Magnetostratigraphic constraints on the fossiliferous Ulantatal sequence in Inner Mongolia, China: Implications for Asian aridification and faunal turnover before the Eocene-Oligocene boundary

a Department of Geosciences and Geography, University of Helsinki, P.O. Box 64, Helsinki 00014, Finland b Department of Physics, University of Helsinki, P.O. Box 64, Helsinki 00014, Finland c Key Laboratory of Vertebrate Evolution and Human Origin of the Chinese Academy of Sciences, Institute of Vertebrate Paleontology and Paleoanthropology, Chinese Academy of Sciences, Beijing 100044, China d CAS Center for Excellence in Life and Paleoenvironment, Beijing, China e University of Chinese Academy of Sciences, Beijing, China


Introduction
The onset of aridification in the continental realm has been increasingly documented in terrestrial sequences of interior Asia. Important drivers for the development of aridity are attributed to the combined effects of global cooling, the Tibetan uplift and retreat of the Paratethys Sea, subsequent decrease of moisture and eustatic sea level changes (e.g. Dupont-Nivet et al., 2007;Guo et al., 2008;Bosboom et al., 2014a;Wang et al., 2016;Bougeois et al., 2018;Caves Rugenstein and Chamberlain, 2018;Meijer et al., 2019). The stepwise aridification (Meijer et al., 2019) culminates in the transition from Eocene to Oligocene (EOT) ∼34 Ma which is, in general, considered to signify major climatic and environmental changes on a global scale (e.g., Prothero and Heaton, 1996;Dupont-Nivet et al., 2007;Hren et al., 2013;Sun et al., 2014).
The current understanding is that the faunal changes associated with the EOT had different forms in Europe (Hooker et al., 2004), North America (Prothero and Heaton, 1996) and inland Asia (Meng and McKenna, 1998). In the inlands of continental Asia, the transition, also known as the "Mongolian Remodeling", was outlined by replacement of Eocene medium-sized perissodactyl-dominant faunas with the Oligocene rodent-lagomorph-dominant faunas (Meng and McKenna, 1998;Kraatz and Geisler, 2010). Considerable attention has been paid to establish the relationships between the climatic and faunal changes at the Eocene-Oligocene transition in inland Asia: Meng and McKenna (1998) and Sun et al. (2014) have stimulated discussion of the idea of nearly synchronous turnover across the epoch boundary associated with change in the physical environment. The scarcity of chronostratigraphic control restricts the insight in the timing and magnitude of climatic and faunal events and thereby hinders the identification of their causes and effects. In addition, the geographical coverage corroborating the environmental impacts remains scarce and the stratigraphic frameworks are often rudimentary. Furthermore, recent results suggest largely heterogeneous response in vegetation across the EOT globally (Pound and Salzmann, 2017), but implying biomes responding to increased aridity in Central to East Asia already since early Eocene, without any major shifts across the boundary from the Eocene to the Oligocene.
In order to overcome this inconsistency we conducted integrated litho-, bio-and magnetostratigraphic investigations in the Ulantatal area. Ulantatal provides an optimal location to examine the timing of the "Mongolian Remodeling" and its association with the EOT because of the area's largely undeformed sedimentary succession and diverse fossil fauna in stratigraphic superposition (e.g. Vianey-Liaud et al., 2006;Gomes Rodrigues et al., 2014;Zhang et al., 2016). This combination provides a unique opportunity to study the evolution of terrestrial environments and faunas during a critical period of the Cenozoic era. Despite the abundant and diversified Oligocene mammalian fossil assemblages (e.g. Zhang et al., 2016 and references therein), proper stratigraphic and temporal correlation, understanding of the depositional environment and biogeographic relationships with the Mongolian plateau, has remained unsettled. In this paper, we present a comprehensive lithoand magnetostratigraphic framework for the Ulantatal deposits. These data provide age constraints for the sequence of faunas and for the transition from fluvial to eolian dust-dominated depositional environment. The results will provide new insights in the link of the "Mongolian Remodeling" to the continental aridification and EOT, and highlights a more asynchronous faunal turnover than previously thought.

Geological setting and study area
Landscape of the Ulantatal area is characterized by a large open area with little topographic relief, consisting of a series of low gullies. The area is situated around 40 km northwest from Bayanhot, a major city in Alxa Left Banner in western Inner Mongolia. The Ulantatal Formation, named after the Ulantatal gulley traversing the northern part of the sequence, rests on the Bayanhot Basin of the North China Block, next to the N-NE trending Helan Mountains. Thick Cenozoic sedimentary cover overlies the region, and it is surrounded by Cha-Gu fault to the south, Bayanwulan Mountains to the N-NW and the Western Helan Mountain thrust fault associated with the Helan Mountains to the E (Zhang et al., 2009). The Bayanhot Basin developed on a junction of three tectonic formations, the Ordos Block, Alashan Block and Qilian Mountains Fold belt (Wang et al., 2017). The basin and its peripheral areas have developed to their current state through five evolutionary stages from Early-Middle Proterozoic rift to Late Proterozoic-Cambrian-Ordovician aulacogen, the Silurian-Devonian foreland basin, the Carboniferous-Permian compound basin, and finally into the Mesozoic-Cenozoic fault basin (Xiong et al., 2001). The Precambrian basement of the North China Block is overlain by marine and non-marine Carboniferous sediments, which are then covered by weakly or non-deformed Mesozoic-Cenozoic terrestrial sediments (Wang et al., 2017). The Helan Mountains some 50 km to S-SE from Ulantatal, faced at least three events of major uplift, in the Late Cretaceous, in the Eocene and in the Pliocene (Zhao et al., 2007;Wang et al., 2017).
Originally, some fossils from Ulantatal were found by workers from the Ningxia Geological Survey in 1977, and surveyed briefly by the IVPP (Institute of Vertebrate Paleontology and Paleoanthropology) in 1982 (Zhang et al., 2016, and references therein). These first relative age determinations suggested a Middle Oligocene age for the sediments according to similar lithological characteristics to the Qingshuiying Formation in Ningxia, China, and similar faunal composition to the Hsanda Gol Formation in the Valley of Lakes, Mongolia (Zhang et al., 2016, and references therein). Later a Chinese-German team returned to the area and made an extensive geological and paleontological survey producing abundant small mammal fossils (Vianey-Liaud et al., 2006). These investigations generated more precise information on the geological context of the area, leading to defining of three stratigraphical units: Ulan I, Ulan II and Ulan III. The Kekeamu locality was found during 1988 and 1989 near the main Ulantatal section with new fossiliferous horizons in lithological context, leading to extensive studies on these fossil collections (see Zhang et al., 2016, and references therein).
Our study region spans an area of approximately 11 km 2 , ranging from 39 • 10 to 39 • 12 N in latitude and from 105 • 31 to 105 • 33 E in longitude (Fig. 1). The Ulantatal region can be informally subdivided into three main sites, Kekeamu (KK), Shangjing (SJ) and Ulantatal main section (UTM). The sampled sites include subsections of KK (1 to 5), SJ and UTM (A to F) (Fig. 1). Zhang et al. (2016) have presented a preliminary profile of the sequence with thickness of over a hundred meters. The Ulantatal sequence has an unconformable lower contact to the brick red beds of the Eocene Qaganbulag (Chaganbulage) Formation, whereas unconformably overlying the Ulantatal Formation is the Wuertu Formation of early Miocene in age (Zhang et al., 2016).

Sedimentological methods
In the present work, twelve 15-25-m-thick exposures between Kekeamu 1 and Ulantatal Main Section E (stratigraphically F precedes A in the sequence) were selected for detailed sedimentological investigations (Fig. 1). The succession in this area is relatively well-exposed, with gentle bedding dips ranging from 1-2 degrees to SW. Standard logging techniques were used to construct detailed graphic logs of the sediment sequences in order to determine vertical and lateral facies changes throughout the area. Observations of the primary sedimentary structures, texture, sorting of the sediments, paleosol formation, bioturbation, palaeocurrent directions, spatial relationships and contacts between discrete beds were recorded from each unit. Colors were obtained by comparison to Munsell™ Color soil charts. The relative amount of CaCO 3 in the sediment was estimated in the field with diluted hydrochloric acid. The grain-size distributions in the size range of 0.15-2000 μm were determined for 57 selected samples using a Malvern Panalytical Mastersizer 2000 laser particle sizer. The ratio between coarse silt and medium to fine silt (U-ratio: 16-44 μm/5-16 μm; Vandenberghe et al., 1997) was calculated to the selected samples.
Multiple largely overlapping sections were investigated, but for the purpose of the publication, the composite lithostratigraphical columns are presented for the Kekeamu, Shangjing and Ulantatal main sections (Figs. 2, 3 and 4). Lithostratigraphic correlations within the KK and most of the UTM areas were made by means of intermediate lithostratigraphic sections, which enabled physical correlation. Although a direct lithostratigraphic correlation between the KK -SJ -UTM was hampered due to lack of distinct marker beds, correlation within a couple of meters was possible following the gentle strike and dip of the beds.

Rock magnetic and paleomagnetic methods
Altogether 1152 samples for paleomagnetic analyses were collected throughout the Ulantatal sequence every 10-50 cm from the twelve sections, yielding the composite sequence with thickness of approximately 110 m. All samples collected derived from fresh surfaces exposed by digging into the outcrop for at least 50 cm. Most of the samples were collected with a handheld water-cooled gasoline drill and oriented in the field with a magnetic and sun compass, with additional samples collected by using an electric portable drill and as oriented block samples. Both the block samples and drill cores were then prepared in the laboratory to either cubic or cylindrical specimens, respectively. Of the 1152 samples a total of 1087 samples were analyzed. The paleomagnetic and rock magnetic measurements were carried out at the Solid Earth Geophysics Laboratory of the University of Helsinki, Finland. Used rock magnetic and paleomagnetic methods are described in the Supplement.
During AF cleaning, an initial viscous component is isolated in the range of 0-5 mT and the other secondary component mainly by 15 mT. The characteristic remanent magnetization (ChRM) component was separated at 15-120/160 mT. The initial intensity drops to half of its original value (Median Destructive Field, MDF) with fields less than 40 mT indicating magnetite as the main remanence carrier. In addition, presence of magnetically harder phase is evident since the remanence intensity does not decay to less than 10% of the original value. Nevertheless, ChRM shows both reversed and normal polarity components trending towards origin (Fig. 5).
Obtained declination and inclination of ChRM are plotted in stratigraphic order and presented as composite logs (Figs. 2, 3 and 4), constructed from magnetostratigraphic and lithostratigraphic correlation of the sub-sections. Short normal polarity magnetozones are observed in lower and middle parts of the KK section. The upper half of the KK and the 25-m-thick SJ sections are dominantly reversed polarity but a large number or single samples show normal polarity. UTM shows five reversed and six normal polarity magnetozones, but a number of single samples showing reversed and normal polarity.

Descriptions
In general, the Ulantatal sequence has a strikingly uniform appearance dominated by laterally continuous massive silts, which show alternating reddish yellow (Munsell colors 5YR 6/6 and 5/6)  and reddish (Munsell colors 2.5YR 5/6 and 6/6) colors. Occasionally these massive beds contain signs of burrows, weak signs of pedogenic alteration and are calcareous in places. The sediments are organized in an overall fining-upward trend reflected in a subtle decrease of median grain size from the bottom of Kekeamu to the top of UTM. Strata in the Ulantatal area are not significantly de- formed in the sites studied, with beds only slightly (1-2 degrees) tilted to the SW.
Kekeamu (KK) section (Figs. 1 and 2) rests on the Eocene brickred Qaganbulag Formation with an unconformable contact and shows a generally upward fining trend. The basal part of the section consists of coarse silt interfingering with fine sand units that range in thickness from a few centimeters to few decimeters. Some of these beds are present as thin tabular beds while some have a shallow channelized geometry. Primary sedimentary structures, when present, include thin parallel lamination and ripple cross-lamination. Main part of the sequence is composed of lat-erally extensive beds of brownish red clayey siltstones and brownish yellow siltstones, occasionally truncated by thin beds of silty sands. The siltstones are mostly massive, well-cemented and often slightly calcareous. Subtle traces of bioturbation, i.e. marks of burrowing, alongside with convolute laminations and clay pebbles are recorded in the upper portion of the section. Vertebrate fossils are produced along the whole Kekeamu section with the richest occurrences in the lowermost part of the section within the lower 3 m (locality KM01), in the middle of the section ca. 23 m, and near the top of the section at ca. 30 m (locality 140909zzq01) level. West from the KK section, there is a series of low cliffs and hillocks, 2-4 m high, which expose a composite section of ca. 26 m (Fig. 3), called the Shangjing (SJ) section (Fig. 1). The sediment succession at the SJ comprises exceptionally monotonous massive yellowish brown-reddish brown siltstones with a mean bed thickness of around two meters. Slightly calcareous beds occur sparsely in the lower half of SJ whereas no distinguishable features exist in the upper half. Fossils can be found along the whole section with the main excavated fossil localities at ca. 9.5 m and 12.5 m.
Moving towards SW from the SJ is the Ulantatal main section (UTM; Figs. 1 and 4), which has a total thickness of ca. 60 m. UTM is made up by massive variegated clayey silt and silt beds with sporadic intercalations of thin parallel and cross-laminated fine sand lenses in the middle levels with non-scoured bases (at ca. 20 and 26 m levels of the local section). A coarse silt unit, 2-3-m-thick, is prominent at ca. 15-m level in the UTM section. This silt has a well-defined sharp base, resting over the clayey siltstones, is well-sorted and exhibits plane-parallel and ripple cross-   and Jingle Red Clay (Shang et al., 2016) from the Chinese Loess Plateau. C) A view of Kekeamu sub-section 5 (height 14.3 m) towards the Shangjing section to SW. D) GSD of selected samples from the Ulantatal sequence compared to Xifeng loess (Sun et al., 2002) and Red Clay from Xi'an (Lu et al., 2001). lamination. The silt unit shows a significant change laterally over a distance of a few hundred of meters. Much thicker exposure of this unit occur to the SW, where it attains a thickness of >10 m and forms a channelized sandy body, whereas it grades laterally into a decimeter-thick carbonate-nodule rich conglomerate towards NW. Another distinguishable feature of the UTM sequence is a distinctive 3-m-thick dark red (2.5YR 3/6) clayeysilt unit at about 18 m level that provides a widespread important stratigraphic marker traceable over kilometers along the outcrops. Subangular clay pebbles, usually 1-2 mm in size, have been recorded in the lower and middle portions of the sequence. The zones containing the clay intraclasts commonly occur with more abundant manganese precipitates associated with thin laminations, peds and ripple cross lamination.
One of the richest fossil levels of the UTM is located just below the coarser silt bed at ca. 12.8 m, associated with slightly calcareous material, manganese precipitates and slickensides. The highest fossil level of UTM encompasses numerous localities (UTM-E-01) ranging from ca. 42 to 48 meters also including one of the three main localities (130916ly02) at ca. 44 m. In addition, Ulan II (Vianey-Liaud et al., 2006) can be easily correlated to 12 to 15 mlevel.
Grain-size distributions (GSD) for the whole sequence are typically bimodal with a modal grain size of coarse silt (20-50 μm), and a minor modal value at ca. 4-7 μm (clay and fine silt) (Fig. 6D). The sand-sized fraction (>63 μm) is almost negligible. U-ratios mainly show low values typically ranging from 0.8 to 1.3.

Interpretations
The coarser siltstones with sand intercalations at decimeterscale that dominate the lowermost Kekeamu section are interpreted as floodplain deposits, where sandstone bodies with hydraulic structures relate to occasional enhanced fluvial activity.
The bulk of the Ulantatal sequence is characterized by the ubiquitous silty texture, massive structure and spatially correlative lithostratigraphy over vast areas, all features indicative of eolian dust lithofacies. These field observations are supported by the grain size distributions that clearly demonstrate the eolian origin of the deposits: most of the GSDs are characterized by two distinctive modal sizes (Fig. 6D) with a major peak at coarse silt and a minor peak at clay and fine silt, a feature similar to the Neogene Red Clay deposits in the Chinese Loess Plateau (e.g. Lu et al., 2001;Shang et al., 2016). Textural similarity with Jingle Red Clay (Shang et al., 2016) and Dadiwan loess   (Fig. 6B) further attest to an eolian origin for the studied Ulantatal sediments of the Chinese Loess Plateau (CLP). Furthermore, the U-ratios are comparable to Quaternary loess-paleosol deposits in the CLP, and resemble values from interglacial times (Vandenberghe et al., 1997).
The thin tabular beds of coarse silt and very fine sand recorded in the upper portion of the sequence are considered to represent short-lived surface runoff in the area. The rare occurrence of thin laminated clays/clayey silts indicates standing water bodies, while the zones containing subangular clay pebbles/rip-up clasts indicate subaerial exposure of these water-laid deposits and the incorporation of the clay sediments into the sediment load, corresponding to brief periods of flooding events.

Mammalian fossils in Ulantatal
From 2009 on, nine field seasons of exploration and excavation have been conducted in this area. These excavations have yielded a large amount of well-preserved specimens including jaws, skulls and postcranials totaling over ten thousand specimens. The specimens are stored in the Institute of Vertebrate Paleontology and Paleoanthropology, Beijing. Possibly, due to the tranquil depositional environment, most of the specimens are small mammals including rodents, lagomorphs, and insectivores. Of the localities, one from the lowermost KK section, and two levels from the UTM section with coarser sediments, are the richest with dense accumulations of fossils suitable for screen washing. Other localities have produced fewer specimens with less prominent fossil concentrations. Detailed biostratigraphy is a subject of future publications. Fossil levels were physically traced to the local sections by using stratigraphical markers and correlated to the sections with magnetostratigraphic control. Main fossil levels are shown in Figs. 2 to 4. Selected fossils from the localities KM01 and 130916ly02 are shown in the Figs. 7 and 8. Imaging was done with a Zeiss MA EVO25 Scanning Electron Microscope (SEM) in the Key Laboratory of Vertebrate Evolution and Human Origins of Chinese Academy of Sciences.

Fossils of Kekeamu
The lowermost fossil level from KK section is the locality KM01 (Fig. 2). Originally, Wang and Wang (1991) listed 12 taxa from the lowermost Kekeamu. With our new findings, the updated fossil list yields more than 17 taxa ( Fig. 7  Kekeamu fauna earlier than the middle Oligocene faunas from Hsanda Gol, Ulantatal main section level, Saint Jacques, and Wulanbulage, therefore suggesting an early middle Oligocene age. The Kekeamu fauna was later clustered with other Oligocene fauna in Meng and McKenna (1998), representing the earliest post "Mongolian Remodeling" fauna.

Fossils of the Ulantatal main section
The UTM (Fig. 4) hosts the richest fossil occurrences of the area. These include the fossil localities UTL1-8 (Vianey-Liaud et al. (2006) grouped them into Ulan I, II, and III). Based on our field survey, newly collected fossils, and the geographical location pre- sented in Vianey-Liaud et al. (2006), Ulan I ought to be in the lower part of the UTM, Ulan II at ∼15 m level of UTM, and Ulan III likely in the subsection C of the UTM (Fig. 4; 20∼25 m). Hence, the localities UTL1-8 cover less than 30 m of the strata in the lower and middle parts of the UTM section. Rodent fossils in the UTM area have been studied by Huang (1992) The highest fossil level (locality 130916ly02) in the UTM section (Fig. 4) is situated in the lower part of subsection E (ca. 44 m), and produced fossils (Fig. 8) including Ruminantia indet., Didymoconus indet., Palaeoscaptor acridens, Sinolagomys cf. kansuensis, Ochotonidae gen. et sp. nov. Leporidae indet., Tsaganomyidae indet., Parasminthus tangingoli, Tataromys parvus, Yindirtemys grangeri and Eucricetodon jilantaiensis. These fossils resemble those from Yindirte (Yandantu) fauna from Taben-Buluk area, Gansu, which is considered to represent late Oligocene (Wang et al., 2003). Wang et al. (2008) estimated the Yindirte fauna to be around 23-25 Ma by referring to an unpublished paleomagnetic study in the nearby Tiejianggou section. However, the more primitive characters of Sinolagomys cf. kansuensis and the Ochotonidae gen. et sp. nov. from locality 130916ly02 than those from Yindirte fauna, indicates an earlier age.

Biostratigraphical correlation
Despite the full description of Ulantatal fossils is still ongoing, the present fossils discovered from the Ulantatal sequence provide a temporally continuous and successive faunal sequence in contrast to the local faunas presented in most of the previous studies. The only comparable long sequence in Asia is from the Valley of Lakes area in Mongolia . Although biogeographic differences may exist, fossil composition from the locality 140909zzq01 in the upper part of the KK section, including Karakoromys decessus, Anomoemys lohiculus, Zaaralestes, and Desmatolagus gobiensis, and those from the SJ section, are all characteristic elements of Biozone A from the Valley of Lakes area .
The KM01 is situated around 25 meters below the 140909zzq01 and taxonomic composition of small mammal assemblage from this site displays a constellation of primitive and derived morphological characteristics that biochronologically resembles the sites from the late Eocene and early Oligocene of China and Mongolia. Presence of archaic taxa Ageitonomys neimongolensis (Ctenodactyloidea), and Palaeoscaptor sp. nov. (Erinacidae), Karakoromys cf. decessus (Ctenodactylidae), and Allosminthus sp. (Dipodidae) at the KM01 suggests the fauna being biochronologically older than that in Biozone A (∼33.5 to 31.5 Ma), based on the evolutionary stage of their morphological traits. Palaeoscaptor sp. nov. has double rooted third premolar but shows similarities with Palaeoscaptor tenuis from Biozone A in size and morphology. Karakoromys cf. decessus has isolated metacone cusp on most of the upper molars. Larger anteroconid on lower first molar, a less developed mesoloph on the upper molars, and a less developed mesoconid on the lower molars differentiate Allosminthus sp. from Allosminthus khandae from Biozone A. In addition, Desmatolagus cf. vetustus has lower tooth crown than the early Oligocene Desmatolagus gobiensis. Comparing with the late Eocene faunas from China, the KM01 is most similar with the Qujing fauna by sharing Allosminthus (Dipodidae), Heosminthus (Dipodidae) and Eocricetodon (Muroidea) in the genus level with more derived species. Additionally, presence of odd-toed ungulates Schizotherium cf. avitum (Chalicotheriidae) and Ardynia mongoliensis (Hyracodontidae) may bracket, or likely document, the Eocene-Oligocene transition (Wang and Wang, 1991). Overall, the primitiveness of comparable taxa and absence of typical elements from Biozone A, e.g. Cricetops, Selenomys and Anomoemys suggest the KM01 fauna obviously earlier than those from Biozone A of the Valley of Lakes area, Mongolia.
The lower and middle part of the UTM section are well represented by units Ulan I-III. The three units are very similar sharing most of the Ctenodactylidae and Cricetidae taxa, with the presence of Karakoromys desessus (only two teeth) in the Ulan I, and some more advanced taxa from Ulan III. Fossil Zapodids are similar in Ulan II and III (Huang, 1992). Based on the presence and absence of some key taxa, Schmidt-Kittler et al. (2007) correlated Ulan I to Biozone B of the Valley of Lakes sequence, Ulan II between Biozones B and C or C, and Ulan III to Biozone C1. However, the abundance of Huangomys and the presence of Cricetops in Ulan II fit the Cricetops taxa range zone, and Huangomys abundance subzone in Biozone B according to updated data of Harzhauser et al. (2017).
The stratigraphically highest locality 130916ly02 in the UTM section produced an obviously later age than Ulan III, with e.g. Tataromys parvus, Yindirtemys grangeri and Sinolagomys cf. kansuensis. The comparison with the Biozones of the Valley of Lakes area is not obvious. The earliest occurrence of Sinolagomys cf. kansuensis from locality 130916ly02 may suggest its correlation with Biozone C or C1.

Correlation to the geomagnetic polarity time scale
In this work, we present for the first time a magnetostratigraphic framework for the fossiliferous Ulantatal sequence. Overall, our magnetostratigraphy suggests an age range of ca. 35-27 Ma based on fossil evidence and structural characteristics of the sequence (Figs. 2-4 and 9). In the composite section (Fig. 9), the magnetic polarity record can be subdivided into eight normal polarity (N1-N8) and seven reversed polarity (R1-R7) magnetozones, the single site polarity intervals are considered uncertain and hence not numbered. The lowermost part of the section is characterized by two long reversed-polarity intervals (R1 and R2). In accordance with the biostratigraphic age control placing the lowermost fossil locality (KM01) to the latest Eocene-earliest Oligocene, the most plausible correlation of these reversed zones is to chrons C13r and C12r (Gradstein et al., 2004): The mammal fauna, especially Palaeoscaptor, Karakoromys and Allosminthus from locality KM01, is clearly more primitive than those forms found at locality 141409zzq01 (uppermost Kekeamu) and Biozone A from the Valley of Lakes, Mongolia, thus indicating an older age for KM01. The age of Biozone A, in turn, is well-constrained by the overlying basalt with an 39 Ar/ 40 Ar age of ca. 31.5 Ma and the fauna residing within magnetochron C12r . With these constraints, the long reversed polarity R2 containing locality 140909zzq01 should be correlated to C12r, and R1 with locality KM01 to C13r. Consequently, the basal normal polarity interval (N1) correlates to chron C15n and the short normal N2 to C13n. Albeit short, N2 is well established as it is recorded in the several overlapping subsections of the KK. The distinctive pattern of two long normal polarity zones N3 and N8 separated by four shorter normal (N4-N7), four short reversed (R3-4, R6-7) and one longer reversed (R5) polarity intervals in UTM section allows for correlation to polarity chrons from C12n to C9n. An alternative correlation placing the long reversed zones R1 and R2 to C12r is considered to be unlikely as it omits N2 and suggests too young age for KM01 placing it within Biozone A. This correlation can be ruled out as since the fauna in KM01 is necessarily older than that of Biozone A.
Magnetochron C15n was identified at the base of the sequence, and by interpolation within the C13r, the Eocene-Oligocene boundary (33.9 Ma; Gradstein et al., 2004) can be placed at 13-m level in the KK section, ca. 10 m above the KM01 fossil locality.
Magnetostratigraphic correlation between the KK and the SJ sections has a degree of uncertainty, since the overlap falls within the long interval of reversed polarity. However, lithostratigraphical correlation within a couple of meters was possible tracing the gentle strike and dip of the beds.
High-resolution analyses revealed several polarity intervals defined by single stratigraphical level only. Particularly notable are the ten short normal polarity zones within the reversed polarity intervals R1 and R2 (C13r and C12r;Figs. 2,3 and 9). Even though these are not included in standards, similar short-duration events are a feature frequently discerned in sequences of early Oligocene in age (e.g. Tauxe and Hartl, 1997) and may provide separate evidence for the proposed age range. These thin zones may imply either short periods of low geomagnetic field intensity or frequent reversals of the geomagnetic field (e.g. Tauxe and Hartl, 1997).
The newly excavated fossil localities (Figs. 2 and 4) and units Ulan I-III of Vianey-Liaud et al. (2006) can be correlated with age-depth model constructed from the thickness of the sediments and biostratigraphical and paleomagnetic data (Fig. 10). Fossil sites close to the sections can be easily plotted but others traced over long distances have some stratigraphic uncertainty. The KM01 locality with its pre-Biozone A fauna can be dated to ∼34.8 Ma, placing the earliest post "Mongolian Remodeling" fauna (e.g. Meng and McKenna, 1998) in the latest Eocene. 140909zzq01 locality in the uppermost KK section is dated to ∼32.5 Ma, in good agreement with the age inferred from the fossil record. With the EO boundary lying between these two localities, more exploration of this interval is highly needed for a better understanding of faunal transition across the boundary from the Eocene to the Oligocene. Fossils from SJ section are all within the C12r (R2), having an approximate age range of 32 to 31 Ma. Ulan I of Vianey-Liaud et al. (2006) in the UTM section can be correlated to C12n and interpolated to be of ∼31 Ma. Ulan II has an age range of ca. 30.2 to 30.6 Ma, fitting well with the Cricetops taxa range zone, and Huangomys abundance subzone  in Biozone B, which is dated to ∼31.5 to 28 Ma . Ulan III has an estimated age range between 28 to 29 Ma. The youngest fossil level (130916ly02) falls in the interval of C9r, having an interpolated age of ∼27.7 Ma.
The age-depth model (Fig. 10) shows sediment accumulation rates ranging from ca. 10 to 18 cm/kyr. Sedimentation rate can be used to infer interrelations between rock denudation and sedimentation within the studied region. Preservation of red clay and loess deposits require stable tectonic environment both during and after the deposition (Guo et al., 2008). Generally, accumulation rates stay low and constant through time, suggesting relatively stable erosional conditions prevailing in the area without any major tectonic events. However, two intervals of lower sedimentation rate from around 35.0 to 33.2 Ma (0.5-3 cm/kyr) and across 30 Ma (3-5 cm/kyr) can be inferred from the age-depth-model. The second interval at around 30 Ma may be attributed to increased erosion associated with the deposition of laterally widespread horizons of coarser sediments observed at ca. 15 to 18 m of UTM. This erosional and higher energy event might be a reflection of a period of minor uplift of the nearby Helan Mountains during this time (Zhao et al., 2007).

Paleoclimatic implications
Based on the massive structure of major part of the lithologies, fine-grained sediment typical for transport in suspension by wind, and bimodal grain-size distribution strikingly similar to younger loess and Red Clay deposits in China (e.g. Lu et al., 2001;Shang et al., 2016), we interpret the Ulantatal sediments to be primarily eolian in origin. The onset of eolian deposition in Ulantatal is dated at ca. 34.8 Ma, 10 m below the Eocene-Oligocene boundary in the Kekeamu section, which corroborates the onset of aridification in mid-latitude Asia during the Eocene (e.g. Dupont-Nivet et al., 2007;Abels et al., 2011;Licht et al., 2016). In principle, largescale and thick loess and Red Clay sedimentation requires a sizable and arid source area to allow deflation, and strong enough winds to transport the eolian dust, in addition to suitable geomorphological conditions to capture the deposited particles (Pye, 1995). Even Fig. 9. Correlation of obtained polarity zones to the GTS2004 (Gradstein et al., 2004). Polarity zones within the KK and the UTM were constructed with magneto-and lithostratigraphic correlations of sub-sections, whereas the shaded columns between KK-SJ and SJ-UTM represent correlation with lithostratigraphic and structural characteristics of the sequence. Main fossil localities (KM01, 140909zzq01, 130916ly02) and the EOB are indicated in the figure. though the exact sources of this dust may be various, including alluvial fans, reworked fluvial sediments and dry lake beds (e.g. Nie et al., 2014), the most important implication is the requirement of a long-lived source area that allows a continuous dust supply over millions of years. Such long lasting period of sedimentation is indicative of regionally enhanced aridification (e.g. . The onset of major eolian deposition has generally been considered a late Oligocene phenomenon when dust accumulation initiated on the CLP some 25 to 22 Ma (e.g. Licht et al., 2016). Recent evidence, such as the presence of Eocene dust deposits in the Xining and Xorkol basins and dust component in the fluviolacustrine sediments at Xijiadian, has indicated eolian deposition may have begun already during early Cenozoic in the western China and CLP (Licht et al., 2016, and references therein;. Ulantatal sequence provides an unambiguous and well-constrained record of widespread and long-term deposition of eolian dust in the vicin- ity of the CLP. The main contributor for this semi-arid to arid climate in Central Asia in Eocene and Oligocene has likely been the withdrawal of the Paratethys Sea and subsequent abatement of moisture (e.g. Bosboom et al., 2014a;Meijer et al., 2019). Indicated by highly seasonal temperatures, the shallow Proto-Paratethys Sea did not have a strong dampening effect that may imply monsoonal circulation was active in Central Asia (Bougeois et al., 2018), supporting the regional evidence of strong Eocene monsoons (Licht et al., 2014). The sea provided some moisture to Central Asia through mid-latitude westerlies but was strongly seasonal receiving the moisture during winters (Bougeois et al., 2018;Caves Rugenstein and Chamberlain, 2018), in contrast to anticyclonic circulation during summer (Botsyun et al., 2019).
Our new fossil discoveries combined with the earlier results (Vianey-Liaud et al., 2006;Gomes Rodrigues et al., 2014;Zhang et al., 2016) show a striking lack of faunal change through the Eocene-Oligocene transition. However, we present the first evidence of post "Mongolian Remodeling" fauna (Meng and McKenna, 1998;Sun et al., 2014) existing already before the marine EO boundary, contradictory to earlier studies suggesting synchronous faunal turnovers post-dating the EO boundary in East Asia (Kraatz and Geisler, 2010;Zhang et al., 2012;Sun et al., 2014). Our lowermost Kekeamu fossil locality KM01, mainly composed of rodents typical for the "Mongolian Remodeling" fauna, can be interpolated to be of 34.8 m.y. of age, suggesting the faunal turnover from medium sized perissodactyl to rodent and lagomorph predominance significantly predating the marine EO boundary. Unfortunately, the lack of outcrops and productive localities in the late Eocene prevents a precise dating of the faunal turnover. It must have preceded the change in lithology from fluvial to predominantly eolian by an unknown but probably not very long interval. That the faunal turnover could not coincide with the lithological change is shown by the late Eocene fossil locality KM01, with a post-turnover fauna at the lithological transition.
In principle, this could be interpreted as an anomaly in a separate basin, but since there is evidence of terrestrial faunal responses predating the marine EO boundary also in Europe (Joomun et al., 2010) and South America (Kohn et al., 2004), it is likely these responses have been driven by a climatic and environmental change which had begun before the Eocene-Oligocene transition. In addition, recent paleobiome and MAT models reconstructed from pollen data indicate no uniform changes across the EOT globally; in Central Asia, the spread of arid biomes was suggested already during the early Eocene (Pound and Salzmann, 2017). The increase of xerophytic shrublands in Central Asia (Pound and Salzmann, 2017) and mammalian communities mainly composed of rodents and lagomorphs recovered in this study, are in concert with the evidence of the widespread Eocene aridification in mid-latitude Asia suggested previously (e.g. Dupont-Nivet et al., 2007;Abels et al., 2011;Licht et al., 2016;Bougeois et al., 2018;Meijer et al., 2019). Therefore, our results are consistent with the association of "Mongolian Remodeling" with increased aridification (e.g. Meng and McKenna, 1998), but challenge the concept of synchronous biotic turnovers with the EOT in East Asia, and highlight the ability of terrestrial systems to respond more rapidly to climatic and environmental changes.
In summary, our new age-depth model allows dating of terrestrial biochronostratigraphical units during the Eocene-Oligocene transition and for the first time introduces the Ulantatal sediments and rich fossil faunas in a high-resolution temporal context. The sequence provides clear evidence for the onset of eolian dust accumulation in the latest Eocene at 34.8 Ma. It also reveals the establishment of post "Mongolian Remodeling" fauna well before the marine EO boundary and before the lithological change to primarily eolian mode of deposition. While the shift towards more arid environments is consistent with the accumulating evidence of the onset of aridification of mid-latitude Asia already during the Eocene (e.g. Abels et al., 2011;Bosboom et al., 2014b;Licht et al., 2016;Bougeois et al., 2018;Caves Rugenstein and Chamberlain, 2018), and its association with the "Mongolian Remodeling" (e.g. Meng and McKenna, 1998), it does not support the scenario of a simultaneous faunal turnover coupled with the EOT but a stepwise or gradual change instead. Our discoveries highlight more dynamic responses of terrestrial systems in changing climatic and environmental conditions across the Eocene-Oligocene transition.

Declaration of competing interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.