al., 2007a, 2008a, 2008b). This model may also explain the observations that, whereas pre-Quaternary basin fills are widely exposed in the northern Qaidam Basin, they are rare in the southern Qaidam Basin. Traditional provenance analyses of sandstone petrography, heavy minerals, paleocurrents, and conglomerate counts usually indicate that Cenozoic sediments adjacent to basin margins are directly linked to the adjacent mountain belts (Hanson, 1999; Jian et al., 2013; Wu et al., 2012; Zhuang et al., 2011).

Cenozoic outcrops are thickest in the Dahonggou and Lulehe localities in the northern Qaidam Basin (QBGMR, 1984), and they show an initial upward-fining and then an upward-coarsening trend (Bush et al., 2016; Fang et al., 2007; Ji et al., 2017; Lu and Xiong, 2009; Wang et al., 2017). This exceptionally thick succession has been divided into seven formations, including, in ascending order, Qigequan, Shizigou, upper Youshashan, lower Youshashan, upper Ganchaigou, lower Ganchaigou, and Lulehe. The Lulehe Formation at the Dahonggou section consists of poorly sorted, clast-supported, granule-cobble conglomerates interbedded with medium- to coarse-grained, trough cross-stratified sandstones representing an alluvial-fan and braided fluvial system. The lower Ganchaigou Formation is dominated by interbedded sandstone, or siltstone and mudrock deposits, and it represents a fine-grained fluvial system. The upper Ganchaigou Formation is predominantly composed of the interbeds of laminated or massive mudstones and ripple-laminated siltstones and likely also represents a fine-grained fluvial system. The lower Youshashan Formation consists of alternating brown laminated or bedded mudstone and gray-green massive sandstone, or conglomerate, or siltstone, and it represents a coarse-grained fluvial system. The upper Youshashan Formation is mostly composed of interbedded conglomerate and sandy conglomerate, with brown or yellow massive sandstone intercalated with yellow massive siltstone, and it represents a sandy to gravelly braided fluvial system. The Shizigou Formation consists of coarse-grained conglomerate and medium- to very coarse-grained sandstone, representing a sandy to gravelly braided fluvial or alluvial-fan system. The horizontal Qigequan Formation unconformably overlies the Shizigou Formation and consists of conglomerate, sandy conglomerate, sandstone, and siltstone deposits. Typically, the stratigraphic ages of the Qaidam Basin sediments are bracketed between the Paleocene–Eocene and Quaternary (Chang et al., 2015; Fang et al., 2007; Ji et al., 2017; Lu and Xiong, 2009; Sun et al., 2005). Recently, however, this long-held view was challenged by a new magnetostratigraphic study of the Dahonggou section, bolstered by newly discovered mammalian fossils (Wang et al., 2017). The latter suggested a far younger basal age of ca. 25 Ma for the Lulehe Formation.

Paleocurrent directions were measured from fluvial and alluvial strata of the Lulehe Formation in the northern Qaidam Basin, using cross-stratification (eight measurements at one station) and pebble-cobble imbrication (517 measurements at eight stations), in the localities of Lulehe, Yuqia, Dahonggou, and Yinmaxia (Fig. 3). The Dahonggou locality is ∼5 km east of the previous section of Bush et al. (2016) and Wang et al. (2017). The dip direction and dip angle of planar paleocurrent indicators (cross-strata, pebble-cobble imbrication) were measured in the field with a Brunton compass. We measured 62 and 42 pebble-cobble imbrications in two horizons at the Lulehe locality. Two layers with 93 and 59 pebble-cobble imbrications were recorded at the Yuqia locality, and two horizons with 50 and 50 pebble-cobble imbrications, plus one horizon with eight cross-strata data, were measured at the Dahonggou locality. Finally, we measured 56 and 105 pebble-cobble imbrications in two layers at the Yinmaxia locality. Detailed global positioning system (GPS) locations of all measuring points and attitudes (dip directions and dip angles) of all the pebble-cobble imbrications and cross-beddings are listed in the GSA Data Repository Table DR1.¹ Structural restoration of the paleocurrent data was accomplished using a stereonet computer program.

Five unweathered medium-grained Jurassic–Cretaceous sandstone samples were collected from five localities in the northern Qaidam Basin (Figs. 1C–1E). We attempted to minimize the effects of weathering by excavating and collecting the freshest samples possible. For each sample, an ∼5 kg aliquot of medium-grained sandstone was collected and processed according to standard mineral separation techniques. Subsequently, ∼250 zircon grains were randomly mounted in epoxy resin and then polished to obtain a smooth flat internal surface. Reflected and transmitted light microscopy, as well as cathodoluminescence (CL) imagery, was used to identify ideal grains for U-Pb analysis. Approximately 100 zircon grains were randomly selected for analysis using laser ablation–multicollector–inductively coupled plasma–mass spectrometry (LA-MC-ICP-MS) at the Key Laboratory of Continental Tectonics and Dynamics, Institute of Geology, Chinese Academy of Geological Sciences, Beijing, China. Analyses were conducted using a laser spot size of 32 mm and a repetition rate of 5 Hz under 70% energy conditions, and with a 91500 zircon standard for calibration. Detailed methods have been described in Yuan et al. (2004).

The results of paleocurrent analyses are reported in Table DR1, and a rose diagram is illustrated in Figure 4. Two horizons at the Lulehe locality display a consistent NE-directed flow for the Lulehe Formation deposits. Given the occurrence of coarse-grained boulders at the Lulehe section, the results most likely reflect a high-energy depositional system and proximal sediment source. Therefore, we attribute the two NE-directed paleoflow directions to thrust-related exhumation of the Saishenteng Shan, immediately west of the Lulehe section. Two horizons at the Yuqia locality show a consistent SE-directed flow for the Lulehe Formation. The paleocurrents indicated by the Lulehe Formation at the Dahonggou locality flow SW, and then SE, whereas those indicated by the Lulehe Formation at the Yinmaxia locality flow consistently SW.

Representative CL images of detrital zircon grains from the five sandstone samples are displayed in Figure 5, and the analytical data are listed in Table DR2. It is generally acknowledged that younger ²⁰⁶Pb/²³⁸U ages (younger than 1000 Ma) are often more precise, whereas older ²⁰⁶Pb/²⁰⁷Pb ages (older than 1000 Ma) are often more precise (Gehrels et al., 2011). Provenance determinations were based mainly on age clusters that included at least three analyses (Gehrels et al., 2011). Major age groups and their corresponding peak ages were evaluated by visual inspection of the detrital zircon U-Pb age probability plots for all five samples (Fig. 6A). In this study, the major peak refers to age populations having more than 30% of the total number of data, whereas minor peaks refer to those with less than 20%. A total of 500 detrital zircon grains produced data of sufficient precision for geochronological interpretation (Fig. 6A).

The zircon grains of Early Jurassic sample SD-1 exhibited both euhedral and abraded shapes, with sizes ranging from 50 to 200 µm (Fig. 5). The crystals displayed half distinct oscillatory and half faint zoning on CL images, suggesting both magmatic and metamorphic origins. The Th/U ratios varied from 0.01 to 3.23. Detrital zircon grains of sample SD-1 are 260–2830 Ma in age, with one major peak at 850 Ma and several minor peaks at 266, 463, 1820, and 2451 Ma. The zircon grains of Early Jurassic sample SD-2 showed both euhedral and abraded shapes, with sizes ranging from 50 to 300 µm (Fig. 5). Most of the crystals (∼70%) displayed faint zoning on CL images, but the rest of the crystals showed distinct oscillatory zoning. The Th/U ratios ranged from 0.01 to 3.03. The age-distribution diagram of sample SD-2 shows one major peak at 446 Ma and two subordinate peaks at 240–292 and 952 Ma. The zircon grains of Cretaceous sample SD-3 showed both euhedral and abraded shapes, with sizes ranging from 50 to 300 µm (Fig. 5). Most of the crystals (∼60%) displayed faint zoning on CL images, but the rest showed distinct oscillatory zoning. The Th/U ratios ranged from 0.01 to 3.69. The sample has a unimodal distribution, with one major peak at 460 Ma and rare Paleoproterozoic and Neoproterozoic grains. The zircon grains of sample SD-4 showed both euhedral and abraded shapes, with sizes ranging from 60 to 280 µm (Fig. 5). Most of the crystals (∼70%) displayed faint zoning on CL images, and the rest showed distinct oscillatory zoning. The Th/U ratios ranged from 0.04 to 3.81. The sample yielded a unimodal Paleoproterozoic peak of 2452 Ma, which is consistent with other age results for the same section (Qian et al., 2018; Yu et al., 2017). The zircon grains of sample SD-5 showed both euhedral and abraded shapes, with sizes ranging from 60 to 200 µm (Fig. 5). Most of the crystals (∼60%) displayed faint zoning on CL images, but the rest showed distinct oscillatory zoning. The Th/U ratios ranged from 0.17 to 3.76. The sample exhibited diverse zircon age spectra, with one major peak at 449 Ma and two minor peaks at 983 and 2503 Ma.

Calculation errors and differences in the boundaries assigned to each formation by different researchers may result in differences in the thickness of each formation at the same section. Several studies previously measured the thickness of the Lulehe Formation at the Dahonggou locality (Fig. 1B), and most of them obtained a relatively consistent thickness of 386–490 m (Zhuang et al., 2011; Ji et al., 2017; Wang et al., 2017). Our investigations in the field yield a thickness of ∼546 m for the Lulehe Formation at the eastern Dahonggou section (Fig. 1B). This slight inconsistency is due to different boundary assignments for the Lulehe Formation. However, Bush et al. (2016) reported a thickness of 1100 m for the Lulehe Formation, which is more than twice that of other estimates. After careful examination, we discovered that Bush et al. (2016) misidentified the underlying Cretaceous Quanyagou Formation as the lower part of the Lulehe Formation (Fig. 1B). There is an unconformable contact between the Quanyagou and Lulehe Formation (QBGMR, 1984), but the poor exposure of the Quanyagou and Lulehe Formation at the western section, likely the result of intensive weathering and erosion caused by underground springs, may blur the boundary between the Quanyagou and Lulehe Formation. The near-indistinguishable boundary of the two formations may have led to this error.

Bush et al. (2016) collected samples for detrital zircon U-Pb geochronology and paleocurrent analysis at the Dahonggou section to determine provenance (Fig. 1B). Through strata correction, we are certain that four samples for zircon U-Pb dating and five horizons for paleocurrent analyses from the lower part of the Lulehe Formation were in fact collected from the Cretaceous Quanyagou Formation (QBGMR, 1984). Provenance analyses of the Cretaceous Quanyagou Formation indicated that large amounts of clastic materials from the western Kunlun Shan (Qimen Tagh) were transported to the Dahonggou section by an E-directed axial fluvial system (Bush et al., 2016). However, this view is inconsistent with previous provenance analyses of Cretaceous strata in the northern Qaidam Basin margin, which suggested the derivation of siliciclastic detritus from the Qilian Shan (Ritts and Biffi, 2001). However, discussion of the provenance of Cretaceous strata is beyond the scope of the present study. To compare the provenance results of Cenozoic strata from different studies, we excluded the detrital zircon U-Pb geochronology and paleocurrent data from the Cretaceous Quanyagou Formation (Bush et al., 2016) in the following discussion.

The Lulehe Formation is named for its type section at the Lulehe locality (QBGMR, 1980). It is widely exposed within the northern Qaidam Basin, where four localities were investigated in this study (Figs. 1 and 7). Typically, the Lulehe Formation unconformably overlies Mesozoic strata and conformably underlies the lower Ganchaigou Formation and is composed of terrigenous clastic red beds (Figs. 7 and 8) with a thickness of 400–500 m (Zhuang et al., 2011; Ji et al., 2017; Wang et al., 2017). The division between the Lulehe Formation and the lower Ganchaigou Formation is easily determined by the presence and absence of sandy and conglomeratic red beds, respectively. The Lulehe Formation deposits contain large amounts of cobble-boulder conglomerates (Fig. 7) and are usually interpreted as a synorogenic coarse-grained conglomerate deposited by high-gradient depositional systems (Zhuang et al., 2011; Ji et al., 2017; Yin et al., 2008a). The depositional environments of the Lulehe Formation can be reconstructed based on sedimentological observations of the 546-m-thick measured stratigraphic interval of the eastern Dahonggou section (Fig. 1B).

The Lulehe Formation at the Dahonggou section is divisible into two parts based on lithofacies assemblages and lithological characteristics. The lower part of the section (0–284 m) is characterized by a diverse assemblage of poorly sorted, matrix-supported pebble-boulder conglomerates (mainly Gcm, Gmm, and Gcmi) mixed with brick-red mudstones, siltstones, and sandstones with mottling (Fig. 8G). Beds are usually more than 2 m thick, and occasionally exceed 40 m. The pebble to boulder conglomerates are composed of clasts derived from local basement lithologies (granitoid intrusive rocks, metamorphic rocks, and sedimentary rocks; Zhuang et al., 2011). The sandstones of this unit are texturally immature and matrix-supported, with poor sorting and grain sizes up to granular and pebbly sand (Fig. 8D). However, the sandstones, predominantly composed of monocrystalline quartz, are compositionally hypermature, reflecting derivation from a stable, deeply weathered source area or orogenic recycling of quartzose sedimentary rocks (DeCelles et al., 2014). The texturally immature, compositionally mature aspect of the sandstones in the lower part of the Lulehe Formation indicates that the sediments were derived from a highly weathered terrane and were transported rapidly from source to sink by concentrated density flows (DeCelles et al., 2014). We interpret this sequence to represent debris-flow deposition on an alluvial fan or on fan-delta slopes (Table 1; Miall, 1978; DeCelles et al., 1991).

A distinguishing feature of the upper part of the section (284–546 m) is the presence of moderately sorted, clast-supported, unstratified, imbricated pebble to cobble conglomerates (mainly Gmm and Gcmi) alternating with brick-red siltstones, sandstones, and mudstones (Figs. 8A, 8E, and 8F). These deposits have an erosive and sharp boundary (Figs. 8E and 8F). The conglomerate beds are usually 0.5–5 m thick, and the interbedded sandstones, siltstones, or mudstones are 0.5–10 m thick. The unit is characterized by normal-graded bedding, climbing ripple laminations, parallel bedding, soft-sediment deformation, and trough cross-stratification (Fig. 8H). The sandstones are subangular to subrounded and moderately sorted (Figs. 8B and 8C). The sandstone compositions fall into the feldspatholithic group (McBride, 1963), based on the presence of large amounts of feldspar and sedimentary lithic grains (Figs. 8B and 8C). In contrast to the lower part of the Lulehe Formation, the sandstones of the upper part have submature textures and immature compositions. The abundance of feldspar and sedimentary lithic grains indicates the presence of sedimentary rocks in the source terrane. Based on these observations, we interpret the sequence to represent braided-fluvial deposits (Table 1; Miall, 1978; DeCelles et al., 1991).

Our field observations, together with sedimentological, paleocurrent, and detrital zircon U-Pb age data, substantially improve our understanding of the tectono-sedimentary evolution of the Cenozoic strata in the northern Qaidam Basin. Comparison of detrital zircon U-Pb ages between the Dahonggou sedimentary rocks and potential sources (the northern Qaidam Basin margin and the southern Qilian Shan, East Kunlun Shan, and Qimen Tagh), in conjunction with sedimentology, paleocurrent, seismic reflection data, helps to determine the prime candidate for the source region of the Dahonggou section: the northern Qaidam Basin margin and the southern Qilian Shan or the East Kunlun Shan and Qimen Tagh.

Sedimentological analyses indicate that the strata of the Lulehe Formation at the Dahonggou locality are dominated by conglomeratic and sandy brick-red beds and were deposited initially in an alluvial fan and subsequently in a braided-fluvial system. This conclusion is broadly consistent with previous studies at the localities of Dahonggou (Ji et al., 2017; Song et al., 2013; Bush et al., 2016; Zhuang et al., 2011), Mahai (Zhuang et al., 2011), or Lulehe (Zhuang et al., 2011), which indicated that the deposits of the Lulehe Formation represent a high-gradient depositional system and synorogenic sedimentation, suggestive of a proximal uplifted source terrane (e.g., the northern Qaidam Basin margin and the southern Qilian Shan). Photomicrographs of the Lulehe Formation sandstones reveal the presence of large amounts of unstable feldspar and sedimentary lithic grains (Figs. 8B and 8C), which have also been identified in previous sedimentary petrological studies at Lulehe, or at other localities in the Qaidam Basin (F. Cheng et al., 2016a; Fu et al., 2013; Jian et al., 2013). Presumably, the source area was uplifted and eroded so rapidly that there was insufficient time for physical weathering of the feldspar and sedimentary lithic grains. These results demonstrate the significant derivation of sedimentary rocks from a proximal source terrane (e.g., Mesozoic sedimentary rocks in the northern Qaidam Basin margin), which has important implications for identifying potential source terranes based on detrital zircon U-Pb geochronology, as discussed below.

According to previous studies, continental red beds typically develop as alluvial fan and fluvial deposits and are the earliest deposits associated with the onset of compressive deformation, and they are assumed to represent the initiation of syntectonic sedimentation (Turner, 1980). They have been regarded as clastic wedges of sediment formed along the flanks of actively rising mountain regions, implying strong tectonic control in the form of marginal faulting and uplift of the source regions (Turner, 1980). Interestingly, this wedge-shaped structure was identified by seismic reflection profiling in the Qaidam Basin (Cheng et al., 2017; Wei et al., 2016; Yin et al., 2008b). The Lulehe Formation deposits in sections A–B and C–D of Figure 9 are thickest in the northern Qaidam Basin margin; they then thin to the southwest and then disappear before reaching the basin center (Yin et al., 2008b; Wei et al., 2016). A transverse section (E–F) along the central axis of the basin also supports the absence of the Lulehe Formation deposits at the center of Qaidam Basin, south of the Dahonggou section (Fig. 9D; Cheng et al., 2017). These findings are consistent with an isopach map of the Lulehe Formation deposits in the Qaidam Basin (Fig. 9A; Yin et al., 2008b). This distribution pattern of the Lulehe Formation deposits indicates that the Qaidam Basin was initiated as a common foreland basin during its earliest stage of sedimentation (Yin et al., 2002).

Sedimentology, facies analysis, and seismic reflection profiling provide evidence that the Lulehe Formation strata in the northern Qaidam Basin were deposited in a proximal foreland basin system. Therefore, we conclude that the Lulehe Formation deposits in the locality of Dahonggou originated from the proximal northern Qaidam Basin margin and the southern Qilian Shan.

The paleocurrent orientations of the deposits of the Lulehe Formation at Dahonggou are disputed (Fig. 1A). Three previous paleocurrent measurements of the continuous Cenozoic sequence at Dahonggou have been conducted (Fig. 10): Bush et al. (2016) and Ji et al. (2017) inferred dominantly SW- and SE-directed flows, transverse to and away from the southern Qilian Shan, throughout the Cenozoic. Wang et al. (2017) also inferred four SW-directed paleocurrent orientations within the lower Youshashan, upper Youshashan, and Shizigou Formation; however, they inferred two NE-directed paleocurrent directions within the Lulehe Formation and the lower part of the lower Ganchaigou Formation. Therefore, the sediment-dispersal pathways may have reversed during the interval between the lower Youshashan and lower Ganchaigou Formation.

To clarify the dispute over the provenance of the Lulehe Formation, we measured three paleocurrent orientations (PC-5, PC-6, PC-7) at three different horizons of the eastern Dahonggou section (Figs. 1B and 4). The results indicate SW- and SE-directed flows and a northerly source area. In addition, we obtained six paleocurrent estimates for the Lulehe Formation from the localities at Lulehe, Yuqia, and Yinmaxia in the northern Qaidam Basin (Figs. 1C–1E and 4). Except for two NE-directed flows for Lulehe, the other localities showed SW- to SE-directed flows (Fig. 4). We conclude that the two NE-directed flows at the Lulehe locality reflect the tectonic uplift of the adjacent Saishenteng Shan (Fig. 4).

In summary, all our paleocurrent analyses spanning a wide range of the northern Qaidam Basin confirm that the widespread Lulehe Formation deposits in the northern Qaidam Basin were derived from the northern Qaidam Basin margin and the southern Qilian Shan. These paleocurrent results are supported by other paleocurrent studies in the northern Qaidam Basin (Zhuang et al., 2011; Meng and Fang, 2008).

Comparisons of age distributions based on the presence and absence of specific ages or age groups are the most reliable means for provenance identification (Gehrels et al., 2011). Since numerous geological factors may bias zircon age distributions, including recycling of zircons from older sedimentary units exposed in source terranes and/or incorporation of detrital zircons during transport, variations in the proportions of ages or age groups should be used with caution (Gehrels et al., 2011).

Two previous studies have examined the detrital age structure of the Dahonggou sedimentary rocks, both based on seven sandstone samples (Bush et al., 2016; Wang et al., 2017). Compared to the sampling strategy of Wang et al. (2017), the samples of Bush et al. (2016) had a more uniform distribution. Both studies revealed two major age populations, Permian–Triassic (200–300 Ma) and late Cambrian to Early Devonian (500–400 Ma), and two minor age populations, Neoproterozoic (1000–700 Ma) and Paleoproterozoic (2.5–1.6 Ga). The depth of 3200–4300 m is an exception, and here the Permian–Triassic peak is minor or absent (Fig. 11). This age structure is inconsistent with the concept of a simplified two-stage evolution with a major 250 Ma peak being replaced by a major 440 Ma peak (Wang et al., 2017).

We combined large amounts of data to characterize several potential source terranes for the Cenozoic Dahonggou strata, including the East Kunlun Shan and Qimen Tagh (Fig. 6E; He et al., 2016; Li et al., 2013; Xia et al., 2015), Qilian Shan and Nan Shan (Fig. 6D; Chen et al., 2012; Gehrels et al., 2011, 2003a; Menold et al., 2009), and Mesozoic strata in the northern Qaidam Basin (Fig. 6B; Yu et al., 2017; Bush et al., 2016; this study). Compared to the age populations of the Cenozoic Dahonggou strata (Fig. 6C), those of the Qilian Shan and Nan Shan have fewer Permian–Triassic (300–200 Ma) ages, and those of the East Kunlun Shan and Qimen Tagh have fewer Paleoproterozoic (2.5–1.6 Ga) ages. Although their proportions vary considerably, the four age populations of the Cenozoic Dahonggou strata are all indicative of three potential source terranes. Therefore, it is difficult to tie zircons of these ages to specific source terranes, if we determine source terrane based on the presence or absence of specific age populations, rather than by variations in the proportions of the age populations.

Wang et al. (2017) suggested a two-stage provenance evolution model, with the earlier source area of the East Kunlun Shan being replaced by the Qilian Shan. This model is based mainly on the observation that granite bodies in the Qilian Shan are dominantly late Cambrian to early Devonian (500–400 Ma) in age, whereas those in the East Kunlun Shan are Permian–Triassic (300–200 Ma) in age. However, this assumption overlooks the fact that large amounts of zircons with crystallization ages of 300–200 Ma are present in the northern Qaidam Basin. Besides the occurrence of isolated plutons immediately north of the Dahonggou section in the northern Qaidam Basin (Chen et al., 2012; Cheng et al., 2017; Menold et al., 2009), the Jurassic and Cretaceous sedimentary rocks have a major zircon age peak of ca. 250 Ma (Bush et al., 2016; Yu et al., 2017; Qian et al., 2018).

Geological mapping and analyses of seismic reflection profiles indicate that the northern Qaidam Basin margin and the southern Qilian Shan–Nan Shan thrust belt developed a SW-directed thrust-wedge duplex (triangle-zone structure), the southward propagation of which has accommodated a significant amount of crustal shortening (20% to >60%; Yin et al., 2008a). The structural development of the orogenic wedge controls the geometry and shape of the northern Qaidam Basin and therefore has a strong influence on the prolonged exhumation of older strata within the fold-and-thrust belt and rapid sedimentation of younger strata within the basin interior. The Jurassic and Cretaceous sedimentary rocks crop out discontinuously within the fold-and-thrust belt of the northern Qaidam Basin margin with a greater altitude than the basin interior (Ritts and Biffi, 2001; Yu et al., 2017). It is believed that these Mesozoic rocks experienced multiphase exhumation and prolonged erosion since, or soon after, the beginning of the India-Eurasia collision (Yin et al., 2008a), and therefore they supplied large amounts of detritus to be redeposited within the basin interior. This interpretation was substantiated by recent detrital apatite fission-track analyses of Mesozoic strata in the northern Qaidam Basin margin, which suggested >6 km of exhumation of Mesozoic sedimentary rocks based on the observation that their apatite fission-track ages are younger than their depositional ages (Jian et al., 2018). This sedimentary recycling model is further supported by sandstone compositional data. Numerous Cenozoic sandstone samples from the northern Qaidam Basin represent recycled orogenic or, more specifically, fold-and-thrust belt materials according to classical Dickinson-Gazzi variation diagrams (Fig. 12; Lu et al., 2014; Bush et al., 2016; Jian et al., 2013). Therefore, the Jurassic and Cretaceous sedimentary rocks in the northern Qaidam Basin margin could be one of the major source regions for the Cenozoic Dahonggou section, and recycling of these Mesozoic sedimentary units would yield Permian–Triassic–aged zircons, such as the 300–200 Ma population observed in the Cenozoic sandstones. Rieser et al. (2006b) reported a uniform 280–220 Ma age cluster for detrital white mica in Cenozoic sedimentary rocks from the Lulehe section in the northern Qaidam Basin (Fig. 1C). It was also concluded that these late Paleozoic–Early Triassic ages were either derived from recycled Triassic–Jurassic cover sequences, or from Permian intrusive bodies in the northern Qaidam Basin and the southern Qilian Shan–Nan Shan (Rieser et al., 2006b).

The widespread occurrence of 300–200 Ma detrital zircon ages in Mesozoic strata can be attributed to the fact that the bedrock of the Qaidam Basin contains granite bodies of 290–280 Ma to 215 Ma in age (Wu et al., 2016; Cheng et al., 2017; see Fig. 2C), which are overlain by Mesozoic–Cenozoic sedimentary rocks (Cheng et al., 2017). The episodic rejuvenation of fold-and-thrust belts (Ritts and Biffi, 2001), or normal faults (Yin et al., 2008a; Wu et al., 2011), in the northern Qaidam Basin since the Mesozoic may have exposed subsurface granites, which represent the principal and ultimate source for the 300–200 Ma detrital zircons in the Mesozoic and Cenozoic strata.

We obtained 500 detrital zircon ages from five Mesozoic sedimentary rocks, which were combined to supplement detrital zircon age data for Mesozoic strata in the northern Qaidam Basin. The integrated detrital zircon ages of Mesozoic strata show three major age populations: Permian–Triassic (300–200 Ma), late Cambrian to Early Devonian (500–400 Ma), and Paleoproterozoic (2.5–1.6 Ga), and one minor age population: Neoproterozoic (1000–700 Ma). Based on previous studies, the detrital zircon ages of the Cenozoic Dahonggou strata display two major age populations: Permian–Triassic (300–200 Ma) and late Cambrian to Early Devonian (500–400 Ma), and two minor age populations: Neoproterozoic (1000–700 Ma) and Paleoproterozoic (2.5–1.6 Ga). However, the four age populations are all reported in three potential source terranes, in varying proportions, including the East Kunlun Shan and Qimen Tagh, the Qilian Shan and Nan Shan, and Mesozoic strata in the northern Qaidam Basin. Therefore, it is difficult to tie zircons of these ages to specific source terranes using the presence or absence of specific age populations, rather than examining variations in the proportions of the age populations. Therefore, the use of a single detrital zircon dating method cannot provide unique source information.

Sedimentological analyses indicate that the Lulehe Formation strata at the Dahonggou locality are dominated by conglomeratic and sandy brick-red beds and were deposited first in an alluvial fan and then in a braided-fluvial system (Fig. 8). The Lulehe Formation deposits represent a high-gradient depositional system and synorogenic sedimentation, suggestive of a proximal uplifted source terrane (e.g., the northern Qaidam Basin margin and the southern Qilian Shan). Seismic reflection profiling also testifies that the Lulehe Formation strata in the northern Qaidam Basin were deposited in a proximal foreland basin system (Fig. 13). Therefore, sedimentological and stratigraphic evidence together indicate that the Lulehe Formation deposits at the Dahonggou locality originated from the proximal northern Qaidam Basin margin and the southern Qilian Shan (Fig. 13). This is further supported by our detailed paleocurrent studies spanning a large part of the northern Qaidam Basin (525 measurements at nine stations), which consistently indicate that the widespread Lulehe Formation deposits in the northern Qaidam Basin were derived from the northern Qaidam Basin margin and the southern Qilian Shan (Fig. 13).

Detrital fission-track dating can also be used to characterize variations in source terrane and has been regarded as a very important supplementary means of investigating changes in provenance for the Qaidam Basin (Du et al., 2018; Jian et al., 2018; Wang et al., 2017; Y. Wang et al., 2015). Detrital apatite fission-track data of the Cenozoic strata, especially the Lulehe Formation in the northern Qaidam Basin, show a prominent age population peak of 80–40 Ma (Du et al., 2018; Wang et al., 2017; Jian et al., 2018). This means that some of the Lulehe Formation sediments in the northern Qaidam Basin were most likely derived from Mesozoic strata in the northern Qaidam Basin margin, the detrital apatite fission-track ages of which range from 81 to 46 Ma (Fig. 13; Jian et al., 2018). Reworking of Mesozoic strata in response to uplift and cannibalization in the fold-and-thrust belt is supported by the presence of significant sedimentary lithic (Ls) grains and detrital modal analyses of Cenozoic sandstones in the northern Qaidam Basin, as was discussed above. It is thought that these Mesozoic strata experienced deep burial (>6 km) and strong annealing, resulting in younger detrital apatite fission-track ages than their depositional ages (Jian et al., 2018). Wang et al. (2017) suggested a drastic change of source area, with an early southerly source area (East Kunlun Shan) being replaced by a subsequent northerly source area (the southern Qilian Shan), based on the observations that an early sharp decrease in peak ages up section was succeeded by a slowly decreasing trend of peak ages. However, this conclusion is unconvincing, since it is based solely on sharp variations in detrital apatite fission-track ages. Jian et al. (2018) also observed similar variations in detrital thermochronological data, but they attributed the phenomenon to a single massive source area (the northern Qaidam Basin margin and the southern Qilian Shan) with distinct cooling ages. Low-temperature thermochronological studies of crystalline basement rocks suggest that the northern Qaidam Basin margin (Lvliang Shan and Xitie Shan) and the southern Qilian Shan had a regionally heterogeneous cooling history, which indicates three phases of rapid exhumation since the Mesozoic (Early Cretaceous, Paleocene–Eocene and middle–late Miocene; X. Cheng et al., 2016; Jolivet et al., 2001; Wang et al., 2004; Zhuang et al., 2018).

This provenance model has important implications for the timing of the tectonic uplift of the northern Qaidam Basin margin and the southern Qilian Shan. The age of the basal Cenozoic sedimentary rocks in the Qaidam Basin (the basal age of the Lulehe Formation) is bracketed to 65–50 Ma (Ji et al., 2017; Yin et al., 2008a), or ca. 25 Ma (Wang et al., 2017). Since these massive terrigenous clastic sedimentary rocks in the northern Qaidam Basin are the product of surface uplift, tectonic exhumation, and erosion of the northern Qaidam Basin margin and the southern Qilian Shan, the timing of the initiation of surface uplift of the northern Qaidam Basin margin and the southern Qilian Shan was likely no later than 65–50 Ma, or ca. 25 Ma. Irrespective of which age is correct, the substantial uplift of the southern Qilian Shan should be far older than 12 Ma, as suggested by Wang et al. (2017).

Our integrated study of the sedimentology, paleocurrent directions, detrital zircon U-Pb ages, detrital apatite fission-track ages, and seismic reflection profiles enables us to assemble a reliable reconstruction of the provenance of the Lulehe Formation in the northern Qaidam Basin.

• (1) The detrital zircon age populations of the Cenozoic Dahonggou strata are all reported in three potential source terranes, in varying proportions, including the East Kunlun Shan and Qimen Tagh, the Qilian Shan and Nan Shan, and Mesozoic strata in the northern Qaidam Basin. Thus, the use of a single detrital zircon dating method cannot provide unique source information.

• (2) The sedimentological and paleocurrent analyses, in combination with existing paleocurrent data, seismic reflection data, and detrital apatite fission-track dating, point consistently to a unified proximal northerly source area (the northern Qaidam Basin margin and the southern Qilian Shan).

• (3) The orogenic recycling effect of Mesozoic strata in the fold-and-thrust belt of the northern Qaidam Basin, which was ignored in recent studies, played a significant role in producing the Permian–Triassic–aged zircons observed in the Cenozoic sandstones.

We thank Longgang Fan, Guoli Wu, and Xuxuan Ma for help with data processing and interpretations; Aijun Sun for help in the field; and Min Lei and Xiangzhen Xu for providing laboratory support. Laurent Godin (Science Editor) and two anonymous reviewers are thanked for constructive and valuable suggestions. This research was supported by the National Natural Science Foundation of China (grant 41472187), the Strategic Priority Research Program of the Chinese Academy of Sciences (grant XDA19050104), and the China Geological Survey (grant DD20160022–03).