Earth-Science Reviews 190 (2019) 460–485

Contents lists available at ScienceDirect

Earth-Science Reviews

journal homepage: www.elsevier.com/locate/earscirev

Cenozoic magnetostratigraphy of the Xining Basin, NE Tibetan Plateau, and T its constraints on paleontological, sedimentological and tectonomorphological evolution ⁎ Xiaomin Fanga,b, , Yahui Fanga,b, Jinbo Zana, Weilin Zhanga, Chunhui Songc, Erwin Appeld, Qingquan Mengc, Yunfa Miaoe, Shuang Daic, Yin Lua,d, Tao Zhanga a CAS Center for Excellence in Tibetan Plateau Earth Sciences and Key Laboratory of Continental Collision and Plateau Uplift, Institute of Tibetan Plateau Research, Chinese Academy of Sciences, Beijing 100101, China b College of Resources and Environment, University of Chinese Academy of Sciences, 19A Yuquan Road, Beijing 100049, China c School of Earth Sciences & Key Laboratory of Western China's Environmental Systems (MOE), Lanzhou University, Lanzhou 730000, China d Department of Geosciences, University of Tübingen, Hölderlinstr. 12, 72074 Tübingen, Germany e Key Laboratory of Desert and Desertification, Cold and Arid Regions Environmental and Engineering Institute, Chinese Academy of Sciences, Lanzhou 730000,China

ARTICLE INFO ABSTRACT

Keywords: The Xining Basin is well known for its thick and continuous sequence of fine-grained Cenozoic sediments. In Cenozoic magnetostratigraphy addition, it contains important Xiejia fauna which define the Xiejian Stage of the early Miocene in the standard Xiejia fauna Chinese land mammal zonation, and it also provides detailed histories of eolian dust deposition which document Eolian dust deposition the aridification of Asia, monsoon evolution and the initiation of the Yellow River in relation to the upliftof Xining Basin Tibetan Plateau. However, the results of magnetostratigraphic dating of the fossil-bearing sequence has yielded NE Tibet uplift conflicting ages for the fauna and stratigraphy, hindering the use of the sequence for addressing majorpa- leoenvironmental and paleoclimatic issues. Here, we review the paleontological, lithological and magnetos- tratigraphic record of the Xining Basin and provide a new, long and continuous high-resolution magnetos- tratigraphy from the basin center. The results show that the observed magnetic polarity zones from the various sections at different sites in the basin exhibit similar magnetic polarity patterns which can be readily correlated, in terms of both magnetic polarity zonation and lithofacies; in addition, they show that the sedimentary se- quence of the basin was sub-continuously deposited from ~54 Ma to 4.8 Ma. The magnetostratigraphic corre- lations confirm previous constraint of the Xiejia fauna in the late Oligocene at ~25 Ma, challenging thecurrent Chinese mammal land zonation for the early Miocene, and we suggest that it provides a superior record of eolian dust deposition and river incision in the late Pliocene than previous studies. There is a close match of the climatic proxy records of the Xiejia section, dated by magnetostratigraphy, with the global climatic record which cor- roborates the paleomagnetic correlations. The evolution of the sedimentary environment provided by the li- thofacies demonstrates a complete cycle of basin formation and termination, which records the eastern Qilian Shan experienced three main phases of uplifts: slow episodic uplifts at ~54 Ma and 22.5 Ma and a late rapid episodic uplifts at 8–7 Ma and since 4.8–3.6 Ma. This tectonic uplift in the NE Tibetan Plateau was mostly nearly synchronous responses to the initial and continuing collision of India and Asia since ~50 ± 5 Ma.

1. Introduction et al., 1996a,b; Li and Fang, 1999)(Fig. 1). It is especially important in relation to the standard Chinese land mammal zones (CLMZ) of the The Xining Basin (XB) lies in a critical transition zone, in terms of early Miocene Xiejian Stage (equivalent to the Aquitanian Stage of the both tectonics and landforms, between the NE Tibetan Plateau and the European marine fossil zones or the Agenian Stage (MN 1–3) of the Loess Plateau, and in climatic terms between the Asian monsoon- European land mammal zones (ELMZ)) (Li and Qiu, 1980; Li et al., dominated region and the dry, westerly-influenced Asian interior (Li 1981, 1984; Qiu et al., 1981; Qiu and Qiu, 1990; Deng et al., 2006). It

⁎ Corresponding author at: CAS Center for Excellence in Tibetan Plateau Earth Sciences and Key Laboratory of Continental Collision and Plateau Uplift, Institute of Tibetan Plateau Research, Chinese Academy of Sciences, Beijing 100101, China E-mail address: [email protected] (X. Fang). https://doi.org/10.1016/j.earscirev.2019.01.021 Received 1 September 2017; Received in revised form 17 January 2019; Accepted 23 January 2019 Available online 25 January 2019 0012-8252/ © 2019 Elsevier B.V. All rights reserved. X. Fang et al. Earth-Science Reviews 190 (2019) 460–485

Fig. 1. (a)DEM image showing the location of the Xining Basin with respect to the geomorphologic and tectonic framework of the NE Tibetan Plateau. (b) Distribution of the Cenozoic stratigraphy of the Xining Basin and locations of the sections referenced in the text (modified from Dai et al., 2006). GJS: Guanjiashan; TS: Tashan and drilling site; 1: Xiejia section; 2: Tashan section; 3: East Xining-Shuiwan-Guanjiashan section; 4: Mojiazhuang section; 5: Bingling Shan section; 6: Yagou section; HW: Houwan section; CJB: Caijiabao section; DDL: Dadunling loess section.

461 X. Fang et al. Earth-Science Reviews 190 (2019) 460–485

Fig. 2. Cross sections a-b and c-d showing the relationships between the Cenozoic stratigraphy and tectonics. c-c′ indicates the East Xining-Shuiwan-Guanjiashan composite sections with its extension upward to the Houwan and Caijiabao sections for the top sequences. Topographic surface outlines are taken from DEM topographic data and 1:50,000 topographic maps. See Fig. 1 for the locations of cross-sections. Numbers in circles indicate the mammalian fauna listed in Table 1. develops the continuous sequence of predominantly fine-grained sedi- great potential for providing high resolution paleoclimatic records ments spanning the early Eocene to the Pliocene (QBGM, 1991; Dai documenting Asian aridification and monsoon evolution, the uplift of et al., 2006; Fang et al., 2015; Yang et al., 2017)(Fig. 2). Thus, it has NE Tibet and the origin and incision of the Yellow River (Li et al.,

462 X. Fang et al. Earth-Science Reviews 190 (2019) 460–485

1996a,b; Lu et al., 2004; Dupont-Nivet et al., 2007; Lin et al., 2001; tectonosedimentary and geomorphologic (river incision) framework of Licht et al., 2014; Fang et al., 2015; Nie et al., 2015). The sequences are the XB. We hope that this work will provide the basis for future studies especially important for studying the time interval from the Oligocene of the deformation and uplift of the NE Tibetan Plateau and will help to the early Miocene, which marks a critical transition in both climatic unravel the paleontological and environmental evolution of the region, change and tectonomorphological evolution on both regional and as well as the record of Asian aridification and monsoonal climate global scales, and marked the shift from a greenhouse to an icehouse change provided by the unusually continuous, fine-grained sedimentary Earth (Zachos et al., 2008). The interval is also important in terms of sequence. changes in global planetary wind systems and regional monsoons in Asia (Liu et al., 1998; Sun and Wang, 2005; Guo et al., 2008), as well as 2. Geographic and geologic setting for crustal thickening, extrusion and the growth and uplift of the Ti- betan Plateau, and the associated large-scale reorganization of fluvial The XB is in the eastern extension of the Qilian Shan (Mts.) at the systems (Tapponnier et al., 2001; Molnar, 2005; Zheng et al., 2013). northeastern corner of the Tibetan Plateau (Fig. 1). Its dimensions are However, intense tectonic deformation and broad uplift have also given about 100 × 80 km and the basin surface elevations are mainly be- risen to widespread erosion and development of unconformities in the tween about 2000 and 3000 m. It is confined by the Daban Shan (Mts.), Oligocene-to-early Miocene stratigraphy in Asia. These themes high- Laji Shan (Mts.) and Riyue Shan (Mts.) to the north, south and west, light the importance of the continuous Oligocene-to-Pliocene fine- respectively, while the eastern side is open to the Minhe Basin (Fig. 1b). grained sedimentary record of the Xining Basin as an archive of in- The Huang Shui (River), a tributary of the Huang He (Yellow River), formation on paleontological, climatic and tectonic evolution, which is originates in the Riyue Shan and Daban Shan and flows from west to reflected by increasing regional and global research attention (e.g., Li east, incising the basin center. Up to 16 terraces have developed as a et al., 1996a,b; Horton et al., 2004; Lu et al., 2004, 2007; Dai et al., result, upon which thick (up to 200 m) eolian loess was deposited (Li 2006; Deng et al., 2006; Dupont-Nivet et al., 2007, 2008a,b; Xiao et al., et al., 1996a,b; Lu et al., 2004, 2007; Zhang et al., 2017)(Figs. 1 and 2). 2010, 2012; Abels et al., 2011; Long et al., 2011; Hoorn et al., 2012; Chi Climatically, the XB is now dominated by an arid/semiarid con- et al., 2013; Licht et al., 2014, 2016; Zhang and Guo, 2014; Fang et al., tinental climate with mean annual temperature (MAT) of 5–6 °C and 2015; Zan et al., 2015; Han et al., 2015; Zhang et al., 2017; Yang et al., mean annual precipitation (MAP) of 350–500 mm. Rainfall is mainly 2017). from June to August, following a typical monsoon pattern. The Xiejia (XJ) fauna was initially discovered in the XB (Li and Qiu, The collision of the Qaidam Block and minor Qilian Blocks with the 1980; Li et al., 1981; Qiu et al., 1981) and was later classified as the North China Block during the Caledonian Orogeny terminated the standard Chinese land mammal stage of the early Miocene (Li et al., Qilian Ocean and initiated the Paleozoic Qilian orogen, forming the 1984; Qiu and Qiu, 1990; Deng et al., 2006). As a result, the XB has basement rocks of the intermontane Mesozoic-Cenozoic basins (Gehrels attracted increasing research attention in the areas of paleontology, et al., 2003). These basins were formed due to post-collisional relaxa- magnetic stratigraphy, sedimentology, tectonics, and climatic change. tion and collapse of the orogen and were infilled with thick continental Wu et al. (2006) carried out the first paleomagnetic dating of the Mesozoic-Cenozoic red beds (Feng and Wu, 1992; Yin and Harrison, stratigraphic interval containing the fossils in the XJ section and de- 2000; Fang et al., 2013). Most of the Cenozoic basins developed from termined an age of 21 Ma for the XJ fauna. Later, high resolution pa- Mesozoic basins (Yin and Harrison, 2000; Fang et al., 2013) due to the leomagnetic dating of the entire XJ section constrained the age of the response to the collision of India with Asia since 50 ± 5 Ma (Rowley, XJ fauna to ~25.5 Ma, challenging the reliability of the classification of 1996; Zhu et al., 2005; Molnar and Stock, 2009). The XB succeeded the the XJ fauna as a standard mammalian zone of the Early Miocene (Dai large Mesozoic Xining-Minhe-Lanzhou Basin (Horton et al., 2004), et al., 2006). Subsequent drilling of the uppermost part of the XJ sec- where the youngest Mesozoic stratigraphy is the early Cretaceous red tion, and a parallel study of the nearby Tashan section, confirmed the beds, which form the base of the Cenozoic XB and are widely exposed reliability of the magnetostratigraphy of Dai et al. (2006) (Xiao et al., along the rim of the basin, while the Jurassic stratigraphy is exposed 2012; Zan et al., 2015). This was further supported by the litho- and only at a small site in the basin center (QBGM, 1991)(Fig. 1b; Suppl. magneto-stratigraphy of the Shuiwan section in the central XB (Dupont- Fig. 1). Nivet et al., 2007, 2008a,b) (see Fig. 1 for site locations). Zhang et al. Two sets of transpressional thrust faults border the basin. The (2017) obtained a detailed magnetostratigraphy for the uppermost part roughly E-W- oriented elongated sinistral thrust faults border the of the Shuiwan section (otherwise known as the Guanjiashan section) northern and southern margins of the basin and the major ranges of the and found that it may correlate with the equivalent part of the XJ Qilian Shan (Fig. 1). A set of NW-oriented short dextral transpressional section. Most recently, a late Miocene Hipparion fauna was discovered thrust faults occur between the E-W faults, separating or dividing the in the Mojiazhuang (MJZ) or Jiangjiagou section, about 10 km east of roughly E-W-elongated basins from each other, or dividing sub-basins the Shuiwan section (Han et al., 2015). Accordingly, this section com- within basins. For the XB, the North Laji Shan Fault (NLSF) in the south, prises the uppermost sequence of the Cenozoic stratigraphy of the XB and the North Central Qilian Shan fault (NCQF) in the north, thrust the and was paleomagnetically dated to 12.7–4.8 Ma (Yang et al., 2017). mountain basement rocks onto, and consequently deforming, the The continuity of the red bed sequence was found to be in conflict with Xining red beds on the southern and northern margins of the basin, the inferred timing of terrace development of the Huang Shui (River), a respectively; roughly E-W trending folds were formed as a consequence tributary of the Yellow River, and its overlying eolian dust sequence (Figs. 1, 2). The dextral Haiyan fault lifts the Riyue Shan and the which was paleomagnetically dated to about 14 Ma and was claimed to western margin of the XB, possibly giving rise to several local gentle document the subsequent aridification of Asia (Lu et al., 2004). The northeast-trending folds (Wang et al., 2011; Yuan et al., 2013)(Fig. 1). terrace age was much older than the age of 1.2 Ma suggested by Li et al. Several small dextral faults occur in the XB, forming small ramps which (Li et al., 1996a,b). are incised by the Huang Shui as gorges which divide the basin into the These contradictory stratigraphic dating results hinder the clar- Xining and Minhe Basins (Fig. 1b). ification of the major scientific questions described above. Here,we The Laji Shan and the Daban Shan, to the south and north of the XB, present the results of new investigations of stratigraphic sequences and respectively, with average elevations 3500–4000 m, belong to the South detailed magnetostratigraphic dating of the central XB, provide a and Central Qilian Shan and consist mainly of pre-Cambrian slates, comprehensive review of the entire Cenozoic stratigraphic sequence of schists and marble, Caledonian grano-dioritic intrusive rocks, Paleozoic the XB, and assess and supplement the previous magnetostratigraphy. limestones and sandstones, and occasional Triassic limestones and Our overall aim is to assess the reliability of the paleontological and Jurassic black shale (QBGM, 1991). paleomagnetic constraints of the XJ fauna, and to clarify the The formation of the entire NE Tibetan Plateau, including the XB

463 X. Fang et al. Earth-Science Reviews 190 (2019) 460–485 and its surrounding Laji Shan and Daban Shan has generally been re- muddy gypsum or gypsiferous mudstone - siltstone and medium-to-thin cognized as a remote response to the collision of India with Asia (Yin lenticular grayish sandstone or gravelly sandstone with occasional and Harrison, 2000; Horton et al., 2004; Fang et al., 2003, 2005a,b, cross- or massive bedding (QOET, 1977; Dai et al., 2006; Fang et al., 2013; Lease et al., 2011; Yuan et al., 2013). The uplift histories of the 2015)(Table 1 and Figs. 3, 4 and Suppl. Fig. 2). Laji Shan, Daban Shan and other parts of the Qilian Shan ranges have The Qijiachuan Fm. (S13–15, thickness 22.4 m), at the base of the been well constrained by basin sediments and thermochronological XJ section, is composed of a distinct unit of red-to-grayish black cal- ages. They show two long phases of uplift, the early Cenozoic slow rise careous mudstones intercalated with thin yellowish-grayish marlite, around 50–30 Ma and the late Cenozoic rapid rise starting at ca. 22 Ma greenish-gray muddy gypsum layers and a few thin massive sandstone and accelerating at ca. 13–8 Ma (e.g., Fang et al., 2003, 2005a,b, 2007, beds (QOET, 1977)(Table 1 and Figs. 3, 4 and Suppl. Fig. 2). We infer 2013, 2016; Yan et al., 2006; Lease et al., 2007, 2011, 2012; Clark that these sediments were deposited in a transitional zone from a distal et al., 2010; Yuan et al., 2013; Zhang et al., 2015, 2016; Yang et al., floodplain with swamps and a brackish lake to a playa mudflat withan 2017; Nie et al., 2017), interpreted as adjustments to the slip history of ephemeral salt-lake (Reading, 2009; Fang et al., 2016)(Fig. 4). the Altyn Tagh fault (Yin and Harrison, 2000; Wu et al., 2012; Li et al., The Honggou Fm. (S16–23, thickness 177.22 m) is characterized 2017) and the northeastward extension of the Tibetan Plateau mainly by red mudstones intercalated with thin-to-thick greenish-gray (Tapponnier et al., 2001; Lease et al., 2011). gypsum beds (with gypum bed thicknesses and frequency increasing upwards) and occasional thin grayish marlite layers. The Mahalagou 3. Litho- and bio-stratigraphic sequences and sedimentary facies Fm. (S24–33, thickness 344.2 m) is very distinctive and consists of al- ternating thick (~2.5 m) greenish-gray gypsum beds and thin 3.1. Previously-studied sections (~0.5–1 m) brown gypsiferous mudstone beds with the gypsum layer frequency decreasing upwards, especially in succession S33 (60.52-m 3.1.1. Xiejia-Chetougou-Tashan sections thick) which has only a single thin gypsum layer (QOET, 1977)(Table 1 The Cenozoic stratigraphy in the XB comprises a thick (over and Figs. 3, 4 and Suppl. Fig. 2). These two formations clearly indicate a 1200–2000 m) continuous sequence of dominantly fine-grained fluvio- playa salt-lake and mudflat system (Reading, 2009; Dai et al., 2006; lacustrine to saline-playa sediments (QBGM, 1991; Dai et al., 2006; Dupont-Nivet et al., 2007; Abels et al., 2011)(Fig. 4). Fang et al., 2015). It disconformably overlies the Cretaceous alluvial The Xiejia Fm. has various definitions, leading to different lower sedimentary rocks of the Hekou and Minhe Groups, or unconformably and upper boundaries in succession (QOET, 1977; Li et al., 1981; overlies older basement rocks (QBGM, 1991; Horton et al., 2004). The QBGM, 1991; Dai et al., 2006; Wu et al., 2006; Xiao et al., 2012) sequence is incised by the paleo- and modern Huang Shui River to form (Table 1 and Figs. 3, 4; Suppl. Fig. 2). The QOET originally defined a series of high and low terraces, which are capped by thick eolian loess successions S34–40, with a thickness of 153.11 m, as the XJ Fm.; it (up to 200 m) (Li et al., 1996a,b; Lu et al., 2004, 2007; Zhang et al., consists of brownish-to-yellowish-brown mudstones intercalated with 2017). The Cenozoic stratigraphy was first investigated and named by occasional thin greenish, yellowish-green muddy gypsum beds and Young and Bien (1937) as the Paleogene Xining Group and Neogene occasional sandstone lenses (QOET, 1977)(Table 1 and Fig. 4h; Suppl. Guide Group. The type succession was established by the Qinghai Oil Fig. 2). We interpret the mudstones and thin, massive sandstones as Exploration Team (QOET) in 1977 in the Xiejia - Chetougou section overbank sediments and sediments deposited in the small river chan- (hereafter the XJ section) (36o31’ N, 101o52’ E), about 1 km north of nels of a distal floodplain, with thin greenish muddy gypsum beds de- Xiejia village in Tianjiazhai Town in the mid-southern part of the XB, posited in ephemeral salt lakes on the floodplain (Reading, 2009; Fang about 15 km south of Xining City (Fig. 1b; Suppl. Fig. 1). The QOET et al., 2016)(Fig. 4). The XJ fauna occurs in the lowermost greenish divided the XJ section (~1000-m-thick) into 47 lithologic successions muddy-to- sandy gypsum (in succession 39) and it is also the lowermost (S) (QOET, 1977)(Table 1 and Figs. 3, 4b, h; Suppl. Fig. 2). According level of fossil mammals found in the XB; thus for convenience we to lithology and ages estimated by pollen and spore assemblages, other designate it the 1st level of fossil fauna. It contains several fossil microfossils, and fossil mammals found in the section, successions mammals that have been classified as the Early Miocene Xiejian Stage S1–12 were assigned to the early Cretaceous and the remainder to the of the widely-accepted CLMZ (equivalent to the Agenian Stage (MN Cenozoic. The Cenozoic successions were further divided into two 1–3) of the ELMZ) (QOET, 1977; Li and Qiu, 1980; Li et al., 1981, 1984; groups, the Paleogene Xining Group (S13–33) and the Neogene Guide Qiu et al., 1981; Qiu and Qiu, 1990; Deng et al., 2006)(Table 2 and Group (S34–47) (Table 1 and Figs. 3, 4b, h; Suppl. Fig. 2). The Paleo- Figs. 3, 4; Suppl. Fig. 2). gene Xining Group was subdivided into the Paleocene Qijiachuan For- Changes in the definition of the upper boundary of the XJ Fm.,as mation (Fm.) (E1q) (S13–15), the Eocene Honggou Fm. (E2h) (S16–23) discussed above, inevitably result in changes in the lower boundary of and the Oligocene Mahalagou Fm. (E3m) (S24–33); and the Guide Group the overlying Chetougou Fm. (QOET, 1977; QBGM, 1991; Dai et al., into the early Miocene Xiejia Fm. (N1x) (S34–40), the middle Miocene 2006; Wu et al., 2006; Xiao et al., 2012)(Table 1 and Figs. 3, 4; Suppl. to Pliocene Chetougou Fm. (N1c) (S41–44), and the Pliocene Xianshuihe Fig. 2). The QOET originally defined successions S41–44, with a Fm. (N1xs) (S45–47) (QOET, 1977)(Table 1 and Figs. 3, 4h; Suppl. thickness of 119.21 m, as the Chetougou Fm. and described it as pre- Fig. 2). This division has generally been accepted as basis for the dominantly light-brownish to light-yellowish-brown massive siltstones Cenozoic stratigraphy in the XB (e.g., Zhai and Cai, 1984; QBGM, 1991) and mudstones intercalated with occasional thin, brownish-gray fine- however, the ages have been subsequently better constrained as pre- grained sandstones, with the base consisting of a thin gravel sandstone sented below. bed with obvious cross-bedding (QOET, 1977). We infer that these se- The lower part of the Xining Group is characterized by the dom- diments were also deposited in a distal floodplain but with slightly inance of gypsum deposits, and the upper part by a the dominance of more energetic river channel flow (Reading, 2009; Fang et al., 2016) red beds. The gypsum dominated lower part consists of rhythmic white, (Fig. 4). Fossil mammals of rodents Eumyarion sp. (?), Rodentia indet., greenish gypsum with muddy gypsum beds of near-uniform thickness and Rhinocerotidae indet. were found in the striking cross-bedded (~1.5–2.5 m), and reddish gypsiferous mudstone or silty mudstone gravel sandstone bed in succession 41, at the base of the formation beds. The red bed dominated upper part consists mainly of thick, (QOET, 1977)(Tables 1, 2 and Figs. 3, 4; Suppl. Fig. 2). However, in the massive purplish-red gypsiferous silty mudstone or mudstone beds in- equivalent stratigraphy at other sites in the XB, two levels of abundant tercalated with occasional greenish gypsum or muddy gypsum beds. In fossil mammals were found and named respectively the Chetougou contrast, the Guide Group is almost devoid of gypsum and is char- fauna (hereafter called the 2nd level of fauna) and the Danshuilu fauna acterized by brown to yellowish-brown or brownish-yellow massive (the 3rd level of fauna) (Suppl. Fig. 1). Correspondingly, they have been mudstones and siltstones, occasionally intercalated with thin greenish identified in composition and age as mammals of the late Shanwangian

464 .Fn tal. et Fang X. Table 1 Division of the Cenozoic stratigraphy in the Xining Basin.

QOET, 1977 Li et al., 1981 QBGM, 1991 Horton et al., 2004 Dai et al., 2006

Section location Xiejia-Chetougou Xiejia and other sites Xiejia and other sites Xiejia and other sites Xiejia Evidence stratigraphy Pollen, ostracods stonewort mammal Mammal Pollen, ostracods stonewort, Pollen Paleomag., mammal mammal Guide group Linxia Fm. Guide Fm. Pliocene Linxia Fm. 7–4.8 Ma (Bahean-Baodean /Vallesian- Turolian) Xianshuihe Fm. S45–47 L. Miocene S45–47 Miocene (L. Tunggurian/L. Astaracian) 18- < 17 Ma 18- < 16 Ma E. Middle Mio. (16–13.8 Ma) (E. Tunggurian / E. Astaracian) Chetougou Fm. S41–44 M. Miocene (MN5–6) S42 (upper)-44 L. Early Mio. (20.4–16 Ma) (Shanwangian / Orleanian) M. Mio. to Pliocene. M. Miocene - Pliocene 17.3- < 15 Ma 23–18 Ma 24.0-17 Ma Xiejia Fm. S34–40 S38 S38–42(lower) S33–40 S38–42 E. Miocene –41 30–23 Ma 25.5–20 Ma Early E. Miocene (MN2–4) E. Miocene (25.5–24 Ma) Xining Group Mahalagou Fm. S24–33 S24–37 Oligocene < S31–S34 Oligocene Oligocene 43-34 Ma Honggou Fm. S16–23 S16–23 Eocene Eocene Eocene 50.0–41.5 Ma 51.8-43 Ma Qijiachuan Fm. S13–15 S13–15 Paleocene S13–15 Paleocene Paleocene 52.5–50.0 Ma 465

Xiao et al., 2012 Wu et al., This study Mammalian fauna found in the stratigraphy (see Table 2 for details) 2006

Tashan Xiejia Xiejia-Tashan, East Xining-Shuiwan-Guanjiashan, Houwan, Lithology and fauna Age Mojiazhuang Pollen Paleomag., Paleomag., mammal Paleomag., mammal mammal P- Linxia Fm. S67–69 Mojiazhuang Fm. Very thick gray pebble-to-cobble conglomerates intercalated with thin, light Late Miocene (11.6–5.3 Ma) brownish-yellow gravelly siltstones. S45–47 S47–48 S47–66 Guanjiashan Fm. Brownish yellow siltstones and mudstones intercalated with many thin-bedded L. Middle Mio. 17–7 Ma white marlite, sandstones and reddish brown paleosols with nodules in the (13.8–11.6 Ma) basin center, but increasing sandstones and developing grayish fine Pliocene (MN7–8) conglomerates to the basin margins. Sandstone beds in the lower, middle and upper contain the Danshuilu, Diaogou and Xiamen faunae respectively.

S41–44 S43–46 S mid. 42 < S43 Upper S39–46 Chetougou Fm. Light brownish-yellow to light-orange calcareous mudstones intercalated with Earth-Science Reviews190(2019)460–485 L. 20–18 Ma some thin massive to cross-bedded sandstones and occasional grayish muddy marlite in the basin center, but increasing sandstone beds and developing reddish brown paleosols to the basin margins. The distinct thick cross-bedded sandstone bed (S41) near the bottom contains the Chetougou fauna. S35–S42 (mid.) S28-Lower Xiejia Fm. Brownish to light brownish-red siltstones Late Oligocene* 39 and mudstones intercalated with three (28.4–23 Ma) 21.6–17.3 Ma (21 Ma) 34–24.0 Ma distinct 1–2-m-thick grayish-greenish (Tabenbulukian / Chattian) S24–32 gypsiferous sandstones bearing the Xiejia S > 27–37 (continued on next page) fauna at the top and a few thin to thick greenish gypsiferous mudstones and gypsum beds in the lower and middle. 41.0–30.0 Ma > 35.5–25.- > 24.0–21.6 Ma S24–27 Mahalagou Fm. Interbeddings of nearly equally-spaced thick white gypsum beds and brownish 5 red or greenish gypsiferous mudstones. X. Fang et al. Earth-Science Reviews 190 (2019) 460–485

and early Tunggurian Stages of the CLMZ (early Middle Miocene, cor- related with the late Orleanian (MN 5) to early Astaracian (MN 6) of the ELMZ) (Li et al., 1981; Liu, 1992)(Table 2 and Figs. 1, 2, 4). The remaining successions S45–47, with a thickness of 93.44 m, are described by QOET (1977) as light brownish-yellow to light-orange or dull-yellow massive calcareous mudstones intercalated with occasional thin sandstones and white, light-grayish muddy marlite (Table 1 and Figs. 3, 4; Suppl. Fig. 2). We interpret them as being deposited in a distal floodplain setting with brackish lakes (Fig. 4b). Fossil mammals were not found in these successions in the XJ section; however, many fossil mammals were found in equivalent stratigraphy in the nearby Diaogou section (about 2 km northwest of the XJ section) and in many other sections in the basin; they are named the Diaogou fauna (here- after, the 4th level of fauna) (Table 2 and Figs. 1, 2, 4; Suppl. Fig. 1). for details) The components of the Diaogou fauna have been thought to resemble those found in the Xianshuihe Fm. in the neighboring Lanzhou Basin Table 2 (Fig. 1). They are characterized by the well-known fossil elephant Gomphotherium sp., and are thus also called the Gomphotherium fauna, and were assigned to the late Tunggurian Stage (Late Miocene) by the CLMZ (Li et al., 1981). This was later changed to the Middle Miocene,

Brownish red gypsiferous siltstones andgreenish mudstones gypsum with beds some or thin occasionally grayish- gray sandstones or marlite. equivalent to the late Astaracian Stage (MN7–8) of the ELMZ (Liu, 1992)(Table 2 and Figs. 3, 4). The stratigraphic name ‘Xianshuihe Fm.’ in the Lanzhou Basin was also used to name the S45–47 successions in the XB (QOET, 1977; Li et al., 1981) and was subsequently widely ac- cepted (QBGM, 1991; Liu, 1992; Dai et al., 2006; Xiao et al., 2012) (Table 1 and Figs. 3, 4; Suppl. Fig. 2). We re-measured the XJ section and determined the following for- mation thicknesses: Qijiachuan Fm., 20–69 m; Honggou Fm., 69–224 m; Mahalagou Fm., 224–504 m; XJ Fm., 504–667 m; Chetougou Fm., Honggou Fm. Brownish purple mudstones, gray marlite and black calcareous mudstones intercalated with some thin sandstone and grayish - greenish gypsum beds in the basin center; Orange red fine conglomerates, sandstones and gravely siltstones in the basin margins. Mammalian fauna found in the stratigraphy (see 667–778 m; Xianshuihe Fm. 778–819 m (Dai et al., 2006)(Figs. 3, 4b, h). The sediment lithology and facies exhibit a slight coarsening-up- ward trend and decrease in gypsum beds. The alternating gypsum layers and gypsiferous mudstones and siltstones below 388 m are in- terpreted as the cyclic deposits of a playa salt-lake and its marginal mudflats. The massive mudstones and siltstones (red beds) areocca- sionally intercalated with thin gypsum layers (with the final layer at about 500 m), between 388 and 560 m, are recognized as the deposits of playa mudflats and an ephemeral salt-lake (Dai et al., 2006; Dupont- Nivet et al., 2007)(Fig. 4b). Above 560 m, the sequence is mainly in- terpreted as flood plain mudstones and siltstones intercalated with occasional channel deposits consisting of fine sandstones with clear cross-bedding (Dai et al., 2006)(Fig. 4b). In addition, above 700 m, we

S16–23 Qijiachuan Fm. found many brownish paleosols with a clear granular argillic layer (Bt horizon) and carbonate nodule layer (Bk horizon) that can roughly be classified as luvic cambisols (FAO, 1988) (see below for more detailed descriptions). A 194-m drill core was obtained from the top of the XJ section at Tashan to obtain a complete succession (Zan et al., 2015)(Fig. 3). The core revealed that the upper Xianshuihe Fm. is light yellowish-brown, mainly massive, with occasional mottled siltstones and mudstones with S10 –S 15 54–51.8 Ma This study many layers of greenish fine sandstones and weakly-to-moderately de- veloped brownish paleosols. There are two fine conglomerate beds near the top. The paleosols are usually relict soils, ~0.2–0.5-m-thick, light brown to brownish, argillous, massive-to-blocky or weakly-developed peds, occasionally mottled, and with many small carbonate nodules

Wu et al., 2006 below (Zan et al., 2015)(Fig. 3). From the pedological features, they can be roughly classified as cambisols affected by groundwater (FAO, 1988). In fact, the stratigraphy comprises several cycles of fining-up- wards sedimentary associations from greenish fine sandstones/fine conglomerates to light yellowish-brown, massive and occasionally mottled siltstones and mudstones, and finally brownish paleosols (Zan et al., 2015)(Figs. 3, 4a). These are typical features of a distal flood- plain (Reading, 2009; Fang et al., 2016)(Fig. 4). For convenience and ( continued ) S16–23 Xiao et al., 2012 continuity, we name these parts of the stratigraphy as successions S48–56 (Fig. 4a) and interpret them as floodplain deposits (channel

Table 1 S: Succession; MN: European land mammal zone ( Steininger, 1999 ). sandstones/fine conglomerates and overbank siltstones-mudstones-

466 X. Fang et al. Earth-Science Reviews 190 (2019) 460–485

Fig. 3. The photos on top showing the drilling of the top loess-covered sequence at Tashan and the representative lithology of cores (Top two: Brownish pedo- genetically-impacted mudstone layers or weakly developed paleosols; Bottom two: Grayish-light yellowish massive siltstones). Cross-section showing the strati- graphic sequence in the Tashan syncline and the relationship between the Xiejia section, Tashan section and Tashan drilling. Note that the stratigraphic division and its linked successions are used with the division presented in this study (see Table 1 for details and differences with previous divisions). Numbers in circles indicate mammalian fauna listed in Table 1. Mahalagou Fm. (gypsum member) + Shuiwan Fm. (mudstone member) = previous Mahalagou Fm.; Guanjiashan Fm. = previous Xianshuihe Fm. (See Table 1 and Suppl. Fig. 2 for details). paleosol complexes) (Reading, 2009; Fang et al., 2016). river, to alluvial fan (Figs. 4g, 5). The fossil mammals were found to However, actually in the field the most impressive stratigraphic occur at two levels, at the depths of 175–195 m for extinct rhino Par- boundary is the end of the thick gypsum dominated unit (S24–27) elasmotherium sp. and three-toed horse Hipparion dongxiangense (Fig. 3k, within the Mahalagou Fm., while the boundary between the upper part l) (Han et al., 2015), and at 230–235 m for extinct rhino Chilotherium (S28–32/33/37) of the Mahalagou Fm. and the above Xiejia Fm. is wimani by ourselves (Yang et al., 2017). We classified them as the 5th unclear from their lithology (Fig. 4h and Table 1). Furthermore, the level of fauna in the XB (Table 2; Figs. 4g and 5i, j). Many equivalent boundary between the Xiejia Fm. and the above Chetougou Fm. is also fossil mammals, called the Xiamen fauna, have been found at other sites variable depending on an unstable occurrence of a lenses of a 1–2 m in the XB (Table 2 and Fig. 1b). All of these mammal sites typically thick cross-bedded sandstone bed (S41). But the closely occurred three contain Hipparion sp. of the well-documented Hipparion fauna of the thick greenish gypsiferous sandstone beds that contain the Xiejia fauna Baodean-to-Bahean Stages of the CLMZ (equivalent to the Turolian to form a basin-wide mark bed (three greenish belts-in-one group) (S37- Vallesian (MN9–13) of the ELMZ (Table 2)). lower S39) (Fig. 4). Thus, we re-divide the boundaries of the Maha- The lower part of the section (336–199 m) is characterized by yel- lagou, Xiejia and Chetougou Fms. at S27/28 and lower / upper S39, lowish-brown massive mudstones and siltstones, intercalated with respectively (Fig. 4 and Table 1) (see Section 7.1 for details). several thin grayish-green massive-to-horizontally laminated marlite beds and numerous paleosols that are very similar to, but generally 3.1.2. Mojiazhuang section better developed than, those in the Guanjiashan and Chetougou Fms. in The Mojiazhuang (MJZ) section (36o41′07.08″ N, 102o04′15.78″ E the SW-GJS sections (Figs. 4, 5). These paleosols can all be classified as to 36o41′06.95″ N, 102o03’04.47″ E) is immediately south of luvic cambisols (FAO, 1988). The upper part of the lower section lacks Mojiazhuang village in the eastern XB, ~25 km northeast of Xining City paleosols but has occasional thin grayish-yellow sandstone and fine (Fig. 1 and Suppl. Fig. 1). It contains abundant newly-discovered fossil conglomerate beds. The gravels in the conglomerate beds are composed mammals that constitute the uppermost (youngest) mammal stage of primarily of clasts of metasandstones and siliceous rocks, such as the famous Hipparion fauna, thus forming the uppermost part of the quartzite and schist (Fig. 5), and they exhibit obvious imbricate struc- basin sedimentary sequence (Han et al., 2015; Yang et al., 2017) tures that indicate a dominant southerly transport direction (Fig. 4g). (Tables 1, 2 and Figs. 4g, 5). It is 336-m-thick and characterized by a Like those in the Guanjiashan and Chetougou Fms. in the SW-GJS sequence of mudstones and siltstones interbedded with paleosols, sections, these sediments are interpreted as floodplain sediments with coarsening upward into imbricated boulder conglomerates, indicating a the massive siltstones and mudstones deposited by overbank sheet progressive change of sedimentary facies from floodplain, via braided floods; the paleosols developed on these sediments after the floods, the

467 X. Fang et al. Earth-Science Reviews 190 (2019) 460–485

(caption on next page)

468 X. Fang et al. Earth-Science Reviews 190 (2019) 460–485

Fig. 4. Cenozoic stratigraphic sequences of the Xining Basin that form the basis of this study (Dai et al., 2006; Zan et al., 2015; Yang et al., 2017; Zhang et al., 2017). Note that the stratigraphic sequence of the Houwan section mentioned in the text is not shown here due to space limitations but shown in Fig. 9a. For convenience, we continue the numbering of the successions below succession 13, but show the succession numbers in italics to distinguish them from the same numbers that are used for the Cretaceous stratigraphy in the XJ section (QOET, 1977). Stratigraphic divisions and ages in sections are newly determined in this study, with the previous divisions plotted in column h (see text for details). The Xiejia-Tashan drilling sections and the Bingling Shan section are redrawn from Dai et al. (2006) and Zan et al. (2015); the Houwan section from Zhang et al. (2017); the Mojiazhuang section from Yang et al. (2017); and the combined red clay-loess sequences from Lu et al. (2004) and Zhang et al. (2017). thin green massive-laminated marlite was deposited in small short-lived inferred paleocurrent directions are also similar to those of the GJS shallow lakes on the floodplain, and the sandstones and fine con- section (Fig. 4c, d). glomerates are the deposits of small tributary channels (Reading, 2009; The middle terrace sequence is composed of about 5 m (HW section) Fang et al., 2016). Since the lowermost part of the paleosol complexes and 3 m (CJB section) of relatively loose gravels with a striking im- forms the marker bed (later confirmed by magnetostratigraphy) (cf. bricated structure showing a general westward direction with an Fig. 9), which correlates with succession 59 in the Houwan section, for overlying ~0.5-m-thick sandy siltstone bed (Fig. 4c, d). The composi- convenience we continue to name this part of the sequence as succes- tions of the terrace gravels in the HW and CJB sections are different sions S58 to S69 (Fig. 4c, g). from those in the basin sediments but are the same as that of the present The middle part (199–84 m) of the MJZ section consists mainly of Huang Shui, and thus they are regarded as higher terraces (T13 and thick brownish-yellow massive mudstones and siltstones, intercalated T14) of the Huang Shui (Zhang et al., 2017)(Fig. 4c, d). with several ca. 1–2-m-thick bluish-gray fine-to-coarse conglomerates The upper eolian sequence is continuous from the terrace sequence and grayish-yellow sandstones. Several cycles of conglomerate-sand- and consists of two parts. The lower part is a brownish red clay se- stone to siltstone-mudstone are present. The gravels are mainly com- quence which is 36-m- and 39-m-thick in the HW and CJB sections, posed of clasts of metasandstone and quartzite and have an imbricate respectively. The upper part is a yellowish loess sequence with many structure with a primarily southerly direction (Fig. 4g). We interpret thin, yellowish-brown paleosols and is about 60-m and 20-m-thick in them as braided river deposits with the conglomerates, sandstones and the HW and CJB sections, respectively (Zhang et al., 2017)(Fig. 4c, d). siltstones-mudstones being deposited by main channel, bar and over- Given the very thick loess deposits in the XB, the loess sequence here is bank processes, respectively (Figs. 4g, 5e–h). By successional order, we undoubtedly partially eroded and this is clearly indicated by erosion designated this part of the stratigraphic sequence as successions S58 to surfaces at the top and bottom of the loess sequence (Fig. 4c; see also S59 (Fig. 4g). Fig. 9). Combined with the loess sequences on terrace T12 at the top of The upper part (84–0 m) of the MJZ section is composed of dis- the GJS section and other Huang Shui terraces, we have constructed a tinctive thick (~5–20 m) gray pebble-to-cobble conglomerates inter- composite of the entire eolian red clay and loess-paleosol sequence, calated with thin, light brownish-yellow gravelly siltstones (Fig. 5b–d). which is illustrated in Fig. 4f. The pebble-to-cobble conglomerates are weakly imbricated or massive, To obtain a complete sequence of the Cenozoic stratigraphy in the poorly sorted and matrix supported, indicative of alluvial fan sediments XB, and to resolve the differences in the interpretation of the previous probably with an airborne dust component (Fig. 5b–d) (Reading, 2009; magnetostratigraphy and the age of the XJ fauna, we investigated and Fang et al., 2016). Again, we named this part of the stratigraphy as sampled several new sections in the central XB, where the stratigraphy successions S60 and S62 and assigned it a new name, the Mojiazhuang is spatially consistent and the upper parts are almost horizontal. Fm., for reasons given below (Fig. 4g). A thin, homogenous, loose, yellow, typical eolian loess layer un- conformably caps the top of the MJZ section (Fig. 5a). All the strati- 3.2. Measured new sections: East Xining-Shuiwan-Guanjiashan composite graphy in the MJZ section occurs nearly horizontally. section

The East Xining-Shuiwan-Guanjiashan-Houwan composite section 3.1.3. Houwan and Caijiabao sections (EXN-SW-GJS) is in the basin center. The basal part is in East Xining The Houwan (HW) and Caijiabao (CJB) sections are in fact upper- City at the mouth of the Huzhu River, the lower and middle parts are to most extensions of the stratigraphic sequence of our newly dated the northwest in Shuiwan, and the upper part is in Guanjiashan (Figs. 1, Shuiwan-Guanjiashan (SW-GJS) section below (Figs. 2, 4c, d). They are 2 and Suppl. Fig. 1). The measured and sampled East Xining (EXN) about 2 and 3 km north of the GJS section, respectively (HW: section (36°34.77′N, E101°53.81′E to 36°35.05′N, 101°53.56′E) (Horton 36°41.456′ N, 101°50.041′ E; CJB: 36°42.93′ N, 101°49.70′ E) (Figs. 1, 2 et al., 2004) is 350-m-thick. We were able to trace a visually-distinctive and Suppl. Fig. 1). All the stratigraphic marker beds can be traced marker bed, consisting of the thick gypsum layers of the Mahalagou horizontally for considerable distances. Fm., to the 722-m-thick Shuiwan-Guanjiashan (SW-GJS) sections The sequences of the HW and CJB sections consist of three parts: the (36°39.48′N, 101°52.31′E to 36°40.55′N, 101°50.24′E) in the north upper eolian sequence of red clay-loess deposits, the middle Huang Shui (Dupont-Nivet et al., 2004, Dupont-Nivet et al., 2007; Dai et al., 2006) terrace sequence of gravels and overbank siltstone, and the lower basin and to replicate some 150 m stratigraphy, which was later confirmed by sequence of fluvial-alluvial sediments. The terrace and basin sequences magnetostratigraphy (cf. Fig. 8). Thus, the entire red bed sequence has are separated by a parallel unconformity (Fig. 4c, d). The measured a thickness of 912 m (Figs. 4d, e, 6). basin sequences of the HW and CJB sections are 136-m and 120-m- The lithology and stratigraphy of the red bed sequence in the EXN- thick, respectively. They are very similar to and can readily be corre- SW-GJS composite sections resemble those of the XJ section (Fig. 4). lated with the GJS section, especially the maker bed corresponding to The Qijiachuan Fm., at the bottom of the EXN section, is pseudo con- an association from grayish carbonate-strongly-cemented (concrete formably superimposed upon the Cretaceous red beds. It is composed of like) channel sandstones to overbank siltstone and mudstone-paleosol 61 m of brownish-red mudstone and siltstone intercalated with thin complexes, but with a significantly increasing content and thickness of gypsum layers in the lower part, and 22 m of distinctly alternating conglomerate and sandstone beds in the upper parts (Zhang et al., medium (~2 m thick) beds of grayish mudstone and grayish gypsum in 2017)(Fig. 4c, d). Thus, in the studied sections the upward-increasing the upper part (Figs. 4e, 6b). The Honggou Fm. is 141 m thick (from 61 conglomerate and sandstone beds were interpreted as reflecting the to 202 m) brownish-red massive mudstone and siltstone intercalated encroachment of a braided river on a floodplain (Reading, 2009; Fang with many irregularly occurred thin-to-thick (~1–5 m) white massive et al., 2016; Zhang et al., 2017). The composition of the gravels and the gypsum beds (Figs. 4e, 6b). The Mahalagou Fm. is characterized by

469 .Fn tal. et Fang X.

Table 2 Fossil mammals found in the stratigraphy in the Xining Basin.

Division of stratigraphy (succession & Fossil mammals (Compiled from Li et al., 1981; Qiu et al., 1981; QBGM, 1991; Wang et al., 2013; Han et al., 2015 and this study) age)* Locality and components Suggested age and ecologic environment

Red clay-loess sequences (3.6–0 Ma) Equus sp., Bos sp., Capra sp., H. sp. Mazegouan-Nihewanian (Villanyian-Villafranchian, Late Pliocene-Pleistocene Desert steppe MN16–18)** (3.6–0.01 Ma) Upper Guanjiashan Fm. to Mojiazhuang Xiamen (Hipparion) fauna: Xiaomen: Hipparion cf. Coelophyes, H. Bahean – Baodean (Vallesian - Turolian, MN9–13) Late Miocene Steppe Fm. (S61–69; 11–4.8 Ma) dongxiangense, H. sp.*, Chilotherium wimani*, Parelasmotherium sp., Gazella sp., (11.6–5.3 Ma) Cervidae, Giraffoidea, etc. Mojiazhuang: Hipparion cf. Coelophyes, H. sp. Lower to Middle Guanjiashan Fm. Diaogou (Gomphotherium) fauna: L. Tunggurian (L. Astaracian, MN7–8) Late Middle Miocene Steppe-Forest (Xianshuihe Fm.) (S47–60 Gomphotherium wimani, G. connexus (13.8–11.6 Ma) 17–11 Ma) Ledu: Stephanocemas chinghaiensis Qijia: Alloptox chinghaiensis, Plesiodipus leei, Gomphotherium wimani, Micromeryx sp. Lierbao: Gomphotherium wimani, Bunolistiodon minheensis,Sinomioceros noverca, Micromeryx sp., Oioceros (?) noverca Leidazhuang: Elasmofherini Danshuilu fauna: E. Tunggurian (E. Astaracian, MN6) Early Middle Miocene Steppe-Forest

470 Danshuilu: Megacricetodon sinensis, Heterosminthus orientalis, (16–13.8 Ma) Protalactaga tungurensis, Rhinocerotidae indet., (L.Shanwangian-E.Tunggurian/L. Orleanian-E. (E. Middle Mio.) Cervidae indet. Astaracian,MN5–6) (16–13.8 Ma) Qijiagoukou: Megacricetodon cf. sinensis. Longzhigou: Protalactaga tungurensis, Heterosminthus orientalis. Hanjiashan: Palaeomeryx sp., Sinomioceros, Stephanocemas sp., Cervidae indet. Shijia: Rhinocerotidae indet. Shuiwan: Elasmofherini, Rhinocerotidae indet., Antiodactyla indet. Chetougou Fm. (Upper S39–4624–17 Ma) Chetougou fauna*: Shanwangian-Gaolanshanian (Orleanian-Agenian, Early Miocene Forest Chetougou-Xiejia:? Eumyarion sp., Rodentia indet., Rhinocerotidae indet. MN1–5)* (L. Shanwangian-E. Tunggurian /L. Orleanian- (23–16 Ma)* Shuiwan: Rhinocerotidae indet., Antiodactyla indet. E. Astaracian,MN5–6) (E. Middle Mio.) (16–13.8 Ma) Xiejia Fm. (S28-Lower S3934–24) Xiejia fauna: Tabenbulukian (Chattian)* (Xiejian / Agenian, MN1–2) Late Oligocene Woodland(North Xiejia: Sinolagomys pachygnathus, S. cf. pachygnathus, Leporid indet., Sciurid (28.4–23 Ma)* subtropic shrub-tree sp., **Cricetodon youngi + [Eucricetodon youngi], Parasminthus (E. Early Mio.) grass) xiningensis + [Plesiosminthus xiningensis], Parasminthus huangshuiensis + [Pl. (23–20.4 Ma) huanghuiensis], Litodonomys lajeensis + [Pl. lajeensis], Yindirtemys

suni + [Tataromys suni], Yindirtemys xiningensis + [T. sp.], Tachyoryctoides Earth-Science Reviews190(2019)460–485 kokonorensis, Turpanotherium elegans + [Brachypotherium sp.], Elasmotheriinae gen. et sp. indet. + [Br. sp.], Sinopalaeoceros xiejiaensis + [Oioceros (?) xiejiaensis], and Mustelid sp., Ctenopharynodon. Nanchuanhe: Tataromys sp.

*This study; **Equivalent stage of the European mammal zone (Steininger, 1999*); [ ] indicates original identifications, and + indicates revised identifications byG.F. Chen, 1988, B.Y. Wang, 1997 and X.M. Wang et al., 2013; Words in bold indicate typical typical mammal components in the fauna which are of identical age; words in light-gray italics indicate the previously-interpretated mammal stage and equivalent age. X. Fang et al. Earth-Science Reviews 190 (2019) 460–485

(caption on next page)

471 X. Fang et al. Earth-Science Reviews 190 (2019) 460–485

Fig. 5. Lithology and representative photos of the Mojiazhuang (MJZ) section showing changes in sedimentary facies. (a) Eolian loess unconformably caps the top of the MJZ section. (b, f, h) Mudstones at different depths. (c, d) Overview and detailed photographs of pebble-to-boulder conglomerate-siltstone complexes (71–84m, MJZ Fm.). (e) Boundary of the conglomerate-mudstone complex (115 m). (g) Conglomerate (171 m). (i, j) Mandible of Chilotherium wimani (234 m). (k) Teeth of Parelasmotherium sp. (j) Teeth of Hipparion dongxiangense. (m) Basal paleosol (280 m). (n) Basal marlite (323 m).

198 m alternating near-uniformly thick (~2–3 m) white massive siliceous rocks, quartzite, schist and slate (Zhang et al., 2017). There- gypsum and brownish-red gypsiferous mudstone beds, which occur at fore, we interpret this formation as a typical floodplain. orbital forcing periodicities (Dupont-Nivet et al., 2007; Abels et al., The Guanjiashan Fm. is parallel-unconformably overlain by the 2011), and its top and bottom boundaries are very sharp and easy to river gravel bed of a higher terrace (T12) of the Huang Shui and by a recognize from its above thick mudstone beds and its lower mudstone 22 m of typical early Quaternary yellowish eolian loess (Zhang et al., beds intercalated with irregular thick gypsum beds (Honggou Fm.). 2017)(Figs. 4d, 6). The loess sequence in the XB contains many modern Thus the Mahalagou Fm. definitely serves as a reliable marker bed mammals represented by appearance of the true horse, Equus sp. (Figs. 4d, e, 6). These three formations demonstrate a clear change from (QBGM, 1991; Liu, 1992)(Table 2). a playa mudflat environment (Qijiachuan Fm.) to a playa salt-lake with marginal mudflats (Honggou and Mahalagou Fms.) environment, and 4. Sampling and measurements of the new sections they exhibit an upward-increasing salt-lake component (Fig. 4d, e). The Xiejia Fm. is characterized with a thick (251 m) brownish-red Oriented block samples were collected from the EXN-SW-GJS mudstone-siltstone dominated fine sediment unit intercalated with composite sections at a 50-cm interval; however, for thick gypsum some upper-decreasing thin gypsum beds and upper-increasing thin layers the sampling interval was increased to ~2–3 m, depending on the sandstone beds, forming roughly two parts. The lower part (127 m presence of thin layers of mudstone or siltstone. Totals of 220 and 630 thick) consists mainly of brownish-red massive gypsiferous mudstones cyclindrical samples were collected from the EXN and SW-GJS sections, and siltstones with several white-to-greenish thin gypsum beds in the respectively, using a water-cooled gasoline-powered motor drill. lower part, and several thin massive fine-grained sandstone beds in the Magnetic compasses were used to orient samples in situ. The cylindrical upper part (Figs. 4d, 6). Again, as in the XJ section, we interpret the samples were cut into three subsamples in the laboratory. The speci- lower part as a playa mudflat with an ephemeral salt-lake and the upper mens were subjected to progressive thermal demagnetization in 20 part as a distal floodplain (Fig. 4d). The upper part (124 m thick) is steps at a 50 °C interval from the natural remanent magnetization similar in composition and structure to the lower part, but it exhibits (NRM) to 550 °C, and then at 20–30 °C intervals up to 685 °C. The re- lighter colors and denser and thicker sandstones. It consists of light manences of one set of specimens were measured using a 2G Enterprises brownish-red mudstones intercalated with several 1–2-m-thick grayish- Model 755 cryogenic magnetometer housed in a field-free room (< 150 greenish sandstones with obvious cross-bedding or parallel bedding nT) in the paleomagnetic laboratory at the Institute of Tibetan Plateau (Fig. 6). We still interpret it as a distal floodplain environment with Research, Chinese Academy of Sciences. The other set (only for the increased transport energy (Fig. 4d). EXN-SW-GJS composite sections) was also measured with a 2G cryo- The Chetougou Fm., 166-m-thick, is characterized by numerous genic magnetometer under similar experimental conditions at the pa- small cycles of yellowish-brown mudstones and siltstones intercalated leomagnetism laboratory of the University of Tübingen, Germany. with numerous 1–4-m-thick grayish-yellowish massive-to-parallel or The characteristic remanent magnetizations (ChRMs) were suc- cross-bedded sandstones and distinct reddish-brown paleosols with cessfully isolated after thermal demagnetization to 200–300 °C (Suppl. clear luvic Bt horizons and illuvic carbonate Bk horizons (Fig. 6). The Bt Fig. 3). The ChRMs were determined using principal component ana- horizon is readily identified by a well-developed ped structure (rela- lysis after excluding samples with maximum angular deviations tively loosely compacted, with easily disaggregated fine (~2–4 mm) (MAD) > 15°. Overall, we isolated 159 reliable ChRM directions (72%) granular peds that consist of aggregated clays, silts, organic complexes from the EXN section and 492 (78%) from the SW-GJS sections. The and infilled earthworm burrows, excrements formed by seasonal cli- thermal demagnetization results, and the measurements of the tem- mate-induced swelling and shrinking, and biological activities), nu- perature-dependence of magnetic susceptibility (κ-T curves) using a merous voids and shiny reddish Fe-enriched clay coatings on peds and MFK1-FA Kappabridge equipped with a CS-4 high-temperature furnace void walls. The Bk horizon is characteristically much lighter in color (Agico Ltd., Brno, Czech Republic), demonstrate that magnetite and/or (grayish-yellow) with abundant coatings, infillings and small hematite are the dominant remanence carriers (Suppl. Figs. 3 and 4). In (~0.5–1 cm) carbonate nodules (Fang et al., 2016). From the pedo- the EXN-SW-GJS composite section, the normal and reversed ChRM features, these brown paleosols can generally be classified as haplic-to- directions with Virtual Geomagnetic Poles latitude (VGP lat.) smaller luvic cambisols (FAO, 1988). We interpret the sandstones as small river than 20° were excluded to remove outliers and transitional directions. channel deposits and the mudstones-siltstones as overbank deposits A reversal test was used to assess the nature of the primary ChRMs between channels, with the paleosols resulting from the effects of (McFadden and McElhinny, 1990; Tauxe, 1998). The mean directions of pedogenesis of the overbank deposits during intervals of flooding normal and reversed polarities from the EXN and SW-GJS sections are quiescence, in a floodplain setting (Reading, 2009; Fang et al., 2016). both antipodal, suggesting that they are a primary record of the Earth's They comprise many small fining-upward channel-overbank-paleosol magnetic field (Suppl. Fig. 5A and B). A bootstrap test further demon- cycles (Figs. 4d, 6). strated that the reversed polarities are flipped to their antipodes and the The Guanjiashan Fm. (formerly called the Xiashuihe Fm.) is similar two means are statistically identical at the 95% confidence level (Suppl. in lithology and stratigraphy to the Chetougou Fm; however, the de- Fig. 5C and D). Thus, the paleomagnetic datasets for the EXN and SW- positional cycles are larger and occur less frequently, and the channel GJS sections both pass the bootstrap reversal test (McFadden and deposits mainly consist of 2–3-m-thick grayish fine conglomerates with McElhinny, 1990; Tauxe, 1998). a clear imbricated structure, and dull grayish-yellowish massive sand- stones. The overbank deposits are thick with multiple minor cycles of massive siltstones/mudstones and paleosols or paleosol complexes 5. Magnetostratigraphy of the Xining Basin (Figs. 4d, 6). The brown paleosols resemble those of the Chetougou Fm. Measurements of the ab-faces of gravels in the imbricated conglomer- 5.1. Xiejia section ates indicate a dominant southerly paleocurrent direction (Figs. 4c, d, 5), and the gravel composition consists mainly of meta-sandstones, At the request of the discoverers, Wu et al. conducted detailed pa- leomagnetic dating of the 250-m-thick stratigraphic intervals bearing

472 X. Fang et al. Earth-Science Reviews 190 (2019) 460–485

Fig. 6. Stratigraphy, lithology and sedimentary facies of the East Xining-Shuiwan-Guanjiashan composite section. Legends are the same as in Fig. 4.

473 X. Fang et al. Earth-Science Reviews 190 (2019) 460–485

Fig. 7. Correlation of the magnetostratigraphy of the Xiejia section with those of the nearby Tashan section and Tashan drilling site and the East Xining and Shuiwan sections. Note for comparison and confirmation, new interpretation of magnetic polarities of the XJ fauna-bearing stratigraphic intervalby Wu et al. (2006) and its determined sedimentation rates were plotted on the right (Fig. 7k–q). Numbers in circles indicate the mammalian fauna listed in Table 1. VGP: Latitude of the virtual geomagnetic pole; Inc.: Inclination. the XJ fauna in the XJ section (Wu et al., 2006). They found 18 normal (Fig. 7a–c). The pattern of magnetic polarity zones of the TS section is and 17 reversed polarity zones. Based on the precondition that the XJ almost identical to that of the equivalent stratigraphic interval of the XJ fauna had an age of ~20 Ma, in the Early Miocene Xiejian Stage section, except for the intervals at the beginning of the Miocene where (equivalent to the Aquitanian Stage of the European marine fossil they interpret several stratigraphic hiati which we could not find in zones, or MN 2–3) (Li and Qiu, 1980; Li et al., 1981, 1984; Qiu and Qiu, field evidence (Fig. 7) (we will discuss this later). They correlated the 1990; Deng et al., 2006), they correlated the observed polarities with observed polarity zones to Chrons 16n to 5Br (~35.5–15.8 Ma) and Chrons 5Bn.2n to 6Cr of the geomagnetic polarity time scale (GPTS) determined the age of the XJ fauna to be ~25 Ma, confirming the re- (Lourens et al., 2004). This correlation indicated that the age of the XJ liability of the magnetostratigraphy of the XJ section by Dai et al., 2006 fauna was about 20.5 Ma (Fig. 7l–p). (Fig. 7a–c, g–i). At almost the same time, Dai et al. conducted a detailed paleo- To test the reliability and to expand the magnetostratigraphy of magnetic study of the entire XJ section and observed 23 normal pola- both the XJ and TS sections, Zan et al. (2015) obtained a drill core from rities (N1-N23) and 24 reversed polarities (R0-R23) which they corre- the top of the TS section to obtain a complete sequence of the TS syn- lated with Chrons 5Br to 23r of the GPTS. This correlation constrained cline (Fig. 3). They clearly distinguished 10 normal polarities (N1–N10) the age of the section to ~52–17 Ma and that of the XJ fauna to ~25 Ma and 9 reversed polarities R1–R9 and found that N8–N10 correlated with in the Latest Oligocene (Fig. 7g–j), rather than the previously suggested intervals N1–N3 at the top of the XJ section and with N1–N5 in the TS biostratigraphic age of ~20 Ma in the early Miocene (Li and Qiu, 1980; section over equivalent stratigraphic intervals. Thus, they continued to Li et al., 1981, 1984; Qiu and Qiu, 1990; Deng et al., 2006). correlate them with Chrons 5ABn to 5En of the GPTS, expanding the top sequence of the TS syncline to about 13.5 Ma (Zan et al., 2015) (Fig. 7d–f). 5.2. Tashan (TS) section and drill core

Xiao et al. (2010) carried out a detailed paleomagnetic study of the 5.3. East Xining - Shuiwan - Guanjiashan composite sections TS section which is parallel to the XJ section at the opposite (northern) limb of the TS syncline (Fig. 3). The lithology of the TS section is almost Horton et al. (2004) sampled the lowermost part of the EXN section, identical to that of the XJ section for the duplicated interval (Fig. 7a, g). at a large sampling interval. They found 3 pairs of normal and reversed They found 23 pairs of normal and reversed polarities and correlated polarities (N1–N3, R1–R3) in the EXN section (Fig. 7s). Based on the them with Chrons 5Cn to 16n.1n of the GPTS (Xiao et al., 2010, 2012) microfossils, they tentatively correlated them to chrons 18n–27r

474 X. Fang et al. Earth-Science Reviews 190 (2019) 460–485

Fig. 8. Correlation of Cenozoic magnetostratigraphic sections in the Xining Basin. Circled numbers 1–5 indicate mammal fauna listed in Table 2 and Fig. 4. Light-red dotted lines are marker beds facilitating the correlation. C: Conglomerate; S: Sandstone; G: Gypsum; M: Mudstone/Siltstone; Ma: Marlite; XF: Xiejia fauna; QJC: Qijiachuan Fm. Other abbreviations as in Fig. 7. (For interpretation of the references to color in this figure legend, the reader is referred to the web version ofthis article.)

(~39–62 Ma). Dai et al. (2006) reinterpreted these polarities and very much that of the Xiejia-Tashan sections (Fig. 8). The equivalent thought that they roughly duplicated polarities N19–N22 at the bottom stratigraphic intervals between the measured uppermost EXN section of the XJ section (Dai et al., 2006)(Fig. 7s). Later, Dupont-Nivet et al. and the lowermost SW section, constrained by gypsum marker beds, (2007) conducted detailed sampling of the lowermost part of the exhibit the same sequence of polarity zones from N19 to R22, con- Shuiwan section to constrain the age of the Eocene and Oligocene firming the reliability of the lithological marker bed(Fig. 8k–n). This boundary. They found 6 pairs of normal and reversed polarities observed pattern of polarity zones resembles very well that of the XJ-TS (N1–N6, R1–R6) and correlated them with Chrons 12n to 18n of the drilling section (Dai et al., 2006; Zan et al., 2015) and that of the GPTS; this not only confirmed the reliability of the lithologic and measured interval of the partial SW section (Dupont-Nivet et al., 2007; magnetic correlations of the XJ section for the same stratigraphic in- Abels et al., 2011)(Fig. 8), demonstrating the reliability of the magnetic tervals, but it also indicated that the Eocene and Oligocene boundary polarity records of both sections. The prominent gypsum marker bed was marked by the termination of visually-striking thick gypsum layers between the two sections, and the two levels (2nd and 3rd) of fossil dated to exactly 34 Ma, consistent with the global cooling record from mammals found in the middle (S42) and upper (S48) parts of the SW- marine sediments (Dupont-Nivet et al., 2007)(Fig. 7r). This magne- GJS sections, which suggested Agenian to Astaracian-equivalent ages of tostratigraphic interval was confirmed by Abels et al. (2011) by more the early-to-middle Miocene (Liu, 1992)(Table 2 and Figs. 4d, 8m), detailed paleomagnetic results in conjunction with astronomic forcing were used to constrain the paleomagnetic polarity correlation. An ad- chronology. ditional level (4th, at late Tunggurian stage of the late Middle Miocene) We measured the entire EXN-SW-GJS composite sections (with an of fossil mammal Elasmofherini in the striking thick grayish sand con- accumulated thickness of 912 m) by duplicating the 150-m interval glomerate beds of the nearby Leidazhuang section (Table 2), ~9 km between the EXN and SW sections which was constrained strictly by top northwest (closer to mountains) of the SW-GJS sections (see Fig. 1b and and bottom boundaries of the marker beds of the Mahalagou Fm., Suppl. Fig. 1 for the location), can be roughly traced to the stratigraphy which was identified by nearly equal occurrence of alternating white immediately above succession S52 in the SW-GJS sections (Fig. 4d), and gypsum and brownish red gypsiferous mudstone beds at orbital forcing provides age constraint to the top of the SW-GJS sections in the late periodicities (Dupont-Nivet et al., 2007; Abels et al., 2011); the purpose Middle Miocene (Fig. 8m). Especially the stratigraphic sequence of the was to provide a crosscheck of the paleomagnetic stratigraphy. A total HW section immediately above the SW-GJS sections can well be cor- of 31 normal polarities (N1–N31) and 30 reversed polarities (R1–R30) related both in lithology (marker beds of paleosols-marlite complexes) were obtained in the EXN-SW-GJS composite sections, which resembles and magnetostratigraphy with the lower part of the MJZ section where

475 X. Fang et al. Earth-Science Reviews 190 (2019) 460–485

Fig. 9. Correlation of the uppermost part of basin magnetostratigraphic sequences from the Houwan and Caijiabao sections studied by Zhang et al. (2017). For comparison, the top of the magnetostratigraphic sequence of the MJZ section and of several loess sequences on terraces above the basin sediments are also shown (modified from Zhang et al., 2017). The blue and light-red shaded zones are sandstone and paleosol marker horizons which facilitate correlation of our sections and the Mojiazhuang section (Yang et al., 2017). All the magnetostratigraphic correlations are based on distinctive stratigraphic markers and mammal fossils. Note that lack of reversed polarities in N5 in the Caijiabao section (c) might be due to large sampling intervals and occurrence of thick conglomerate beds. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.) the 5th level of fossil mammals and high quality of magnetostratigraphy (~25–47 Ma) and the upper and lowermost parts show poor quality and place a robust age of 12.7 Ma for the bottom of the stratigraphy (Zhang seem to have lost some short polarities, thus bearing large uncertainties et al., 2017; Yang et al., 2017)(Figs. 8h–j,o and 9). This indicates the in detailed correlations. This means that the prominent normal polarity- top of the SW-GJS sections must be younger than the bottom of the HW dominated zones N19–22, and the long reversed polarity-dominated section at ~12 Ma (Zhang et al., 2017). Taking all these as age con- zones R15–R18 (above) and R22–R24 (below) in the lower part of straints and keeping in mind matching lithology with sedimentation section, can be respectively correlated with the normal polarity Chrons rates determined by interpreted magnetic polarities, we were able to 16n–18n and reversed polarity Chrons 12r–13r (above) and 18r–20r readily correlate the observed polarities of the SW-GJS sections with (below) of the GPTS (Gradstein et al., 2012). Likewise, the prominent Chrons 24n to 5ABn of the GPTS (Gradstein et al., 2012)(Fig. 8g,k,l). normal polarity-dominated zones N8–15 are correlated with the simi- This indicates that the middle part of the observed polarities N8–N25 larly-patterned Chrons 8n–12n of the GPTS. The several striking long show high quality and can well be correlated with chrons 8n–21n normal and reversed polarity zones N1 to R7 in the upper of the SW-GJS

476 X. Fang et al. Earth-Science Reviews 190 (2019) 460–485 sections, constrained by four levels (2nd to 5th) of the fossil mammals section very close to the earliest late Middle Miocene. Thus, we prefer in the Early to Late Miocene found in or above the sections, can ten- our accepted interpretation of the polarity sequence which gives an age tatively correlated with chrons 5ABn to 7r. That is, the long reversed of 13.5 for the top (S52) of the SW-GJS section. It is also consistent with polarity-dominated zones R4–R7 are correlated with Chrons 6r–7r; the the age of S52 in the XJ-TS section within error ranges (Fig. 8). Espe- long intervals N3 and N4 that occur in the time with increasing sand- cially, the marked bed of sandstone-marlite- paleosols complex at the stone and fine conglomerate components, are likely related to anin- bottom of the HW section, the extended top sequence of the SW-GJS crease in sedimentation rate, and might respectively be correlated with section, can be traced well to the bottom of the MJZ section (Fig. 4c, d, the prominent long Chrons 5Dn and 6n; the last long normal polarity g) where both fossil mammals and high quality of magnetostratigraphy intervals N1 and N2 are therefore most likely correlated with the two determined the section at 12.7–4.8 Ma (Yang et al., 2017)(Fig. 8j, o). long Chron complexes 5ABn-5ADn and 5Cn of the GPTS (Gradstein Since the marked bed in the HW section is very close to the top of the et al., 2012)(Fig. 8g,k-m). For the lowermost part of the section, the SW-GJS section, this constrains the age of the top of the SW-GJS section sequences N25–N31 and R25–R30, characterized by alternating, should only slightly older than the marked bed at bottoms of the HW roughly equally-spaced polarity zones, are correlated with the simi- and MJZ sections at ~12 Ma (Fig. 8h–o). Thus, our interpretation of the larly-patterned Chrons 21n–24n (Fig. 8g–i). This interpretation of the magnetostratigraphy of the top of the SW-GJS section at age of 13.5 Ma observed magnetic polarities yields convincing variation of the polarity- is plausible for the stratigraphic sequence. In contrast, the two alter- determined sedimentation rate with the lithology (Suppl. Fig. 7d) and native correlations proposed by the correlation software yield ages of previously dated intervals of the Shuiwan and East Xining sections ~18 Ma and 16 Ma for the top of the SW-GJS section, much older than (Horton et al., 2004; Dai et al., 2006; Dupont-Nivet et al., 2007; Abels both the late Middle Miocene age constraint by the nearby 4th level of et al., 2011; Bosboom et al., 2014)(Fig. 8g,p,q). The interpreted cor- fossil mammals (Suppl. Fig. 6; Table 2) and the reliable stratigraphic relation also agrees well with the Xiejia-Tashan section in respect of the marker bed, fossil mammal and magnetostratigraphy constraints by the polarity pattern and its correlation and also in determining ages of HW and MJZ sections above (Zhang et al., 2017; Yang et al., 2017) stratigraphic boundaries (Fig. 8). Based on this correlation, the EXN- (Figs. 8h–o and 9). SW-GJS composite sections span the interval ~54–13.5 Ma (Fig. 8g,k- Both alternative correlations proposed by the method of Lallier et al. n). (2013) of the SW-GJS-EXN composite sections yield age of ~57 Ma for Alternative interpretations of the observed magnetic polarities of the bottom of the section, causing same but also abnormal matches of the EXN-SW-GJS composite sections can be made based on calculating sedimentation rates with lithology (Suppl. Figs. 6, 7). These correla- the degree of matching with the GPTS. tions lead to four-fold fluctuations in sedimentation rate for the pro- using the method proposed by Lallier et al. (2013) (Suppl. Fig. 6). minent marker beds of the Mahalagou Fm. (at intervals of ca. The principle of the method is a quantitative assessment of the 36–43 Ma) (Suppl. Fig. 7a, e, f). It is also much higher (about five-times) matching degree (correlation costs) between different interpretations of in sedimentation rate than our accepted correlation which confirms observed magnetic polarities and the GPTS and yields best correlation previous magnetostratigraphy (Dupont-Nivet et al., 2007; Abels et al., (Lallier et al., 2013). We calculated the correlation costs for potential 2011; Bosboom et al., 2014) (Suppl. Fig. 7). Furthermore, it would correlations (input parameters: load date from the EXN-SW-GJS com- imply that in similar mudflat-distal floodplain lithology, intervals with posite sections; load reference from the GPTS 2012 (Gradstein et al., more components of reddish mudstones (intervals of ca. 43–57 Ma) the 2012); Nbr Res: 6000; Max path: 6000; Gap Factor: 0; Max Subst: 1; sedimentation rate is about five times lower than in other intervals Subst Dist: 1). The “Gap Factor” of 0 means that we did not add penalty (intervals of ca. 39–43 Ma above) (Suppl. Fig. 7a,b,e,f). The more sig- to a gap between the two columns. The “Max Subst” of 1 means that one nificant constraint perhaps comes from the occurrence of the marker given chron of the reference scale can only be correlated to a maximum bed of black-grayish mudstones of marshland (S15) at the top of the of one polarity interval of the studied section, thus avoiding a large Qijiachuan Fm. (Fig. 4e) where abundant broadleaved trees occur and number of miss-matches in the correlation. Meanwhile, the “Subst Dist” carbon isotope of bulk samples shifts to a large negative values both in of 1 means that one chron is considered to evaluate the local sedi- the EXN and XJ sections which we thought it could be a record of the mentation rates (see Lallier et al., 2013 for details). The major weakness event of the Early Eocene Climatic Optimum (EECO) at about 52 Ma of the method is that the computation seems to be based mostly or (Zachos et al., 2008) (personal communications with Yahui Fang and mechanically on one-by-one correlation of observed polarities with Yongli Wang, paper in preparation). This is in good agreement with our each polarity of the GPTS, rather than grouped polarity patterns. And of preferred correlation of the observed magnetic polarities, rather than course, it will not care about the match with lithology. The calculated the two alternative correlations where this event was determined at “best-fit correlations” derived from the 6000 minimum cost outputof about 50 Ma (Suppl. Fig. 6). More serious is that both alternative cor- the EXN-SW-GJS composite sections suggest that the observed magnetic relations determined the bottom of the EXN section would reach about polarities correlate with the magnetostratigraphic Chrons C5En to C25n 57 Ma (Suppl. Fig. 6). This means it must cover the other striking event, (~18–57 Ma) or C5Cn.1n to C25n (~16–57 Ma), with little significant the Paleogene/Eocene Thermal Maximum (PETM) at about 55 Ma difference between them. This result resembles our preferred correla- where global carbon isotope was suddenly negatively shifted by 1.5‰ tion above, except for the uppermost and lowermost parts (Suppl. (Zachos et al., 2008), while in the Asian continent records it was ne- Fig. 6). It correlates the uppermost four long normal polarities N1–N4 gatively shifted by 4–6‰ (Bowen et al., 2002, 2005). However, we with chrons 5En to 6An or 5Cn to 6n (Suppl. Fig. 6a, c). The first cor- could not see this event at the bottom of the EXN section (even there is relation cost result causes an anormal (about three-fold) pulse increase some buried interval but the buried stratigraphy can be traced to see in of the sedimentation rate during polarity N4, while there is no sig- nearby outcrops). Thus we prefer to reject these two correlations, and nificant lithologic change (Suppl. Fig. 7a,e). The suggested second accept our preferred interpretation above which results in reasonable correlation cost result overcomes the weakness of the pulse increase of matches of sedimentation rates with lithology and with the previous the sedimentation rate, but produces an about four-fold persistent se- magnetostratigraphic and paleontological chronologies (Fig. 8 and dimentation rate increase from N1 to N2 where lithology shows no Suppl. Fig. 7a,d). corresponding coarsening (Suppl. Figs. 6c and 7a,e). Furthermore, the 4th level of fossil mammals, the Gomphotherium Fauna at the equiva- 5.4. Mojiazhuang section lent succession 54 (marker bed, the only striking conglomerate bed), was found in the thick conglomerate bed at the nearby Leijiazhuang The MJZ section yielded 17 normal and 16 reversed polarity zones, section. This fauna determines a late Middle Miocene age designated N1–N17 and R1–R16, respectively and was previously (13.8–11.6 Ma) (Table 2). It constrains the top (S52) of the SW-GJS analyzed and dated in Yang et al. (2017) (Fig. 8o). Since the 5th level of

477 X. Fang et al. Earth-Science Reviews 190 (2019) 460–485 fossil mammals (Hipparion fauna) indicates an age corresponding to the the Chetougou Fm. and N7 in the Xiejia Fm. in the Tashan section, and Bahean-Baodean Stages (Vallesian-Turolian) of the late Miocene they correlated R7 with chron 6r and N7 with chron 6Bn, while the (Table 2 and Fig. 8o). Reliable magnetostratigraphic correlation pre- many frequently occurring short normal and reversal chrons in chrons viously reported in Yang et al. (2017) and Zhang et al. (2017) assigns 6An-6Ar (~1.8 Ma) are missing within the unconformity. Dai et al. ages for the top and bottom of the MJZ section of 4.8 Ma and 12.7 Ma, (2006) thought there is no unconformity between the Xiejia and Che- respectively; in addition, the marker beds of brown paleosol (cambisol) tougou Fms. In the Xiejia section, and the missing of the many short complexes are dated to ~12–12.4 Ma, and are well correlated with si- normal and reversal chrons probably is due to large sampling intervals; milar paleosol complexes both in terms of marker beds and the age of thus they correlated R4–N5–R5 with chrons 6r–6Ar (Fig. 7a–j). We succession S59 of the Houwan section (Fig. 8g–j,o). carefully checked again both sections and other sections in the field, but did not find any convincing evidence for an unconformity between the 5.5. Houwan and Caijiabao sections and loess sequence Xiejia and Chetougou Fms. The boundary between these two formations is the bottom of a channel deposit of sandstone body (S41) that has The results from this section are reported in Zhang et al. (2017). Lu many evident cross-beddings and a normal small scale base erosional et al. (2004) carried out a paleomagnetic study of the red clay se- surface (Reading, 2009)(Fig. 10 and Suppl. Fig. 2). An unconformity quences on the terrace gravel beds of the Houwan (HW) and Caijiabao will present clear large scale erosion surface usually with obvious (CJB) sections. The polarity zone patterns of the red clay sequences they weathering remains on the surface or clear occurrence (strike/dip) observed in the red clay parts of the HW and CJB sections clearly re- difference between two contact stratigraphy. This is not the case forthe semble the one reported in Zhang et al. (2017); Fig. 9a, b). However, contact between the Xiejia and Chetougou Fms. (Fig. 10). Especially, they correlated them with Chrons 3An.2n to 5r and Chrons 5An–5ADn, Wu et al. (2004) did very high quality paleomagnetic work for this part yielding ages of about 6.5–11.3 Ma and 12–14 Ma for the red clay se- of the stratigraphic interval in the XJ section, taking samples at much quences, respectively (see Fig. 6 in Lu et al., 2004). The sedimentation higher resolution of 20 cm intervals. They found nearly all the missed rates calculated from their interpretation of the observed magnetic short normal and reversal polarities in chrons 6An-6Bn in the 22–20 Ma polarities exhibit a pattern of frequent, large amplitude changes which interval when Xiao et al. (2012) claimed an unconformity (see our re- is inconsistent with the homogeneous nature of the eolian red clay; interpretation of their polarities in Fig. 7j,k). The only weakness of Wu moreover, their magnetostratigraphic interpretation is not supported by et al.'s (2006)’s work is that their interpretation of the magnetostrati- any robust independent evidence and independent age control. We graphy of the XJ fauna-bearing stratigraphic intervals in the XJ section reject their interpretations for the following specific reasons. First, our was forced to comply with the predetermined age of ~21 Ma (MN2 of continuous dating of the basin sediments, which is well-constrained by the Agenian in the ELMZ) previously suggested for the XJ fauna (Li fossil mammal ages, indicates that these terrace red clay sequences et al., 1981; Qiu and Qiu, 1990). This led to a significant mismatch of must be younger than the termination of the basin fluvial-alluvial se- the magnetic polarity record with the GPTS: for example, the correla- dimentation in the MJZ section, dated to about 4.8 Ma (Figs. 8o, 9d). tion of the short interval of two normal polarity intervals N5 and N6, Second, there is a continuous sequence contact from red clay to loess separated by the reversed polarity interval R5, with the long time in- below the unconformity occurs in the loess sequences. The loess se- terval corresponding to Chrons 5Dn to 5En; the correlation of the short quences above the unconformity were systematically dated by previous normal polarity N7 with the very long Chron 6n; and the correlation of work to about 2.3–2.2 Ma (e.g., Zeng et al., 1995; Li et al., 1996a,b; Lu the long normal polarity intervals N9, N10 and N17 with the very short et al., 2004; Zhang et al., 2017)(Fig. 9e), the red clay and bottom of the Chrons 6An.2n, 6AAn and 6Cn.3n, respectively (Fig. 7o,p). These cor- loess sequences on the HW and CJB terraces must have been formed relations resulted in many short intervals of abrupt, high amplitude (up between < 4.8 Ma and > 2.3 Ma. Thus, the observed normal-polarity- to almost seven-fold) changes in sedimentation rate (Fig. 7q) within an dominated zones of the red clay sequences in the HW and CJB sections interval of homogenous lithology (dominant brownish-red mudstones must record the Gauss (G) Normal Chron (Chron 2An), with the up- deposited on a floodplain) (Fig. 7l). permost short reversed polarity zone correlating with the lowermost Using the magnetostratigraphy of the XJ and Tashan sections (Dai part of the Matuyama (M) Reversed Chron (Chron 2r), dating the et al., 2006; Xiao et al., 2012), we re-correlated the observed polarity boundary of the red clay and loess sequences to the M/G boundary to zones of the XJ faunal-bearing stratigraphic intervals with Chrons 6n to 2.6 Ma (Zhang et al., 2017)(Fig. 9a–c). This is in excellent agreement the upper part of 12n of the GPTS (about 19.5–30.8 Ma). This corre- with all other records of the boundaries of the red clay and loess se- lation produces a good visual match to the GPTS and an almost uniform quences over the Chinese Loess Plateau (e.g., Ding et al., 1994; An sedimentation rate for the studied homogeneous lithology (Fig. 7j, k, q). et al., 2001; Han et al., 2011). This new correlation, together with that for the entire XJ section (Dai et al., 2006) and the parallel Tashan section (Xiao et al., 2012), all yield 6. Assessment of the magnetostratigraphic correlations and an age for the XJ fauna of ~25 Ma, i.e., latest Oligocene (Fig. 8). The constraints on mammalian evolution good match of the paleoclimatic proxy records from the XJ section, based on the chronology of Dai et al. (2006), with the global climate The magnetostratigraphy of the long continuous XJ section in the record (e.g., Long et al., 2011; Hoorn et al., 2012; Chi et al., 2013; southern limb of the Tashan syncline obtained by Dai et al. (2006), Dupont-Nivet et al., 2007; Zhang and Guo, 2014; Fang et al., 2015; confirmed by the magnetostratigraphy of the parallel Tashan section in Licht et al., 2014), provides further support for the magnetostratigraphy the northern limb of the syncline by Xiao et al. (2012), and by the and age constraint of the XJ fauna by Dai et al. (2006) (see below for Tashan drilling on the top of the syncline by Zan et al., 2015, provides a details). Thus, the XJ fauna should be assigned to the Tabenbulukian more robust interpretation of the XJ section (Figs. 3, 7). Both constrain Stage of the CLMZ (equivalent to the Chattian Stage of the ELMZ) (Li the Xiejia fauna at an age of ~25 Ma, over 4 Ma older than dated by Wu et al., 1984)(Table 2). In fact, many of the mammals are identical to or et al. (2006) (Fig. 7a–j,i–p). The observed magnetic polarities from the resemble those of the Tabenbuluk fauna at the western end of the Qilian Xiejia and Tashan sections show very similar patterns for the duplicate Shan (Li et al., 1984) (see Fig. 1 for location), with only a few species stratigraphic intervals. The only difference in interpretation is how to slightly younger in comparative age than the Tabenbuluk fauna. This correlate the observed magnetic polarities R4–N5 in the Xiejia section led Li and Qiu (1980) and Li et al. (Li et al., 1981) to conclude that that and R7–N7 in the Tashan section for the same stratigraphic intervals the XJ fauna was slightly younger than the Tabenbuluk fauna and fi- (S42–43) at the boundary between the Xiejia and Chetougou Fms., after nally assigned it to the early Miocene (Table 2). Recent detailed re- fixing correlations of other observed polarities in both sections search on the XJ fauna has revised several important species and casts (Fig. 7a–j). Xiao et al. (2012) suggested an unconformity between R7 in doubt on the zonation of the Xiejian Stage, because most of the

478 X. Fang et al. Earth-Science Reviews 190 (2019) 460–485

Fig. 10. Photos showing sequence at the boundary between successions S41 (bottom of former Chetougou Fm.) and S40 (top of former Xiejia Fm.) where Xiao et al. claimed an unconformity of missing 1.8 Ma stratigraphy occurs (Xiao et al., 2012). Note our photos indicate no unconformity lying between S41 and S40, but a conformable contact of a normal cross-bedded sandstone layer of channel deposit sweeps on a floodplain mudstone bed with some minor normal erosion surface by river channel above. reassessed and revised mammals, such as tsaganomyids, ctenodactylids, determining a late Tunggunian Stage of the late Middle Miocene in the tachyoryctoidids, hyaenodontids, and giant rhinoceroses, flourished CLMZ (Table 2). and disappeared in the Oligocene, with only a few Miocene newcomer The MJZ or Xiamen (Hipparion) fauna was dated by our magnetos- insectivores and rodents (Wang et al., 2013)(Table 2). This means that tratigraphy in the MJZ section at about 8–9.8 Ma, belonging to the the continuous use of the Xiejian Stage of the CLMZ to represent the Baodean Stage of the CLMZ (the late Miocene, equivalent to the Chinese land mammal evolution of the early Miocene is inappropriate Turolian (MN9–10) of the ELMZ) (Table 2 and Fig. 8). and misleading; rather, it is a late part of the Tabenbuluk Stage of the The Equus fauna in the loess sequence is dated roughly to the late Oligocene. Other alternative terms such as the Gaolanshanian Stage Quaternary (Mazegouan-Nihewanian Stages of the CLMZ) (equivalent or Suosuoquanian Stage from the fossil mammals (Gaolanshan fauna or to the Villanyian-Villafranchian (MN16–18) of the ELMZ (Table 2 and Suosuoquan fauna) of the real early Miocene in the adjacent Lanzhou Fig. 8). Basin, or in the Junggar Basin in Xinjiang (Wang et al., 2013), are plausible replacements for the Xiejian Stage of the Chinese early Mio- cene mammal zonation. 7. Stratigraphic division of the XB and its constraints on changes Two levels of abundant mammal fossils, the Chetougou fauna and in the sedimentary environment the Danshuilu fauna, are generally found in the lower and upper parts of the Chetougou Fm., respectively. They were assigned approximate 7.1. Stratigraphic division ages of late Shanwangian-to-early Tunggurian Stages of the CLMZ (equivalent to the late Orleanian-to- early Astaracian (MN5–6) of the The Cenozoic sequence of the XB exhibits a clear upward-increasing ELMZ (Table 2). However, paleomagnetic dating constrains the Che- trend in grain-size and sediment lightness, reflecting changes in sedi- tougou Fm. to about 22.5–16 Ma and the Chetougou fauna to about mentary environment in response to basin evolution and tectonic uplift 22.5–19.5 Ma (Fig. 8), which corresponds to the Gaolanshan fauna of of the surrounding mountains, as well as climate change. With this in the early Miocene in the Lanzhou Basin (Qiu and Qiu, 1990). Only the mind, based on lithology and lithofacies, we have proposed a new Danshuilu fauna was dated to about 15.5 Ma in the lowermost Guan- Cenozoic stratigraphic division for the XB with new paleomagnetic age jiashan (Xianshuihe) Fm. (Fig. 8), still belonging to the early Tunggu- constraints (Table 1 and Fig. 8). nian Stage of the early Middle Miocene in the CLMZ (Table 2). The Qijiachuan Fm. and Honggou Fm. follow the original division of The Diaogou (Gomphotherium) fauna has not been found in our the QOET (1977) as successions S13–S15 and S16–S23, respectively, measured sections; however, the original location where the fauna was but they are supplemented with additional successions S12–S10 (we use found and named is only about 2 km north of our measured XJ section italics to differentiate the successions from Cretaceous successions of (Fig. 1b and Suppl. Fig. 1). By lithologic correlation, it is roughly cor- the same number) at the bases of the EXN and Bingling Shan sections related with several thin sandstone layers in the middle Guanjiashan (Figs. 1, 4). Their paleomagnetic ages are 54–51.8 Ma and 51.8–43 Ma, (Xianshuihe) Fm., which have a paleomagnetic age of about 12.5 Ma respectively (Table 1 and Fig. 8). The Qijiachuan Fm. represents the (Fig. 8); thus the mammalian age agrees with the paleomagnetic dating, initial sedimentary infilling of the basin by alluvial-fluvial systems, characterized by channel sandstones and overbank floodplain

479 X. Fang et al. Earth-Science Reviews 190 (2019) 460–485

Fig. 11. Development of the Huang Shui terraces. Ages are synthesized from our new paleomagnetic results (see Figs. 8, 9) and from previous studies (see Zhang et al. (2017) for details). Gray rose diagrams are from Miao et al. (2008) and blue rose diagrams are from Zhang et al. (2017). (Modified from Zhang et al., 2017). TXD: Tuxiangdao; BSS: Beishanshi; PZS: Panzishan; GJS: Guanjiashan; HW: Houwan; CJB: Caijiabao; MJZ: Mojiazhuang. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.) mudstones and swamps (Fig. 4). The Honggou Fm. records a former structures. However, this sandstone bed occurs as a lenses of unstable fluvial system changing into a playa salt-lake (gypsum beds) and river channel deposit and has a large lateral variation. Instead, the mudflat (red mudstone beds) system (Fig. 4). excellent marker bed, the prominent closely-occurring three greenish The previous classifications of the Mahalagou Fm. divided the muddy-to-sandy gypsum beds (S39–S38) bearing the XJ fauna which is stratigraphy into two parts: the lower part, a visually striking marker just below S41, is quite stable and can be traced nearly basin-widely. It bed of gypsum that forms spectacular cliffs and Danxia-like landforms, meets the requirements of easy-to-identify and prominent division of and the upper part of brownish red massive mudstone intercalated with stratigraphy. We suggest the top of the marker bed, the closely-occur- several thin gypsum beds (Table 1 and Fig. 4, Suppl. Fig. 2). This has ring three greenish muddy-to-sandy gypsum beds (S39–S38), as the top resulted in the confusion of the XJ Fm. with the overlying mudstones of the Xiejia Fm. Thus, our division of the new XJ Fm. (S28–S39) has a and a common misidentification of the Mahalagou Fm. as the lower paleomagnetic age range of 34–24 Ma in the XJ section (Table 1 and gypsum unit. Furthermore, the abrupt lithologic change from the thick Fig. 8). gypsum beds to brownish-red mudstone marks a large and abrupt cli- The Chetougou Fm. (S40–46) resembles the XJ Fm. in terms of matic change in response to the initiation of the Antarctic ice sheet and mudstone units, but it is generally lighter in color than the XJ Fm., with a global temperature decrease at 34 Ma (Dupont-Nivet et al., 2007; many intercalated thin sandstone beds with cross- and massive-bed- Zhang et al., 2014; Fang et al., 2015). Thus, we suggest assigning dif- ding. In the SW-GJS section, sandstone beds occur more frequently and ferent names to the two lithologic units of the former Mahalagou Fm.: paleosol complexes occur at the top of the mudstone units. The paleo- i.e., the lower gypsum unit (S24–S27) is the new Mahalagou Fm., and magnetic age of this formation at 24–17 Ma (Table 1 and Figs. 4, 8). the upper mudstone unit (S28–S34) joins the Xiejia Fm.. The paleo- The Xianshuihe Fm. and Linxia Fm. were originally local formation magnetic age of the newly defined Mahalagou Fm. is 43–34 Ma(Table 1 names in the Lanzhou and Linxia Basins, respectively. The ages and and Figs. 4, 8). stratigraphic divisions have changed considerably over time – initially The XJ Fm. was originally designated S34–S40 by the QOET (1977) from early Miocene-to-Oligocene-middle Miocene, or even late Eocene- (Suppl. Fig. 2) which was later modified because of later revisions of its to-late Miocene for the Xianshuihe Fm. in the Lanzhou Basin; and from lower boundary (Table 1 and Fig. 7) - probably because of the absence initially Pliocene-to-late Oligocene-early Pliocene for the Linxia Fm., of visually distinctive marker beds between the upper mudstone part of and finally the abandonment of the name of the Linxia Fm. in theLinxia the former Mahalagou Fm. and differing interpretations of the original Basin) (GRGST, 1984; Qiu et al., 2001; Xie, 2004; Fang et al., 2003). classification because of the inaccessible original report ofthe QOET Consequently, we suggest abandoning the use of these two formation (1977) (e.g., Li et al., 1981; Dai et al., 2006; Xiao et al., 2012). We think names in the XN Basin to avoid misunderstandings and confusion, and that the original assignment of the base of the XJ Fm. to the base of S34 the use of the new names Guanjiashan Fm. and Mojiazhuang Fm. to (QOET, 1977) was not the best choice because the mudstone lithology replace the previous Xianshuihe and Linxia Fms., respectively, for the of S34 is the same as that of the underlying S33, and S34 is only slightly uppermost sequences of the Cenozoic stratigraphy in the XB. redder than S33 (Fig. 4 and Suppl. Fig. 2). Thus, we suggest to unify Thus, the Guanjiashan Fm. (S47–S66) is also similar in lithology to together the previous upper part (S28–S33) of the Mahalagou Fm. and the Chetougou Fm., but the mudstone and siltstone beds have lighter the Xiejia Fm. There is general agreement on the upper boundary of the colors (generally yellowish-brown) and there are more intercalated XJ Fm., because the overlying succession S41 (base of the Chetougou sandstone beds and paleosols in the XJ section than those of the Fm.) is a distinctive thick sandstone bed with many clear cross-bedding Chetougou Fm.; in addition, there is also the first appearance of several

480 X. Fang et al. Earth-Science Reviews 190 (2019) 460–485

Fig. 12. Synthesis of the stratigraphic sequence, lithology, sedimentary facies and environmental proxies (carbonate and sulfate contents, sediment redness and magnetic susceptibility), sedimentation rate, and paleocurrent changes in the Xining Basin (a-g). The suggested sedimentary environment and tectonic evolution are summarized in g and their comparisons with the global climatic record (Zachos et al., 2008) (h) is also presented. Note that sediment lightness is a close proxy of the carbonate content for the upper sequence of the stratigraphy. Redness and magnetic susceptibility are redrawn from Fang et al. (2015) and Zan et al. (2015). Stratigraphic legends are the same as Fig. 4. MJZ: Mojiazhuang section; XJ-TS: Xiejia-Tashan drilling section. fine conglomerate beds in the SW-GJS section and there are morethin, (54–34 Ma), indicate a progressive drying trend. This trend is indicated fine conglomerate beds in the MJZ section (Fig. 4). The combined pa- by the changes from floodplain-derived calcareous sandstones and leomagnetic dating of this formation in the XJ-Tashan drilling section, reddish mudstones-siltstones intercalated with numerous grayish- and the SW-GJS and MJZ sections, yields an age of ~17–7 Ma (Table 1 greenish marlite, several thin gypsum layers and blackish mudstones, and Fig. 8). representing the overbank deposits of shallow brackish lakes and The Mojiazhuang Fm. consists mainly of thick pebble-to-boulder swamps (indicating water with a significant dissolved carbonate con- conglomerate beds intercalated with several light yellowish-brown tent), to alternating thick gypsum beds (with mirabilite beds in the mudstone and siltstone beds and has a paleomagnetic age of 7.0–4.8 Ma middle) and reddish mudstone beds derived from a playa salt lake and (Table 1 and Figs. 4, 8). mudflats (Fig. 4). This trend is clearly confirmed by the higher carbo- We do not assign any specific names to the eolian sediments, butby nate content (mostly > 20–30%) at the base of the sequence of the XB convention we designate those older than the Quaternary as the ‘red sediments and by the higher calcium sulfate content (e.g., gypsum) in clay sequence’, and those in the Quaternary as the ‘loess-paleosol se- bulk samples as indicated by the high Ca/Al ratio (Fig. 12c, d). For this quence’ (Fig. 4). The red clay and loess-paleosol sequences in the XB part of the stratigraphic interval, the color proxy record also exhibits have much lighter colors (generally brownish-yellow to yellowish) and the highest (but progressively decreasing) redness over the entire are paleomagnetically dated to ~3.6–2.6 Ma and 2.6–0 Ma, respectively Cenozoic sequence in the XB (Fig. 12e); the magnetic susceptibility (Zhang et al., 2017)(Figs. 9, 11). (MS) also exhibits a similar change (Fig. 12f) (Fang et al., 2015). It seems global cooling reported in this interval may be expressed in these proxies due to a temperature or aridification change, with highest proxy 7.2. Comparison to climate context records vs. highest temperatures/moisture during the Early Eocene Climatic Optimum, followed by long-term proxy decreases with long The general coarsening- and lightening-upward trend of the term middle-late Eocene cooling (Fang et al., 2015)(Fig. 12). However, Cenozoic stratigraphy of the XB, as indicated by the stratigraphic di- it is important to note that the controlling factors for these paleoclimate vision described above, is more convincingly demonstrated by re- proxies are complicated and warrant further research. ference to the temporal domain of the lithofacies and its proxy records There is a change at the EOT as previously described. The new Xiejia of the XN Cenozoic sediments, coupled with global climate change and Fm. represents a stable playa mudflat with an ephemeral salt-lake and the history of tectonic uplift (Fig. 12). marks the abrupt termination of the thick gypsum beds of the The sedimentary lithofacies of the lower parts of the XN Cenozoic Mahalagou Fm. at 34 Ma, indicating that there was insufficient water to stratigraphy, from the Qijiachuan Fm. to the Mahalagou Fm.

481 X. Fang et al. Earth-Science Reviews 190 (2019) 460–485 sustain a playa lake system and thus further drying of the XB (Dupont- floodplain system in the lower Xiejia Fm. (34–27.5 Ma), accompanied Nivet et al., 2007; Zhang et al., 2014; Fang et al., 2015)(Figs. 4, 12). by a second episode of high sedimentation rates between 34 Ma and The hematite content, Magnetic Susceptibility, chemical weathering 30 Ma, probably indicate a second phase of episodic tectonic uplift of record, and pollen-spore and organic carbon isotope records, confirm the surrounding mountains and gradual deformation of the stratigraphy this sudden shift by exhibiting steep decreases followed subsequently in the proximity of the mountains in the early Oligocene (Fig. 12). We by very stable values (Long et al., 2011; Hoorn et al., 2012; Chi et al., noted that the large shift of the global temperature and paleoclimate 2013; Zhang et al., 2014; Fang et al., 2015). Thus, the lithofacies might contribute to the depositional environment shift at 34 Ma changes of the Qijiachuan Fm. to Xiejia Fm. in the XB record the long- (Dupont-Nivet et al., 2007)(Fig. 11). However, climatic researches term aridification history of the Asian interior within a sub-tropical hot indicate that the climate in Xining turned to dry at 34 Ma with global and dry climatic zone. In addition to the long-term global cooling trend cooling (Dupont-Nivet et al., 2007; Hoorn et al., 2012; Chi et al., 2013; since the Early Eocene Climatic Optimum, the stepwise retreat of the Zhang and Guo, 2014). This would reduce surface erosion and thus proto-Paratethys sea since about 41 Ma (Bosboom et al., 2011; Sun would not increase the sedimentation rate. Tectonic uplift would be et al., 2016), probably also acted as a major driving forces for this long- more plausible for the sedimentation rate increase. Actually, this early term climatic aridification (Dupont-Nivet et al., 2007; Bosboom et al., uplift of the NE Tibetan Plateau was also recorded in other areas parts 2011, 2014)(Fig. 12). of the Qilian Shan (Yin et al., 2002; Dai et al., 2005; Fang et al., 2013), The lithofacies of the middle parts of the XN Cenozoic stratigraphy and it was probably a direct response to the strong strike slip movement are represented by the upper Xiejia and Chetougou Fms. They are all of the Altyn Tagh fault at the western end of the Qilian Shan (Fig. 1) in distal floodplain deposits characterized by the occurrence of channel the Oligocene (e.g., Jolivet et al., 2001; Yin et al., 2002; Zhuang et al., thin sandstone beds and overbank siltstone-mudstone beds and paleo- 2011). sols, indicating an encroaching low energy river in a distal floodplain There is a widely-distributed unconformity between the Paleogene that was replacing the former playa system. The only differences be- and Neogene stratigraphy over the northeastern TP (GRGST, 1984; tween these two formations is the denser occurrence of sandstone beds, QBGM, 1991; Fang et al., 2003, 2005a; Yang et al., 2007; Zhuang et al., with increased thickness and cross-bedding, and the generally lighter 2011; Li et al., 2014). Although we do not observe it in the studied colors (reddish brown), which we interpret as reflecting an enhanced sections in the XB, there is a continuous thick cross-bedded sandstone water supply, hence channel energy, to the floodplain system (Figs. 4 layer, reflecting relatively high river channel energy, at the base ofS41 and 12). in the lower Chetougou Fm. (Figs. 4 and 12). This may be because the The lithofacies of the upper parts of the XN Cenozoic stratigraphy is studied sections are in the central XB. Actually, near the southern represented by the GJS Fm. (17–7 Ma) and the MJZ Fm. (7–4.8 Ma). margin of the Laji Shan to the south of the XB, there is an obvious They represent a rapid change from the previous distal floodplain to a angular unconformity, dated to about 22 Ma, between the Paleogene floodplain environment, and subsequently via braided rivers to anal- Xining Group and the Neogene Guide Group (QBGM, 1991; Parés et al., luvial fan. This is clearly indicated by the rapid increase in the number 2003; Fang et al., 2005a)(Fig. 1). This unconformity may also occur and thickness of sandstone and conglomerate beds, and by the size of near the northern margin of the Laji Shan or southern XB; however, it is the gravels (from fine gravels to pebbles and cobbles) in the conglom- impossible to confirm it, either because of subsequent strong tectonic erate beds (Figs. 4 and 12). The color and MS records also exhibit rapid uplift and erosion of the Neogene stratigraphy in some parts of the decreases, with large amplitude fluctuations, after about 15–14 Ma southern XB, or because of the absence of outcrops (Fig. 1). The rapid (Fig. 12)(Zan et al., 2015). exhumation of the nearby Laji Shan at around 25–20 Ma (Lease et al., The lithofacies of the top sequence of the Xining Cenozoic strati- 2007) recorded this rapid uplift. graphy is mainly represented by river terrace gravels and the overlying From the Guanjiashan Fm. (17–7 Ma) to the MJZ Fm. (7.0–4.8 Ma), eolian red clay sequence (3.6–2.6 Ma) and loess-paleosol sequence there is a rapid increase in grain size and occurrence of sandstone beds, (2.6–0 Ma). reflecting the change from a floodplain to a braided river atabout 8.8 Ma. This is followed by the rapid change to an alluvial fan at about 8. Stratigraphic constraints on the tectonic uplift of the 7 Ma, and finally by the termination of basin deposition at 4.8 Ma.This northeastern Tibetan Plateau rapid grain-size increase was accompanied by a two-to-three-fold in- crease in sedimentation rate, and in addition all measured paleocurrent The sequence of sedimentary environments from a playa-dominated directions in the central basin indicate a dominant southerly flow di- system, through floodplain, to braided river and alluvial fan, largely rection (Fig. 12). The gravel compositions are mainly metasandstone reflects the tectonic evolution of the basin, especially the uplift ofthe (70–90%), with small amounts of siliceous rocks, quartzite, slate and northeastern Tibetan Plateau (TP) (Fig. 12). schist; this suite represents a broadly stable gravel provenance that can The floodplain represented by the Qijiachuan Fm. (54–50 Ma) inthe be traced to Paleozoic (mostly Silurian) metasandstones and pre-Cam- Bingling Shan, and the EXN and XJ sections in the basin center, ac- brian siliceous rocks, quartzites, slates and schists in the eastern ex- companied by the first episode of high sedimentation rates, reflects the tension of the Qilian Shan, the Daban Shan, to the north (Fig. 1b). The initial response of the XB to the initial Cenozoic tectonic depression of particle size of the topmost conglomerates is generally from 30 to the basin or the uplift of the surrounding mountains, following the long- 80 mm, with a maximum of 300 mm, indicating strong flows from the term gradual rise and denudation of the northeastern TP in the late Daban Shan (Yang et al., 2017). Cretaceous (as indicated by the unconformity between the early The coarsening-upward of the sediments could also be caused by Cretaceous and the Qijiachuan Fm.) (Figs. 4, 8, 12). We interpret the climatic change; however, many climatic records from the NE Tibetan onset of the Cenozoic XB at about 54 Ma as a remote response to the Plateau demonstrate the long-term stepwise aridification since collision of the India and Asia at about 50 ± 5 Ma (Rowley, 1996; Zhu ~14–15 Ma (e.g., Li and Fang, 1999; Dettman et al., 2003; Ma et al., et al., 2005; Molnar and Stock, 2009) and locally to the exhumation of 2005; Fan et al., 2006; Li et al., 2014). Thus, we argue that tectonic the Kunlun Shan at ca 55–45 Ma (Clark et al., 2010). uplift was the major factor responsible for the general increasing trends The progressive increase in gypsum beds from the Honggou Fm. in sediment grain-size and sedimentation rate and the according change (51.8–43 Ma) to the Mahalagou Fm. (43–34 Ma), accompanied by the in facies. However, the pulse sedimentation rate increase at 17–14 Ma low sedimentation rates, suggests the continuous and gradual depres- was concordant with the Middle Miocene Climate Optimum (MMCO), sion of the XB and response to the ongoing collision of India and Asia when the climate in Xining and Asian inland was greatly ameliorated (Fig. 12). (Tang et al., 2011; Miao et al., 2012; Zan et al., 2015) and we see no The replacement of playa salt lakes by a playa mudflat–distal clear tectonic activity evidence. Thus we suppose the ameliorated

482 X. Fang et al. Earth-Science Reviews 190 (2019) 460–485

Xining climate might cause the increase of erosion and sediment supply, assistances in the field and laboratory. We thank the editor and thus increasing the sedimentation rate at the MMCO event. Recently, Professor Guillaume Dupont-Nivet and an anonymous reviewer for the sedimentological, thermochronological and tectonic evidence im- valuable comments that improved the manuscript very much. plies that intensified monsoon precipitation during the period 17–14 Ma may cause the establishment of the Mekong River in the Appendix A. Supplementary data southeastern Tibetan Plateau (Nie et al., 2018), due to increased ero- sion. This observation provides further support for our inference. Supplementary data to this article can be found online at https:// Therefore, all these lines of evidence and arguments indicate that doi.org/10.1016/j.earscirev.2019.01.021. the Daban Shan began to uplift acceleratedly since ~8–7 Ma, and rapid uplift from at least 4.8 Ma (Fig. 11). References In fact, numerous tectonic and sedimentary records suggest tectonic uplift accelerated at about 8 Ma across the NE Tibetan Plateau (Fang Abels, H.A., Dupont-Nivet, G., Xiao, G., Bosboom, R., Krijgsman, W., 2011. Step-wise et al., 2005a,b, 2007; Molnar, 2005; Zheng et al., 2006, 2017; Yang change of Asian interior climate preceding the Eocene–Oligocene transition (EOT). Palaeogeogr. Palaeoclimatol. Palaeoecol. 299, 399–412. et al., 2007; Li et al., 2014). Increased sediment grain-size, high sedi- An, Z.S., Kutzbach, J.E., Prell, W.L., Porter, S.C., 2001. Evolution of Asian monsoons and mentation rates and growth strata from ~8 Ma are widely reported phased uplift of the Himalaya-Tibetan Plateau since late Miocene times. Nature 411, across the entire NE Tibetan Plateau (Li et al., 2014), including in the 62–66. Bosboom, R.E., Dupont-Nivet, G., Houben, A.J.P., Brinkhuis, H., Vila, G., Mandic, O., Qaidam Basin (Fang et al., 2007), the Linxia and Guide Basins (Fang Stoica, M., Zachariasse, W.J., Guo, Z.J., Li, C.X., Krijgsman, W., 2011. Late Eocene et al., 2003, 2005a; Parés et al., 2003), and the Qilian Shan and Liupan Sea retreat from the Tarim Basin (West China) and concomitant Asian paleoenvir- Shan (Fang et al., 2005b; Zheng et al., 2006, 2017). onmental change. Palaeogeogr. Palaeoclimatol. Palaeoecol. 299, 385–398. Following the infilling of the XB, the entire region, including theXB Bosboom, R.E., Abels, H.A., Hoorn, C., Berg, B.C.J.V.D., Guo, Z.J., Dupont-Nivet, G., 2014. Aridification in continental Asia after the Middle Eocene Climatic Optimum and the surrounding mountains, was subjected to episodic continued (MECO). Earth Planet. Sci. Lett. 389, 34–42. rapid uplift, causing the successive incision of the basin sediments at Bowen, G.J., Clyde, W.C., Koch, P.L., et al., 2002. Mammalian dispersal at the Paleocene/ about 3.6 Ma, 2.6 Ma, 1.8 Ma, 1.2 Ma, 0.8 Ma, 0.6 Ma, and 0.12 Ma, as Eocene boundary. Science 295, 2062–2065. Bowen, G.J., Koch, P.L., Meng, J., et al., 2005. Age and correlation of fossiliferous late indicated by the development of a series of terraces of the Huang Shui, Paleocene-early Eocene strata of the Erlian Basin, Inner Mongolia. China. Am. Mus. where long term climate cooling and drying were argued for less impact Novit. 3474, 1–26. (Zhang et al., 2017)(Figs. 9, 11). This was accompanied by a shift from Chen, G.F., 1988. Remarks on the Oioceros species (Bovidae, Artiodactyla, Mammalia) from the Neogene of China. Vertebrata PalAsiatica 26, 157–172. the previous southerly-dominated paleocurrent direction to an east to Chi, Y.P., Fang, X.M., Song, C.H., Miao, Y.F., Teng, X.H., Han, W.X., Wu, F.L., Yang, J.W., west direction, together with the deposition of thick dust deposits 2013. Cenozoic organic carbon isotope and pollen records from the Xining Basin, NE which capped the terraces from about 3.6 Ma (Zhang et al., 2017; Yang Tibetan Plateau, and their palaeoenvironmental significance. Palaeogeogr. Palaeoclimatol. Palaeoecol. 386, 436–444. et al., 2017)(Figs. 11, 12). Clark, M.K., Farley, K.A., Zheng, D., Wang, Z.C., Duvall, A., 2010. Early Cenozoic faulting on the northern Tibetan Plateau margin from apatite (U-Th)/He ages. Earth Planet. 9. Conclusions Sci. Lett. 296, 78–88. Dai, S., Fang, X.M., Song, C.H., Gao, J.P., Gao, D.L., Li, J.J., 2005. Early tectonic uplift of the northern Tibetan Plateau. Chin. Sci. Bull. 50 (15), 1642–1652. 1) Our compilation of pre-existing and new paleomagnetic data for the Dai, S., Fang, X.M., Dupont-Nivet, G., Song, C.H., Gao, J.P., Krijgsman, W., Langereis, C., Cenozoic sedimentary sequence of the Xining Basin indicates that Zhang, W.L., 2006. Magnetostratigraphy of Cenozoic sediments from the Xining that there has been sediment deposition from ~54 Ma to 4.8 Ma, Basin: tectonic implications for the northeastern Tibetan Plateau. J. Geophys. Res. 111 (B111), 335–360. followed by pure eolian dust deposition since ~3.6 Ma. Deng, T., Wang, W.M., Yue, L.P., 2006. The Xiejian stage of the continental Miocene 2) Systematic cross-checking of the paleomagnetic data confirms that series in China. J. Stratigr. 30 (4), 315–322 (in Chinese with English abstract). the age of the Xiejia fauna is ~25 Ma (i.e., in the late Oligocene). Dettman, D.L., Fang, X.M., Garzione, C.N., Li, J.J., 2003. Uplift-driven climate change at 12Ma: a long δ18O record from the NE margin of the Tibetan Plateau. Earth Planet. This contradicts and thus suggests to abandon the previously-as- Sci. Lett. 214 (1–2), 267–277. sumed age of ~21 Ma for the Xiejian Stage of the standard early Ding, Z.L., Yu, Z.W., Rutter, N.W., Liu, T.S., 1994. Towards an orbital time scale for Miocene zone of the Chinese Land Mammal Zones. Chinese loess deposits. Quat. Sci. Rev. 13 (1), 39–70. Dupont-Nivet, G., Horton, B.K., Butler, R.F., Wang, J., Zhou, J., Waanders, G.L., 2004. 3) The timing of lithofacies changes suggests that initial uplift of the Paleogene clockwise tectonic rotation of the Xining-Lanzhou region, northeastern NE Tibetan Plateau, and depression of the Xining Basin, began at Tibetan Plateau. J. Geophys. Res. 109, B04401. https://doi.org/10.1029/ ~54 Ma, and that early episodic uplift and later enhanced uplift 2003JB002620. Dupont-Nivet, G., Krijgsman, W., Langereis, C.G., Abels, H.A., Dai, S., Fang, X.M., 2007. occurred in the Oligocene and early Miocene, respectively. From the Tibetan Plateau aridification linked to global cooling at the Eocene-Oligocene tran- Late Miocene, the NE Tibetan Plateau experienced a late phase of sition. Nature 445, 635–638. long-term stepwise, rapid uplifts since ~8–7 Ma and < 4.8–3.6 Ma, Dupont-Nivet, G., Dai, S., Fang, X., Krijgsman, W., Erens, V., Reitsma, M., Langereis, C., 2008a. Timing and distribution of tectonic rotations in the northeastern Tibetan which terminated basin infilling and resulted in the incision ofthe Plateau. Geol. Soc. Am. Spec. Pap. 444, 73–87. basin sediments and forming up to 16 terraces of the Huang Shui Dupont-Nivet, G., Hoorn, C., Konert, M., 2008b. Tibetan uplift prior to the Eocene- River. This tectonic uplift, basin formation and incision in the NE Oligocene climate transition: evidence from pollen analysis of the Xining Basin. Tibetan Plateau were major nearly synchronous responses to the Geology 37, 987–990. Fan, M.J., Song, C.H., Dettman, D.L., Fang, X.M., Xu, X.H., 2006. Intensification of the initial and continuing collision of India and Asia since ~50 ± 5 Ma. Asian winter monsoon after 7.4 Ma: grain-size evidence from the Linxia Basin, northeastern Tibetan Plateau, 13.1 Ma to 4.3 Ma. Earth Planet. Sci. Lett. 248 (1), Acknowledgements 171–182. Fang, X.M., Garzione, C., Van der Voo, R., Li, J.J., Fan, M.J., 2003. Flexural subsidence by 29 Ma on the NE edge of Tibet from the magnetostratigraphy of Linxia Basin, China. This work is co-supported by the National Natural Science Earth Planet. Sci. Lett. 210 (3–4), 545–560. Foundation of China (41620104002) and the Strategic Priority Fang, X.M., Yan, M.D., Van der Voo, R., Rea, D.R., Song, C.H., Pare's, J.M., Gao, J.P., Nie, J.S., Dai, S., 2005a. Late Cenozoic deformation and uplift of the NE Tibetan Plateau: Research Program of Chinese Academy of Sciences (XDA20070201). evidence from high resolution magnetostratigraphy of the Guide Basin, Qinghai We expressed our cordial thanks to Professors Tao Deng, Zhanxiang Province, China. Geol. Soc. Am. Bull. 107, 1208–1225. Qiu, Zhuding Qiu and Banyue Wang for their valuable discussions and Fang, X.M., Zhao, Z.J., Li, J.J., Yan, M.D., Pan, B.T., Song, C.H., Dai, S., 2005b. Magnetostratigraphy of the late Cenozoic Laojunmiao anticline in the northern Qilian comments to improve the early draft of the paper. We also thank Drs. Mountains and its implications for the northern Tibetan Plateau uplift. Science in Han Wenxia, Li Minghui, Li Jiao, Wang Jiuyi, Ying Hong, Bao Jing, He China (series D) 48 (7), 1040–1051. Pengju, Chen Shu, Ma Lifang, Chen Chihao, Hu Chunhua, Zhong Sirui, Fang, X.M., Zhang, W.L., Meng, Q.Q., Gao, J.J., Wang, X.M., King, J., Song, C.H., Dai, S., Miao, Y.F., 2007. High-resolution magnetostratigraphy of the Neogene Huaitoutala Guo Haichao, Yang Rongsheng, Ruan, Xiaobai, Xie Yulong, Lü Shuang, section in the eastern Qaidam Basin on the NE Tibetan Plateau, Qinghai Province, Mao Ziqiang, Lu Shengcheng, Fang Xiaohui and Yang Liye for their

483 X. Fang et al. Earth-Science Reviews 190 (2019) 460–485

China and its implication on tectonic uplift of the NE Tibetan Plateau. Earth Planet. Xining Basin. Qinghai Geology 2, 1–18 (in Chinese with English abstract). Sci. Lett. 258, 293–306. Liu, T.S., Zheng, M.P., Guo, Z.T., 1998. Initiation and evolution of the Asian monsoon Fang, X.M., Liu, D.L., Song, C.H., Dai, S., Meng, Q.Q., 2013. Oligocene slow and Miocene- system timely coupled with the ice-sheet growth and the tectonic movements in Asia. Quaternary rapid deformation and uplift of the Yumu Shan and North Qilian Shan: Quaternary Sciences 3, 194–204 (in Chinese with English abstract). evidence from high-resolution magnetostratigraphy and tectonosedimentology. Geol. Long, L.Q., Fang, X.M., Miao, Y.F., Bai, Y., Wang, Y.L., 2011. Northern Tibetan Plateau Soc. Lond. Spec. Publ. 373, 149–171. cooling and aridification linked to Cenozoic global cooling: evidence from n-alkane Fang, X.M., Zan, J.B., Appel, E., Lu, Y., Song, C.H., Dai, S., Tuo, S.B., 2015. An Eocene- distributions of Paleogene sedimentary sequences in the Xining Basin. Chin. Sci. Bull. Miocene continuous rock magnetic record from the sediments in the Xining Basin, 56 (15), 1569–1578. NW China: indication for Cenozoic persistent drying driven by global cooling and Lourens, L.J., Hilgen, F.J., Laskar, J., Shackleton, N.J., Wilson, D.S., 2004. The Neogene Tibetan Plateau uplift. Geophys. J. Int. 201 (1), 78–89. period. In: Gradstein, F.M., Ogg, J.G., Smith, A.G. (Eds.), A Geologic Time Scale. Fang, X.M., Wang, J.Y., Zhang, W.L., Zan, J.B., Song, C.H., Yan, M.D., Appel, E., Zhang, Cambridge University Press, Cambridge, pp. 409–440. T., Wu, F.L., Yang, Y.B., Lu, Y., 2016. Tectonosedimentary evolution model of an Lu, H.Y., Wang, X.Y., An, Z.S., Miao, X.D., Zhu, R.X., Ma, H.Z., Li, Z., Tan, H.B., Wang, intracontinental flexural (foreland) basin for paleoclimatic research. Glob. Planet. X.Y., 2004. Geomorphologic evidence of phased uplift of the northeastern Chang. 145, 78–97. QinghaiTibetan Plateau since 14 million years. Science in China (series D) 47 (9), FAO, 1988. FAO-UNESCO Soil Map of the World. Revised Legend. World Resources 822–833. Report 60. FAO, Rome. Lu, H.Y., Wang, X.Y., Sun, X.F., Wang, X.Y., Yi, S.W., Zhou, Y.L., Liu, Q.Y., 2007. Loess Feng, Y.M., Wu, H.Q., 1992. Tectonic evolution of North Qilian Mountains and its stratigraphy and palaeoclimate changes during Quaternary in northeastern Tibetan neighbourhood since Paleozoic. Northwest Geosci. 13, 61–74 (in Chinese with Plateau revealed by loess core. Quaternary Sciences 27 (2), 230–241 (in Chinese with English abstract). English abstract). Gehrels, G.E., Yin, A., Wang, X.F., 2003. Detrital-zircon geochronology of the north- Ma, Y.Z., Fang, X.M., Li, J.J., Wu, F.L., Zhang, J., 2005. The vegetation and climate eastern Tibetan plateau. GSA Bull. 115, 881–896. change during Neocene and early Quaternary in Jiuxi Basin, China. Science in China Gradstein, F.M., Ogg, J.G., Schmitz, M.D., Ogg, G.M. (Eds.), 2012. The Geologic Time (series D) 48 (5), 676–688. Scale. Elsevier, Amsterdam, pp. 923–978. McFadden, P.L., McElhinny, M.W., 1990. Classification of the reversal test in palaeo- GRGST (Gansu Regional Geological Survey Team), 1984. Tertiary system of Gansu. Gansu magnetism. Geophys. J. Int. 103 (3), 725–729. Geology 2, 1–40. Miao, X., Lu, H., Li, Z., Cao, G., 2008. Paleocurrent and fabric analyses of the imbricated Guo, Z.T., Sun, B., Zhang, Z.S., Peng, S.Z., Xiao, G.Q., Ge, J.Y., Hao, Q.Z., Qiao, Y.S., fluvial gravel deposits in Huangshui valley, the northeastern Tibetan Plateau, China. Liang, M.Y., Liu, J.F., Yin, Q.Z., Wei, J.J., 2008. A major reorganization of Asian Geomorphology 99, 433–442. climate by the early Miocene. Clim. Past 43, 153–174. Miao, Y.F., Herrmann, M., Wu, F.L., Yan, X.L., Yang, S.L., 2012. What controlled Mid–late Han, W.X., Fang, X.M., Berger, A., Yin, Q.Z., 2011. An astronomically tuned 8.1 Ma eolian Miocene long-term aridification in Central Asia? — Global cooling or Tibetan Plateau record from the Chinese Loess Plateau and its implication on the evolution of Asian uplift: a review. Earth Sci. Rev. 112, 155–172. monsoon. J. Geophys. Res. Atmos. 116 D24114. Molnar, P., 2005. Mio-Pliocene growth of the Tibetan Plateau and evolution of East Asian Han, J.E., Shao, Z.G., Chen, Q.G., Meng, X.G., Zhu, D.G., Zhang, Q.Q., Quan, K., Wang, J., climate. Palaeontol. Electron. 8 (1), 1–23. Xu, B., 2015. Discovery and significance of Hipparion and Parelasmotherium fossils Molnar, P., Stock, J.M., 2009. Slowing of India's convergence with Eurasia since 20 Ma in the Xining Basin, Qinghai Province. Acta Geosci. Sin. 36 (1), 115–120 (in Chinses and its implications for Tibetan mantle dynamics. Tectonics 28 (3) TC3001. with English abstract). Nie, J., Stevens, T., Rittner, M., Stöckli, D.F., Garzanti, E., Limonata, M., Bird, A., Andò, Hoorn, C., Straathof, J., Abels, H.A., Xu, Y.D., Utescher, T., Dupont- Nivet, G., 2012. A S., Vermeesch, P., Saylor, J., Lu, H., Breecker, D., Hu, X., Liu, S., Resentini, A., late Eocene palynological record of climate change and Tibetan Plateau uplift (Xining Vezzoli, G., Peng, W., Carter, A., Ji, S., Pan, B., 2015. Loess Plateau storage of Basin, China). Palaeogeogr. Palaeoclimatol. Palaeoecol. 344–345 (8), 16–38. Northeastern Tibetan Plateau derived Yellow River sediment. Nat. Commun. 6, 8511. Horton, B.K., Dupont-Nivet, G., Zhou, J., Waanders, G.L., Butler, R.F., Wang, J., 2004. Nie, J., Garzione, C., Su, Q., Liu, Q., Zhang, R., Heslop, D., Necula, C., Zhang, S., Song, Y., Mesozoic-Cenozoic evolution of the Xining-Minhe and Dangchang basins, north- Luo, Z., 2017. Dominant 100,000-year precipitation cyclicity in a late Miocene lake eastern Tibetan Plateau: magnetostratigraphic and biostratigraphic results. J. from Northeast Tibet. Sci. Adv. 3 (3), e1600762. Geophys. Res. 109 (B4), 657–681. Nie, J.S., Ruetenik, G., Gallagher, K., Hoke, G., Garzione, C.N., Wang, W.T., Stockli, D., Jolivet, M., Brunel, M., Seward, D., Xu, Z., Yang, J., Roger, F., Tapponnier, P., Malavieille, Hu, X.F., Wang, Z., Wang, Y., Stevens, T., Danišík, M., Liu, S.P., 2018. Rapid incision J., Arnaud, N., Wu, C., 2001. Mesozoic and Cenozoic tectonics of the northern edge of of the Mekong River in the middle Miocene linked to monsoonal precipitation. Nat. the Tibetan Plateau: fission-track constraints. Tectonophysics 343 (1–2), 111–134. Geosci. 11 (12), 944–948. Lallier, F., Antoine, C., Charreau, J., Caumon, G., Ruiu, J., 2013. Management of ambi- Parés, J.M., Van der Voo, R., Downs, W.R., Yan, M.D., Fang, X.M., 2003. Northeastward guities in magnetostratigraphic correlation. Earth Planet. Sci. Lett. 371-372, 26–36. growth and uplift of the Tibetan Plateau: magnetostratigraphic insights from the Lease, R.O., Burbank, D.W., Gehrels, G.E., Wang, Z.C., Yuan, D.Y., 2007. Signatures of Guide Basin. J. Geophys. Res. Solid Earth 108 (B1). mountain building: detrital zircon U/Pb ages from northeastern Tibet. Geology 35, QBGM (Qinghai Bureau of Geology and Mineral Resources), 1991. Geologic History of the 239–242. Qinghai Province. Geological Publishing House, Beijing. Lease, R.O., Burbank, D.W., Clark, M.K., Farley, K.A., Zheng, D.W., Zhang, H.P., 2011. Qiu, Z.X., Qiu, Z.D., 1990. Neogene local mammalian faunas: succession and ages. J. Middle Miocene reorganization of deformation along the northeastern Tibetan Stratigr. 14 (4), 241–260 (in Chinese with English abstract). Plateau. Geology 39, 359–362. Qiu, Z.D., Li, C.K., Wang, S.J., Downs, W., 1981. Miocene mammals fossils from the Lease, R.O., Burbank, D.W., Hough, B., Wang, Z.C., Yuan, D.Y., 2012. Pulsed Miocene Xining Basin, Qinghai Province. Vertebrata PalAsiatica 19 (2), 156–173. range growth in northeastern Tibet: insights from Xunhua Basin magnetostratigraphy Qiu, Z.X., Wang, B.Y., Qiu, Z.D., Heller, F., Yue, L.P., Xie, G.P., Wang, X.M., Engesser, B., and provenance. Geol. Soc. Am. Bull. 124, 657–677. 2001. Land mammal geochronology and magnetostratigraphy of mid-Tertiary de- Li, J.J., Fang, X.M., 1999. Uplift of the Tibetan Plateau and environmental changes. Chin. posits in the Lanzhou Basin, Gansu Province, China. Switzerland Basel: Eclogae Sci. Bull. 44, 2117–2124. Geologicae Helvetiae 94 (3), 373–385. Li, C.K., Qiu, Z.D., 1980. Early Miocene mammalian fossils of Xining Basin, Qinghai. QOET (Qinghai Oil Exploration Team), 1977. Report on 1:100,000 Petroleum Geology Vertebrata PalAsiatica 18 (3), 198–214 (In Chinese with English abstract). Survey of Xining Basin. Qinghai Oil Exploration Team, Xining (in Chinese). Li, C.K., Qiu, Z.D., Wang, S.J., 1981. Discussion on Miocene stratigraphy and mammals Reading, H.G., 2009. Sedimentary Environments: Processes, Facies and Stratigraphy, 3rd. from Xining basin, Qinghai. Vertebrata PalAsiatica 19 (4), 313–320 (In Chinese with Blackwell Science, Oxford, pp. 704 Wiley-Blackwell. English abstract). Rowley, D.B., 1996. Age of initiation of collision between India and Asia: a review of Li, C.K., Wu, W.Y., Qiu, Z.D., 1984. Chinese Neogene: subdivision and correlation. stratigraphic data. Earth Planet. Sci. Lett. 145 (1–4), 1–13. Vertebrata PalAsiatica 22 (3), 163–179 (In Chinese with English abstract). Steininger, F.F., 1999. Chronostratigraphy, geochronology and biochronology of the Li, J.J., Fang, X.M., Ma, H.Z., Zhu, J.J., Pan, B.T., Chen, H.L., 1996a. Geomorphological Miocene "European Land Mammal Mega-Zones" (ELMMZ) and the Miocene and environmental evolution in the upper reaches of the Yellow River during the late "Mammal-Zones (MN-Zones)". The Miocene Land Mammals of Europe 9–24. Cenozoic. Science in China (series D) 39 (4), 380–390. Sun, X.J., Wang, P.X., 2005. How old is the Asian monsoon system?—Palaeobotanical Li, Z., Li, L.Q., Zeng, Y.N., Ma, H.Z., 1996b. The studies of Xining loess. Qinghai Geol. records from China. Palaeogeogr. Palaeoclimatol. Palaeoecol. 222 (3), 181–222. 1–10 (in Chinese with English abstract). Sun, J.M., Widley, B.F., Zhang, Z.L., Fu, B.H., Li, S.H., 2016. Diachronous seawater retreat Li, J.J., Fang, X.M., Song, C.H., Pan, B.T., Ma, Y.Z., Yan, M.D., 2014. Late Miocene- from the southwestern margin of the Tarim Basin in the late Eocene. J. Asian Earth Quaternary rapid stepwise uplift of the NE Tibetan Plateau and its effects on climatic Sci. 116, 222–231. and environmental changes. Quat. Res. 81 (3), 400–423. Tang, Z., Ding, Z., White, P.D., Dong, X., Ji, J., Jiang, H., Luo, P., Wang, X., 2011. Late Li, B.S., Yan, M.D., Zhang, W.L., Fang, X.M., Meng, Q.Q., Zan, J.B., Chen, Y., Zhang, D.W., Cenozoic central Asian drying inferred from a palynological record from the northern Yang, Y.P., Guan, C., 2017. New paleomagnetic constraints on middle Miocene strike- Tian Shan. Earth Planet. Sci. Lett. 302, 439–447. slip faulting along the middle Altyn Tagh Fault. J. Geophys. Res. 122, 4106–4122. Tapponnier, P., Xu, Z.Q., Roger, F., Meyer, B., Arnaud, N., Wittlinger, G., Yang, J.S., 2001. Licht, A.M., Cappelle, M.V., Abels, H.A., Ladant, J.B., Alexander, T., Lanord, C.F., Geology-oblique stepwise rise and growth of the Tibet Plateau. Science 294 (5547), Donnadieu, Y., Vandenberghe, J., Rigaudier, T., Lécuyer, C., Adriaens, R., Guo, Z., 1671–1677. Soe, A.N., Quade, J., 2014. Asian monsoons in a late Eocene greenhouse world. Tauxe, L., 1998. Paleomagnetic Principles and Practice. Kluwer Academic Publishers, Nature 513 (7519), 501–506. Dordrecht, pp. 299. Licht, A., Dupont-Nivet, G., Pullen, A., Kapp, P., Abels, H.A., Lai, Z., Guo, Z., Abell, J., Wang, B.Y., 1997. The mid-Tertiary Ctenodactylidae (Rodentia, Mammalia) of Eastern Giesler, D., 2016. Resilience of the Asian atmospheric circulation shown by Paleogene and Central Asia. Bull. Am. Mus. Nat. Hist. 234, 1–88. dust provenance. Nat. Commun. 7, 12390. https://doi.org/10.1038/ncomms12390. Wang, E., Shi, X.H., Wan, G., Fan, C., 2011. Structural control on the topography of the Lin, A., Yang, Z., Sun, Z., Yang, T., 2001. How and when did the Yellow River develop its Laji–Jishi and Riyue Shan belts in the NE margin of the Tibetan Plateau: facilitation square bend? Geology 29, 951–954. of the headward propagation of the Yellow River system. J. Asian Earth Sci. 40, Liu, M.R., 1992. Stratigraphic sequence and fossil assemblage of Neogene system in 1002–1014.

484 X. Fang et al. Earth-Science Reviews 190 (2019) 460–485

Wang, X.M., Flynn, L.J., Fortelius, M. (Eds.), 2013. Fossil Mammals of Asia: Neogene greenhouse warming and carbon-cycle dynamics. Nature 451, 279–283. Biostratigraphy and Chronology. Columbia University Press, New York 51, 84, Zan, J.B., Fang, X.M., Yan, M.D., Zhang, W.L., Lu, Y., 2015. Lithologic and rock magnetic 124–141. evidence for the Mid-Miocene climatic optimum recorded in the sedimentary archive Wu, L.C., Yue, L.P., Wang, J.Q., Heller, F., Deng, T., 2006. Magnetostratigraphy of stra- of the Xining Basin, NE Tibetan Plateau. Palaeogeogr. Palaeoclimatol. Palaeoecol. totype section of the Neogene Xiejian stage. J. Stratigr. 30 (1), 50–53 (in Chinese with 431, 6–14. English abstract). Zeng, Y.N., Ma, H.Z., Li, Z., Li, L.Q., 1995. A study on terrace formation and development Wu, L., Xiao, A.C., Yang, S.F., Wang, L.Q., Mao, L.G., Wang, L., Dong, Y.P., Xu, B., 2012. of the Huangshui River in Xining area. Sci. Geogr. Sin. 15 (3), 253–258 (in Chinese Two-stage evolution of the Altyn Tagh fault during the Cenozoic: new insight from with English abstract). provenance analysis of a geological section in NW Qaidam Basin, NW China. Terra Zhai, Y., Cai, T., 1984. The Tertiary system of Gansu Province, in Gansu geology. People's Nova 24 (5), 387–395. Press of Gansu, Lanzhou, pp. 1–40 (in Chinese). Xiao, G.Q., Abels, H.A., Yao, Z.Q., Dupont-Nivet, G., Hilgen, F.J., 2010. Asian aridifica- Zhang, C.X., Guo, Z.T., 2014. Clay mineral changes across the Eocene-Oligocene transi- tion linked to the first step of the Eocene-Oligocene climate transition (EOT) inob- tion in the sedimentary sequence at Xining occurred prior to global cooling. liquity dominated terrestrial records (Xining Basin, China). Clim. Past 6, 501–513. Palaeogeogr. Palaeoclimatol. Palaeoecol. 411, 18–29. Xiao, G.Q., Guo, Z.T., Dupont-Nivet, G., Lu, H.Y., Wu, N.Q., Ge, J.Y., Hao, Q.Z., Peng, S.Z., Zhang, J., Wang, Y., Zhang, B.H., Zhao, H., 2015. Exhumation of marginal mountain Li, F.J., Abels, H.A., Zhang, K.X., 2012. Evidence for northeastern Tibetan Plateau ranges of the Xining Basin in NW China and its implication for the evolution of the NE uplift between 25 and 20 Ma in the sedimentary archive of the Xining Basin, Qinghai–Tibetan Plateau during the Cenozoic. J. Asian Earth Sci. 97, 10–23. Northwestern China. Earth Planet. Sci. Lett. 317–318 (2), 185–195. Zhang, J., Wang, Y., Zhang, B.H., Zhang, Y., 2016. Tectonics of the Xining Basin in NW Xie, G.P., 2004. The Tertiary and local mammalian faunas in Lanzhou Basin, Gansu. J. China and its implications for the evolution of the NE Qinghai-Tibetan Plateau. Basin Stratigr. 28 (1), 67–80 (in Chinese with English abstract). Res. 28, 159–182. Yan, M.D., Van der Voo, R., Fang, X.M., Parés, J.M., Rea, D.K., 2006. Paleomagnetic Zhang, W.L., Zhang, T., Song, C.H., Appel, E., Mao, Z., Fang, Y., Lu, Y., Meng, Q., Yang, evidence for a mid - Miocene clockwise rotation of about 25° of the Guide Basin area R., Zhang, D., Li, B., Li, J., 2017. Termination of fluvial-alluvial sedimentation in the in NE Tibet. Earth Plant. Sci. Lett. 241, 234–247. Xining Basin, NE Tibetan Plateau, and its subsequent geomorphic evolution Plio- Yang, S.F., Chen, H.L., Xiao, A.C., Chen, J.J., Fan, M.T., Tian, D.W., 2007. The Structural Quaternary terrace development in the Xining Basin, NE Tibetan Plateau. Characteristics of Northern Qilian Mountain Thrust Belt and its Control on the Oil and Geomorphology 297, 86–99. Gas Accumulation. Chinese Science Press, Beijing, pp. 159 (in Chinese with English Zheng, D.W., Zhang, P.Z., Wan, J.L., Yuan, D.Y., Li, C.Y., Yin, G.M., Zhang, G.L., Wang, abstract). Z.C., Min, W., Chen, J., 2006. Rapid exhumation at~ 8 Ma on the Liupan Shan thrust Yang, R.S., Fang, X.M., Meng, Q.Q., Zan, J.B., Zhang, W.L., Deng, T., Yang, Y.B., Ruan, fault from apatite fission-track thermochronology: implications for growth of the X.B., Yang, L.Y., Li, B.S., 2017. Paleomagnetic constraints on the middle Miocene- northeastern Tibetan Plateau margin. Earth Planet. Sci. Lett. 248 (1), 198–208. early Pliocene stratigraphy in the Xining Basin, NE Tibetan Plateau, and the geologic Zheng, H.B., Clift, P.D., Wang, P., Tada, R., Jia, J.T., He, M.Y., Jourdan, F., 2013. Pre- implications. Geochem. Geophys. Geosyst. 18, 3741–3757. Miocene birth of the Yangtze River. Proc. Natl. Acad. Sci. U. S. A. 110 (19), Yin, A., Harrison, T.M., 2000. Geologic evolution of the Himalayan–Tibetan orogen. 7556–7561. Annu. Rev. Earth Planet. Sci. 28, 211–280. Zheng, D.W., Wang, W.T., Wan, J.L., Yuan, D., Liu, C.R., Zheng, W.J., Zhang, H.P., Pang, Yin, A., Rumelhart, P.E., et al., 2002. Tectonic history of the Altyn Tagh fault system in J.Z., Zhang, P.Z., 2017. Progressive northward growth of the Northern Qilian Shan- northern Tibet inferred from Cenozoic sedimentation. GSA Bull. 114, 1257–1295. Hexi Corridor (northeastern Tibet) during the Cenozoic. Lithosphere 9 (3), 408–416. Young, C.C., Bien, M.N., 1937. Cenozoic geology of the Kaolan-Yungteng area of central Zhu, B., Kidd, W.S.F., Rowley, D.B., Currie, B.S., Shafique, N., 2005. Age of initiation of Kansu. Bull. Geol. Soc. China 16, 221–269. the Indian-Asia collision in the East-Central Himalaya. J. Geol 113 (3), 265–285. Yuan, D.Y., Ge, W.P., Chen, Z.W., et al., 2013. The growth of northeastern Tibet and its Zhuang, G.S., Hourigan, J.K., Ritts, B.D., Kent-Corson, M.L., 2011. Cenozoic multiple- relevance to large-scale continental geodynamics: a review of recent studies. phase tectonic evolution of the Northern Tibetan Plateau: constraints from sedi- Tectonics 32, 1–13. https://doi.org/10.1002/tect.20081. mentary records from Qaidam Basin, Hexi Corridor, and Subei Basin, Northwest Zachos, J.C., Dickens, G.R., Zeebe, R.E., 2008. An early cenozoic perspective on China. Am. J. Sci. 311 (2), 116–152.

485