INTRODUCTION

In 1928, the Central Asiatic Expedition by the American Museum of Natural History found the first “Pliocene” locality near a freshwater well known as Gur Tung Khara Usu in central Inner Mongolia, China (Andrews, 1932; Spock, 1929). Thus began a series of sensational discoveries that produced a fossil mammal assemblage best known for its shovel-tusked elephant Platybelodon grangeri. For a long time Tunggur was the only Chinese middle Miocene locality widely known in the world. This pre-Hipparion (three-toed horse) fauna, one of the richest in the Chinese middle Miocene, became the basis of the Tunggurian land mammal age in East Asia. Despite this promising start, however, geological exploration of the region was discontinued because of wars and political instabilities. With the exceptions of a brief exploration by a Sino-Soviet Paleontological Expedition in 1959 and more recent fieldwork by the Institute of Vertebrate Paleontology and Paleoanthropology in 1986–1987 (see History of Study below), no systematic field study was attempted. As a result, much basic information is lacking, such as the status of the type section and the locations of historically important localities. Without such information, faunas from historical collections were often scrambled into a single heterogeneous assemblage.

The present field studies, carried out between 1993 and 2000, attempt to address five basic issues: (1) to relocate classic localities; (2) to establish lithological and biostratigraphic criteria for correlation within Tunggur strata; (3) to conduct paleomagnetic studies of key sections; (4) to update the faunal lists produced from each localities; and (5) to summarize evidence for paleoenvironments of the regime.

METHODS

Geographic names of the Tunggur area have been complicated by a combination of cross-translations among English (sometimes French), Mongolian, Chinese, and occasionally Manchurian. Mongolian location names in early publications were sometimes translated to Chinese, which were subject to interpretations by individuals with different backgrounds in Chinese dialects, and finally to English, which can be based on either Wade-Gile or Pinyin systems. To compound these problems, Mongolian location names may change within a short time because nomadic tribes have no permanent settlement. With the exception of well-established names, such as Tunggur and Tairum Nor, we use names published in Chinese 1:100,000 topographic maps (published in 1971). These Chinese translations of Mongolian names are further translated to English in the standard Pinyin system of the People's Republic of China.

CAE archive photographs of the Tunggur localities are stored in the general library of the American Museum of Natural History. Latitudinal and longitudinal coordinates were measured by a Trimble Navigation Scout Master GPS (global positioning system) set in the WGS-72 mapping datum. All GPS measurements were taken before May 1, 2000, when the “selective availability” (a form of data degradation) was switched off by the Department of Defense of the United States, and thus they have a degraded accuracy of ±30 m.

HISTORY OF COLLECTIONS

In June 1928, Roy Chapman Andrews of the American Museum Central Asiatic Expeditions made an excursion southeast of the Eren Dabasu (the present day city of Erenhot), a vast area only briefly explored in the southern parts by Swedish geologist J. G. Andersson and French vertebrate paleontologist P. Teilhard de Chardin (Andersson, 1923; Teilhard de Chardin, 1926a, b). In a fluvial sand exposure 50 miles southeast of Erenhot, Andrews' team found a partial “mastodon” skeleton near a well named Gur Tung Khara Usu (Andrews, 1932) (Fig. 1). Fieldwork for more than 10 days resulted in a small collection of two species of proboscideans, a carnivore, a chalicothere, a rhino, a cervid, an antelope, some turtle shells, and freshwater bivalves. The fossil vertebrates were estimated to be “Pliocene” in age, representing a previously unknown fauna, and the bone bed was promptly reported by L. E. Spock (1929) as Tung Gur Formation, a name reversed from the local well Gur Tung Khara Usu.

After a one-year hiatus, the CAE team carried out a more extensive expedition in the Tunggur area in 1930. Collections were concentrated in two new localities, Platybelodon Quarry and Wolf Camp (Fig. 1). The Platybelodon Quarry yielded a series of exquisitely preserved Platybelodon skeletons, ranging from infants to full adults, and the Wolf Camp produced a rich variety of medium-sized mammals. Together with a few other lesser known localities, the Tunggur Formation produced 28 species of mammals, most of them new to science. These fossils, housed in the AMNH, were subsequently described by various specialists: two proboscideans (Osborn, 1929; Osborn and Granger, 1931, 1932), one anchitheriine horse (Colbert, 1939b), one chalicothere (Colbert, 1934a), eight carnivorans (Colbert, 1939a; Hunt and Solounias, 1991), one suid (Colbert, 1934b), two bovids (Pilgrim, 1934), four cervids (Colbert, 1936b, 1940), one giraffe (Colbert, 1936a), four rodents (Stirton, 1934; B.-y. Wang, 1988; Wood, 1936), two lagomorphs (Dawson, 1961), and three rhinocerotids (Cerdeño, 1996) (see table 1 for a list of taxa and their occurrences).

Wars and political difficulties prevented further exploration of these rich deposits until 30 years later when a brief visit to the Tunggur area was organized by a joint Sino-Soviet Paleontological Expedition in 1959 (Chow and Rozhdestvensky, 1960). Materials obtained from this expedition, however, were mostly lost or destroyed as relationships between the two countries deteriorated, and only two new taxa, a beaver and a carnivore, were published as a result of this expedition (Li, 1963; Zhai, 1964). After another long hiatus, a renewed effort by the IVPP in 1986–1987 led to substantial progress both in geology and paleontology (Z.-x. Qiu et al., 1988). In particular, intensive screen-washing at the Moergen locality yielded more than 20 species of small mammals, nearly doubling the previously known number of taxa, and correcting the traditional bias against small mammals in the early collections. The most important result of this latter expedition is a monograph on Tunggur small mammals (Z.-d. Qiu, 1996). In a follow-up study, we attempted to correlate small mammal faunas in and around the Tunggur area (Z.-d. Qiu and Wang, 1999), which includes faunas from early Miocene to Pliocene. Within the Tunggur Formation, we differentiated a new fauna near Tairum Nor, which is older than the classic Tunggur fauna.

Independently, various regional geological survey teams have attempted to synthesize the local geology into a mappable regional framework (e.g., Bureau of Geology and Mineral Resources of Nei Mongol Autonomous Region, 1991), which were often too coarse an outline to be of assistance in field studies. More recently, Dong Mingxing (1993), a graduate student of China University of Geoscience, briefly explored problems related to the regional correlation and depositional environment of the Tunggur Formation. However, with the exception of two brief reports on contact relationships and fossil tracks (Dong, 1999; Dong and Ai, 2001), he did not publish his thesis.

GEOLOGICAL SETTING

Cenozoic deposits east of Erenhot generally lie in the upper part of the fill of a series of northeast–southwest oriented rift subbasins. Collectively, these fault-controlled subbasins form the Eren Basin, initially filled in the Jurassic through Cretaceous by fluviolacustrine deposits and often forming half grabens or asymmetric grabens, up to 3–5 km thick (Lin et al., 2001). The Mesozoic deposits are mostly in the subsurface, but they may crop out in basin margins, as near the Qagan Nor Chemical Factory (Jianchang in fig. 1).

In central Inner Mongolia between the cities of Erenhot and Xilinhot is a large span of rolling plains with few notable landmarks (inset map in fig. 1). Besides occasional patches of granites, usually weathered to ground levels, the most conspicuous features are a series of low tablelands, commonly 40–80 m above ground level and forming elevated benches that extend tens of kilometers in a single stretch. Outcrops along their edges often produce fossils. The Tunggur Tableland is outlined in a northeast–southwest-oriented, tongue-shaped platform, although its northeastern extent becomes gradually less well defined (fig. 1). To the east, another large tableland near Baogeda Ula is capped by a lava, forming a well-defined platform. Late Miocene in age, this latter tableland is similar to another series of lava-capped platforms along the Huiteng River (Chiton Gol) farther southeast of Tunggur, as first explored by Teilhard de Chardin (1926b).

Neogene deposits at Tunggur and surrounding areas are often confined to small basins of just a few square kilometers that are less than a few meters in thickness (e.g., Gashunyin Adege, Shala, Amuwusu; see fig. 1). The Tunggur Formation and the overlying Baogeda Ula Formation, on the other hand, represent much larger depositional basins of several hundred square kilometers in size and up to 100 m in thickness. Over the entire Tableland the Tunggur Formation is flat-lying, and structural disturbance of these rocks is minimal. Teilhard de Chardin (1926b) described volcanic cones along the Huiteng River area south of Xilinhot. Basalt lava sheets from these eruptions, several meters thick in some places, top the presumed Baogeda Ula Formation and form a resistant bench. Their fresh appearance suggests a young age (Whitford-Stark, 1987), and recent K-Ar radiometric dates indicate at least three episodes of eruptions (3.27, 2.08, and 0.47 Ma) during the late Pliocene through Pleistocene (Liu, 1999). These lava sheets and the underlying Baogeda Ula Formation form the most conspicuous tablelands to the south and west of Xilinhot. The Tunggur Tableland, however, is not capped by the lava, which lies too far to the south and east. Instead, the topography of the Tunggur Tableland is mostly shaped by resistant marls or sandstones. Where there is no marl or sandstone, the Tableland tends to gradually blend into the lowland, and exposures are covered by vegetation.

Contact relationships of the Tunggur Formation with rocks above and below were not an issue during the CAE and Sino-Soviet expeditions. Z.-x. Qiu et al. (1988: 403) first cited a locality “25 km northeast to Saihan Gobi” to the north of the Tunggur Tableland as producing Lophiomeryx, “indicating the Oligocene age of the fossil-bearing beds”, although they acknowledged that no direct contact can be observed between Tunggur Formation and its underlying deposits. Dong and Ai (2001) recently reported some fragmentary materials of a brontothere Protitan and a chalicothere Eomoropus in a locality (43°31′21″N, 113°13′07″E) not far from the northern rim of the Tunggur Tableland at an elevation of 986 m. They placed these two forms in the middle Eocene Irdin Manha Formation and recorded a disconformable contact between the Tunggur Formation and the underlying Irdin Manha Formation. To the south of the tableland, Meng et al. (1996) recently reported a new fauna of early Miocene age from Gashunyin Adege (see also Z.-d. Qiu and Wang, 1999). The Gashunyin deposits, however, are confined within a small basin in an area of granite basement rocks, and they are unlikely to be superposed by the Tunggur Formation.

As for the overlying sediments, the Editorial Committee of Stratigraphic Table of Inner Mongolia Autonomous Region (1978) first established the Baogeda Ula Formation based on escarpments near the village of Baogeda Ula, approximately 20 km northeast of the easternmost Tunggur locality, Roadmark 482 Loc. However, the base of the Baogeda Ula Formation, which contains a Hipparion fauna, cannot be observed at the type section. The 482 locality is the only section where a conglomerate is seen to overlie disconformably on the Tunggur Formation (see fig. 8). Although the thin conglomerate (1 m) cannot be directly correlated to the type section of the Baogeda Ula Formation, its geographic proximity and fossil content suggest this contact relationship relative to Tunggur Formation.

DESCRIPTION OF TUNGGUR LOCALITIES

Spock's (1929) original description of the Tunggur Formation is sketchy, and all localities were in reference to a single well “Gur Tung Khara Usu” (see further comments below). Spock reversed the first two words to coin the stratigraphic unit “Tung Gur Formation”. Osborn and Granger (1932: fig. 1) subsequently used “Tung Gur Tableland” to refer to a low plateau, whose escarpments along the northern, western, and southern edges form the exposures of the Tunggur Formation. We continue this usage and the name Tunggur thus refers to “the whole area where the Tung-gur Formation is actually or potentially exposed” (Z.-x. Qiu et al., 1988: 400). The original spelling of “Tung Gur Formation” in Spock (1929) was subsequently modified to “Tung-gur Formation” by Z.-x. Qiu et al. (1988) and further simplified to “Tunggur Formation” in Z.-d. Qiu (1996). We follow this latest form throughout this text.

Precise location of the stratotype and other CAE localities in the Tunggur area was hampered by the lack of precise topographic maps, scarcity of geographic reference points in an essentially featureless terrain, and often inadequate photographic or descriptive documentation of the exposures and rock sequences. Subsequent field studies, widely separated in time and by different groups of people, had only the sketch map published by Osborn and Granger (1932: fig. 1) for reference. The 1959 Sino-Soviet Paleontological Expeditions concentrated on a few quarries, and their collections made minimal reference to the historical localities (Chow and Rozhdestvensky, 1960). The IVPP 1986–1987 field teams attempted to relocate the CAE localities, this time with the benefit of accurate Chinese 1:100,000 topographic maps (Cartographic Bureau of People's Liberation Army, 1972). The IVPP team tried to reconcile their findings on the basis of Osborn and Granger's sketch map and a few distance estimates made by Andrews (1932), which may account for many of the errors in their estimates (see description of localities below). Z.-x. Qiu et al.'s map (1988: fig.1) is essentially that of Osborn and Granger's (1932: fig. 1), whereas Z.-d. Qiu (1996: fig. 1) adopted the Chinese topographic map but still used the relative positioning of the various localities in Z.-x. Qiu et al. (1988). Our own map is based on the same topographic map, with the GPS positions being plotted in Figure 1.

DISTRIBUTION OF TUNGGUR FORMATION

In his original description of the Tunggur Formation, Spock (1929) only described two exposures that were seen in the 1928 field season, namely the stratotype at North Camp and the South Camp at Tairum Nor. After a more extensive exploration in the 1930 field season, Osborn and Granger (1932) were able to publish a map outlining the western part of the Tunggur Tableland, whose escarpment serves as a prominent topographic feature delineating the distribution of the Tunggur Formation. The apparent drawback in equating the Tableland with the Tunggur Formation is the fact that the rim of the Tableland is essentially an erosional feature controlled by resistant rocks such as channel sandstones or marls. The marls are more numerous in the northern and western escarpments where they form resistant hard caps, whereas the southern escarpment follows a continuous channel sandstone that stands out from less resistant red mudstones. No basement rock is exposed along the Tableland rim, and the true extent of the Tunggur Formation (i.e., the original depositional basin) must be considerably larger than is now outlined by the Tableland. Nonetheless, the simple delineation based on the Tableland edges serves well for paleontological investigations since fossils are always found in exposures along the edges, whereas areas of more gradual relief, both on top and below the Tableland, are covered by grasses growing on Pleistocene gravels and sands.

The need for regional stratigraphic correlations by geological survey teams, however, often leads to estimates about the distribution of the Tunggur Formation that are far more extensive than can be justified paleontologically. Therefore, in a comprehensive summary of past surveys recently published by the Bureau of Geology and Mineral Resources of Nei Mongol Autonomous Region (1991), the Tunggur Formation, as a mappable unit, was shown to crop out from 42°00′ to 44°20′ in latitude and 111°30′ to 114°30′ in longitude, a rectangle covering an area of approximately 60,000 km2, many times larger than that outlined by the Tunggur Tableland. Much of the larger distribution was based on superficial lithostratigraphic characteristics, often simply a gross generalization of “red bed”, with little or no knowledge about fossils or other means of age control. Such an approach, relying on gross similarities of sedimentary packages, apparently assumes a large, continuous lacustrine basin, an idea that is prevalent in most Chinese geological survey teams dealing with terrestrial sediments. Our own surveys suggest that Neogene terrestrial sediments in central Inner Mongolia are often scattered in small to medium-sized basins. Localities of different ages can be spaced within a few kilometers of each other. For example, the Amuwusu and Shala localities, which were mapped as the southwesternmost exposure of the Tunggur Formation by the regional geological survey (Bureau of Geology and Mineral Resources of Nei Mongol Autonomous Region, 1991), contain vertebrate fossils younger than the Tunggurian land mammal age (Z.-d. Qiu and Wang, 1999).

Problems further arise because of the different interpretations of stratigraphic relationships between the Tunggur Formation and the Baogeda Ula Formation. As discussed above, the regional geological teams proposed a disconformable relationship between the two formations within the Tunggur Tableland. Assuming a southward dip, the Tunggur Formation was shown croping out only along the northern escarpment of the Tableland while the main body of the Tableland, as well as its southern escarpments, were all interpreted to be Baogeda Ula Formation. Such an interpretation contradicts past paleontological studies (CAE, SSPE, IVPP) that consistently regarded the Tunggur Formation as flat lying and containing vertebrate fossils of Tunggurian age only, without any evidence of the presence of a strata containing Hipparion. Therefore, although the original distribution of the Tunggur Formation must have been larger than the Tableland, for the purpose of paleontological studies we continue to confine the Tunggur Formation to the Tableland edges (fig. 1).

LITHOSTRATIGRAPHIC CORRELATIONS

Stratigraphic relationships among various exposures in the Tunggur Tableland have not been a primary concern until recently. Spock (1929: figs. 4, 5) illustrated two representative sections, one at the stratotype locality at Mandelin Chabu and the other at the Tairum Nor locality. He did not attempt to correlate these beds and only noted (Spock, 1929: 5) that “lack of continuity and a pronounced lens-like form are the outstanding characteristics of the beds.” Both of these characteristics, of course, hinder precise correlations. Such an obstacle is not a problem if one assumes that vertebrate fossils, coming from relatively thin beds, represent a contemporaneous fauna. In practice, the term “Platybelodon grangeri Life Zone” of Osborn and Granger (1932) serves this purpose, and fossils from the Tunggur Formation have generally been regarded as a single fauna. Z.-d. Qiu (1996), however, demonstrated that this is not the case for small mammals—two distinct assemblages were recognized in Moergen locality, corresponding to what he informally recognized as upper (bed V) and lower members (beds I–IV) of the Tunggur Formation. Precise correlations among various localities of the Tunggur Tableland thus become a matter of some importance to prevent a mixing of faunas.

Z.-x. Qiu et al. (1988: 403) redescribed the Mandelin Chabu section (a few hundred meters east of Spock's original North Camp; the top marl and capping cross-bedded sandstone in Spock's section were not mentioned by Z.-x. Qiu et al.) and regarded the section as typical of the Tunggur Formation that “holds considerably stable, thus renders no particular difficulty for section-to-section correlation.” In particular, they singled out the upper red mudstone (bed 4 below) as a marker bed, “owing to its bright color and little changed thickness.” This theme was further fleshed out in Z.-d. Qiu (1996: figs. 3, 4), who recognized five distinct beds as characteristic of the Tunggur Formation. As typified from Moergen, these beds are (from bottom to top): (1) variegated mudstone, (2) mostly sandstones, (3) sandy mudstones with large quantity of concretions, (4) brick-red mudstone, and (5) greenish to brownish mudstone and sands. Five measured sections were presented by Z.-d. Qiu (1996: fig. 4), which were correlated on the basis of the above five beds.

Close examinations, however, reveal complexities not previously noted. First, there are widespread marls throughout the northern and western escarpments. These marls, although lenticular in some places, may serve as possible markers in combination with other beds. Secondly, the stability and uniqueness of the brick-red mudstone is questionable. Two of the five sections measured by Z.-d. Qiu (1996: fig. 4) lack this layer (including the CAE Platybelodon Quarry and IVPP 86021, two of the most productive localities), whereas others, such as the entire southern escarpment, have more than one red bed. As noted by Spock (1929), the stratigraphy is further complicated by a general lack of lateral continuity between exposures and the lenticular nature of the channel sandstones.

A radically different scheme was proposed by the regional geological survey teams (e.g., Bureau of Geology and Mineral Resources of Nei Mongol Autonomous Region, 1991), in which the Tunggur Formation is only recognized on the northern escarpment of the Tableland, and the Tableland proper was mostly assigned to the overlying Baogeda Ula Formation that crops out in the southern escarpment (see also above). Fundamental to this theme are two assumptions: (1) A southward dip characterizes the entire Tunggur Formation, and (2) the late Miocene Baogeda Ula Formation disconformably overlies the Tunggur Formation and can be correlated with the red beds along the southern escarpment. The Tunggur Formation was thus mapped as a thin strip, exposed only in the northern escarpment of the Tableland. The first assumption cannot be substantiated by us (or any previous workers), and it appears to be based on measurements of cross-bedding in channel sands. The second assumption was derived from the stratotype section of the Baogeda Ula Formation and cannot be supported by our own observations. In fact, based on fossil evidence, the southern exposures are lower, not higher, than the northern ones.

Our present correlation scheme (fig. 8) is primarily based on lithological characteristics. Fossils are taken into consideration only in the Tairum Nor section, which is far from other localities and whose lithological characters are sufficiently different from those elsewhere to make direct lithological correlation difficult. As has been previously summarized (Z.-d. Qiu, 1996; Z.-x. Qiu et al., 1988), exposures along the northern and western escarpments generally consist of a package of channel sandstones (often containing numerous bivalve shell fragments) in the lower part (except the Aletexire section, see below), followed in the middle by mudstones of mostly red or brown, and finish at the top by thin layers of greenish or light gray marls interbedded with greenish or variegated mudstones. Generally less than 40 m in total thickness, this package is here informally termed the upper beds of Tunggur Formation (fig. 8). Representing a fining upward sequence, the occurrence of marls near the top, in contrast to abundant channel sands below, seems to indicate a more arid climate. Although serving well as a generalized model of depositional sequence, such a simplified scheme, however, cannot be applied in every section, mostly due to the lenticular nature of the channel sandstones and marls.

The lower beds of the Tunggur Formation, here defined by the lower section of the Aletexire locality (fig. 8) and much of the Tairum Nor section, have traditionally received much less attention because of their lack of abundant fossils. Lithologically, the lower beds are relatively simple and, as typified by the Tairum Nor section, consist of a rather uniform package of red or orange-red mudstone interrupted in the middle by a channel sandstone. Our correlation of the Tairum Nor section with the lower part of the Aletexire section remains speculative. The long distance and lack of intermediate exposures aside, the more mature paleosols at Aletexire have a different appearance from the presumed counterpart at the Tairum Nor section (upper red bed). An obvious implication of such a correlation is that the upper beds of the Tunggur Formation were either eroded away or were never deposited at the southern part of the original Tunggur depositional basin.

MAGNETOSTRATIGRAPHY

Two sections were chosen for paleomagnetic analysis, Moergen and Tairum Nor (346 locality). The Moergen section, sampled in 1995 field season, was selected for its richness in small mammals and the thickness of exposures along the northern margin of the Tunggur Tableland. The main section is located near the western end of the east–westerly-oriented exposure at the Moergen locality, beginning at 43°44′55.0″N, 112°54′02.5″E and ending at 43°44′37.6″N, 112°54′04.3″E. To avoid several meters of loose, coarse-grained sands between two prominent marls, the middle section was laterally shifted 600 m to the east (at 43°44′33.5″N, 112°54′31.3″E), where the exposures consist of fine-grained mudstones. Three segments of various lengths have to be omitted in the sampling due to the presence of a few cross-bedded sands in the lower half of the section.

In contrast to the more varied lithologies in the Moergen section, the section at Tairum Nor was chosen because of its more uniform rock composition of mostly brick-red mudstone with a single layer of cross-bedded sands (less than 3 m in thickness) in the middle. In addition, our discoveries of a small mammal fauna, the Tairum Nor Fauna, which seems to indicate a somewhat older age, constitute an additional reason to sample this locality. Furthermore, some important fossils, such as the holotypes of Platybelodon grangeri, Bellatona forsythmajori, and Distylomys tedfordi were collected from the southern exposures. It is therefore important to learn about the paleomagnetics at the 346 locality, not only as a tool of local correlation but also to place these taxa in chronological perspective. Collected in 1996 field season, the 346 section begins at 43°24′48.3″N, 113°06′57.6″E and ends at 43°24′55.9″N, 113°06′57.8″E.

Oriented hand samples were taken from the Moergen section in 1995 and from the Tairum Nor section in 1996. Sample orientation was measured by geologic compass. Three independent samples were taken in each sampled horizon (sites). All samples were taken from fine-grained mudstones or siltstones in 1-m intervals, except where sediments are too coarse to permit sampling. A total of 90 samples were collected from the Moergen section and 102 from the Tairum Nor section.

The samples were cut into 2 X 2 X 2-cm cubes. The direction of magnetization was measured on a three-axis cryogenic magnetometer in the paleomagnetic laboratory at the University of Florida. All samples were thermally demagnetized in increments of 13–14 steps. Representative plots of thermal demagnetization are shown in figure 9. The demagnetization reveals two magnetic components, a low temperature component that is removed by 100–200°C, whereafter characteristic directions become clearly revealed in trajectories to the origin at temperatures of 670–1000°C.

Seven magnetozones are recognized in Moergen section and six in Tairum Nor section (fig. 10). Correlations with the geomagnetic polarity time scale (Berggren et al., 1995) are difficult because of the relatively short sections and the frequent reversals during the Miocene. Sampling gaps in coarse sandstones in the Moergen section further compromises the reliability of the correlation.

As discussed in the Biochronology section below, the Tunggur faunas are clearly middle Miocene in nature when compared to mammalian faunas from Europe and North America. Within the middle Miocene between 11–16 Ma, the lower and upper parts of the geomagnetic polarity time scale (Berggren et al., 1995) are dominated by relatively long reversed (C5Br and C5r) and normal (C5Bn through C5ACn) magnetozones, in contrast to a distribution of shorter and more temporally equivalent normal and reversed magnetozones in the Tunggur sequence. This leaves C5An.1n through C5ABr as the most suitable segment for correlation. Within this segment, we correlated the composite Tunggur section with the C5Ar.3r through lower part of C5r.3r in approximately 11.8–13 Ma, younger than we had estimated earlier (X. Wang et al., 1996). Such a correlation is consistent with the small mammals, which are comparable to those from European land mammal zones (MN) 7–8 in the late Astaracian (Z.-d. Qiu, 1996), as well as to a number of large carnivorans (see Biochronology below). However, we emphasize the preliminary nature of our magnetic study, which must be tested in additional sections. Based on this correlation, sedimentation rate for the Moergen locality is 63 mm/1000 years (44 m/0.7 m.y.) and for the Tairum Nor locality is 47 mm/1000 years (35 m/0.75 m.y.), with the latter having far fewer channel sandstones than does the former.

BIOSTRATIGRAPHY

Due to their relatively higher densities in fossil records as well as modern collection methods that provide more precise control of horizons, small mammals are ideally suited for fine-scale regional stratigraphic correlations. Z.-d. Qiu (1996) first discerned some faunal changes within the small mammal sequence in the Moergen (MO) section. At the top of the section (horizon MOV, see fig. 8), small mammals begin to show compositional and morphological differences from the main small mammal quarry (horizon MOII), such as the occurrence of Ansomys? and the more advanced status of Gobicricitodon robustus and Plesiodipus progressus. It is now apparent that the absence of Ansomys in Moergen level II is likely due to sampling biases since this aplodontid is also recorded in the lower red bed of Tairum Nor (see below). Based on the magnetic stratigraphy above, these two horizons are approximately 0.5 m.y. apart (fig. 10). However, Z.-d. Qiu did not attempt to establish a new fauna for horizon MOV.

Subsequently, we (Z.-d. Qiu and Wang, 1999) established a new Tairum Nor Fauna to be distinguished from the Moergen Fauna (“Tunggur Fauna” in that paper) from the northern and western edges of the Tunggur Tableland (except the lower part of the Aletexire section), again based on small mammal records only. The Tairum Nor Fauna is here defined on mammals from the lower red bed and middle sandstones along exposures in Tairum Nor. Here we present complete faunal lists of all localities for these two faunas (table 1), part of which is an update published by Z.-x. Qiu et al. (1988).

Faunal distinctions are recognizable within sections where small mammals are finely sampled. As discussed above, faunal composition and evolutionary changes of individual lineages are observed along the Moergen section (Z.-d. Qiu, 1996). In large mammals, however, since most of them belong to early collections that have inadequate data for precise stratigraphic horizons, further stratigraphic resolution within many of the classic localities is, regrettably, still not possible. Nonetheless, in the case of a recently recognized long section—the Aletexire locality that has the longest exposure (approximately 80 m) among all Tunggur localities—recent attention to finer scale collecting has resulted in hints that further divisions are indeed possible. Among others, a new and more primitive form of Stephanocemas appears to be present in the lower beds of this section (under study).

The Tairum Nor locality is another section finely sampled that yields different faunas. The most prominent difference in the small mammals is the presence of Tachyoryctoides in the lower red bed of Tairum Nor, which, together with the middle sandstone, forms the basis of the Tairum Nor Fauna. This ctenodactylid rodent is commonly found in the Oligocene and early Miocene of Asia, and its previous latest occurrence is in the early Tunggurian Dingjiaergou local fauna of Tongxin, Ningxia Hui Autonomous Region (Z.-x. Qiu et al., 1999). Its appearance in Tairum Nor thus becomes the latest occurrence of the genus and extends its range into the late middle Miocene.

Compositional differences of large mammals are also observable in the Tairum Nor localities. Several carnivorans are apparently unique in the middle sandstones of the Tairum Nor section: Leptarctus neimenguensis, a new form of Sthenictis, and a new musteloid possibly related to Aelurocyon. Their absence in the classic localities along northern and western edges of the Tableland is conspicuous and cannot be explained in terms of depositional environments—all are from the prominent channel sandstones between the upper and lower red beds, the kind of deposits well represented elsewhere in the Tunggur Tableland. This lends further evidence that the southern section is older than most of the northern and western localities.

BIOCHRONOLOGY

Although the Tunggur strata have historically been considered Pliocene (Spock, 1929) and late Miocene (Colbert, 1939b; Wood, 1936) following the then-prevalent conventions, that the Tunggur fauna is middle Miocene has not been seriously doubted by recent authors (e.g., Li et al., 1984; Z.-x. Qiu, 1990; Z.-x. Qiu and Qiu, 1995; Z.-x. Qiu et al., 1999). Lack of the three-toed horse Hipparion in Tunggur clearly indicates an age older than early late Miocene Vallesian when Hipparion first appeared in Eurasia. On the other hand, Tunggur fauna cannot be early Miocene given the advanced status of its large mammals (e.g., Chalicotherium, Listriodon, Stephanocemas, Anchitherium) and numerous small mammals shared with the European middle Miocene (Z.-d. Qiu, 1996).

Chinese stratigraphic synthesis in recent years has tended to compare the Tunggur local fauna with the European Neogene Mammal Zone 8 (Z.-x. Qiu, 1990; Z.-x. Qiu and Qiu, 1990, 1995) or to equate it to an undivided European MN 7+8 (Z.-d. Qiu, 1996; Z.-x. Qiu et al., 1999), as is the custom in recent compilation of European MN zones (Agustí et al., 2001; Mein, 1999; Sen, 1997; Steininger et al., 1996; Steininger, 1999).

Since recent attention has been mostly focused on long-distance correlations with European faunas, it may be informative to look toward North America for clues of age relationships, as the Tunggur faunas also have some taxa in common with North America. Among large mammals, the following carnivorans are particularly instructive in their first appearances in East Asia and/or in their well-constrained phylogenetic sequences in the continents of their origins.

Leptarctus neimenguensis Zhai, 1964 is most readily identifiable with North American relatives because of its combination of highly unique cranial and dental morphologies. Strong parasagittal crests and a ventral bullar process combined with a highly hypocarnivorous dentition unambiguously link it to its New World relatives. North American leptarctines have a continuous record from early Hemingfordian to late Hemphillian, and nearly a dozen species have been recognized within this period (see recent summaries by Lim, 1999; Lim and Martin, 2001; Lim et al., 2001; Lim and Miao, 2000). Dental evolution in the leptarctines mainly involved increasing size and hypocarnivory, such as transverse widening of P4–M1, increasing size of P4 hypocone and M1 metaconule, and development of a parastyle on P4. One can divide known species of Leptarctus into primitive and advanced groups based on the above characters. The primitive group includes L. wortmani, L. ancipidens, L. primus, and L. oregonensis from the early Hemingfordian to early Barstovian, except L. primus which has a longer range into the Clarendonian (Lim, 1999; Lim et al., 2001). Generally they are smaller, lack a P4 parastyle, and have narrow P4–M1 and small hypocones. The advanced group, on the other hand, includes L (Hypsoparia) bozemanensis, L. martini, L. kansasensis, L. progressus, and L. supremus from late Barstovian through late Hemphillian, and commonly feature a larger size, possess a P4 parastyle of various sizes, and have wide P4–M1 and large hypocones. L. neimenguensis appears to be a transitional form between the primitive and advanced groups. It is about the same size as L. primus and developed a rudimentary P4 parastyle and M1 metaconule. In overall dental morphology, it is best comparable to materials from the late Barstovian to early Clarendonian of Nebraska, for example, the Norden Bridge Quarry (e.g., F:AM 103337) in the Cornell Dam Member and the Burge Quarry in the Burge Member of Valentine Formation. If the above morphological assessment is accurate, the closest North American relative of L. neimenguensis is expected to fall within a narrow range of 12.5–13 Ma (based on chronology in Tedford et al., 1987), which is consistent with the magnetostratigraphic correlation above (channel sandstones in the Tairum Nor section in fig. 10).

Similar to the case of Leptarctus, a second taxon, Sthenictis, n. sp. (under study), whose relatives are found in North America, is also found in the channel sandstones of the Tairum Nor section. Sthenictis is widely known in the Hemingfordian through Clarendonian of the Great Plains (Matthew, 1924; Matthew and Gidley, 1904; Peterson, 1910). Although the North American forms generally have wider cheek teeth, the Tunggur Sthenictis is best comparable to a specimen from the Burge Quarry (F:AM 25235, an undescribed species) in size and overall morphology, again suggesting an age equivalent to the Barstovian–Clarendonian transition in North America.

Two other large carnivorans are also instructive in age relationships. Although no new material was found for the extremely rare Sansanosmilus (Machairodus sp. in Colbert, 1939a), which is represented by a single well-worn upper carnassial from the Platybelodon Quarry (“5 miles southwest of Wolf Camp”), this P4 is quite informative relative to the evolution of this barbourofeline carnivoran. The Tunggur Sansanosmilus features an upper carnassial with a well-developed parastyle and preparastyle and loss of protocone. Such a combination of features is comparable to that of S. jourdani from La Grive-Saint Alban in France and equivalent beds in Europe (Geraads and Güleç, 1997). A direct descendant from the Old World Sansanosmilus is the North American Barbourofelis (Baskin, 1981; Bryant, 1991; Geraads and Güleç, 1997). The Tunggur Sansanosmilus is comparable in stage of evolution to that of early Clarendonian B. whitfordi, the earliest and most primitive species in North America, but it is less derived than the late Clarendonian B. morrisi from Leptarctus Quarry in Nebraska (e.g., F:AM 79999, holotype) in its less differentiated P4 preparastyle and less reduced P4 protocone root (Geraads and Güleç, 1997; Schultz et al., 1970). The Tunggur Sansanosmilus seems to offer the closest example of an ancestral form that gave rise to the New World barbourofelines, and, as such, may slightly predate the North American early Clarendonian first occurrence. A similar comparison can also be made for the Tunggur hemicyonine ursid Plithocyon teilhardi (Colbert, 1939a) and its North American counterparts. The latter represents another immigrant event in the early Barstovian (Tedford et al., 1987). Overall, the Tunggur materials are best compared to those from the Burge Quarry (e.g., F:AM 54207) in size and dental morphology.

It thus appears that at least four large carnivorans from Tunggur are either closely related to a North American lineage (Leptarctus and Sthenictis) or have a comparable stage of evolution in their North American counterparts (Sansanosmilus/Barbourofelis and Plithocyon), all within the narrow range of approximately 12–13 Ma in the latest Barstovian to earliest Clarendonian.

DEPOSITIONAL ENVIRONMENTS

Sediments in the Tunggur Formation have traditionally been interpreted to represent lake or lakeshore deposits with occasional river channels. A large, continuous water body was envisioned by Spock (1929: fig. 6) and Andrews (1932). Such a lacustrine origin of basin history, however, lacks sedimentological support. In general, there are three types of sediments in the Tunggur Formation: cross-bedded sandstones or unconsolidated sands without clear bedding structures, mudstones of various colors with little or no bedding structures and sometimes with calcareous concretions, and freshwater marls. Only the marl may be related to still-water bodies, but these marls are few, quite thin, and usually lenslike in a scale of less than a few hundred meters in lateral extent. There are never more than two marls in a section and their thickness is usually less than one meter, often much less. Furthermore, the marls often taper off laterally within a single section (e.g., at Wolf Camp) and are consistent with a depositional environment of small, high alkaline ponds, as occur in the present day grassland of the region.

Most rocks, however, are sandstones from channel activities or overbank deposits, or brick-red mudstones that have undergone various degree of paleosol development. Channels are present in Tunggur Formation in nearly all sections and tend to form prominent benches that are resistant to erosions. Lenses containing bivalves are always associated with channel sands and are concentrated by postmortem transportation. Individual sets of cross-beds are commonly less than one meter in thickness, indicating limited water depth in the channels. A single channel in the southern escarpment is the most prominent feature of the Tairum Nor exposures. This paleochannel essentially flows along the direction of the Tairum Nor escarpment and apparently controls the appearance of the exposures. Flow direction, as observed on the cross-beddings, is toward the northeast, consistent with the modern topography of the Tableland which shows a generally northeastern dip. Sandwiched between red beds, the channel sandstone is nearly continuous for about 10 km (toward the northeastern end, the channel can be seen to taper off 5.2 km east of the 346 roadmark at 43°27′27.6″N, 113°08′59.5″E). Meandering loops of the channel can be observed in several locations near the 346 roadmark. Here, a single loop can have a radius of 250 m. Approximately 1 km west of the 346 roadmark (43°24′46.7″N, 113°06′31.8″E) the channel banks are exposed, where the channel loops into the outcrop and reemerges a few hundred meters farther west.

Multiple bands of paleosols can be observed in many localities. Distinct bandings of dark and light red mudstones are seen along the southern escarpments, both in upper and lower red beds, and they are also clearly seen in various northern exposures (they are especially clear in the lower section of Aletexire locality). Variable concentrations of calcareous nodules and sometimes root traces are often associated with the paleosols, although some mudstones are quite pure. Vertebrate fossils are sometimes embedded within the nodules, which help to protect the bones. These dark red mudstones with calcareous nodular horizons seem to indicate deposition in a warm semiarid seasonal climate (Hunt et al., 1999).

Unpublished pollen analysis by Dong (1993) at his “PM-08” locality (approximately 44°53′30″N, 113°30′30″E) at the northeast end of the Tunggur Tableland suggests a predominance of trees (54.7–84.1% in total counts of pollen grains) of mixed conifer and broadleaf types. Conifer genera include Abies, Picea, Tsuga, Keteleeria, Cadrus, and Pinus, and broadleaf genera include Betula, Alnus, Ulmus, Zelkova, Celtis, Juglans, Pterocarya, Carya, and Quercus. Shrubs and grasses consist of 5.7–22.6% of grain counts and include taxa such as Corylus, Ephedra, Artemisia, Polygonum, Potamogetox, Caprifoliaceae, Ranunculaceae, Compositae, Chenopodiaceae, Gramineae, and Umbelliferae. Dong considered the flora of the Tunggur Formation to be characterized by a predominantly mixed forest with small areas of shrubs/grassland.

On the other hand, Z.-d. Qiu's (1996) study of small mammals suggested a dominance of grassland-adapted species such as hedgehogs (Mioechinus), jerboas (dipodids), and picas (ochotonids), and high-crowned, lophodontid cricetids (e.g., Gobicricetodon and Plesiodipus). Transitional areas do seem to exist as indicated by such taxa as Eutamias, Sinotamias, Microdyromys, Miodyromys, and Heterosminthus. The aquatic habitat was occupied by beavers (Anchitheriomys and “Monosaulax”), cyprinid fishes, and bivalves. Furthermore, Z.-d. Qiu (1996) suggested a rather dry environment of arid grassland or even semidesert that are associated with such forms as Atlantoxerus, Protalactaga, and numerous ochotonids, comparable with other north China Tunggurian faunas from the Lanzhou, Tongxin, and Xining basins.

Large mammals, on the other hand, consist of mostly low-crowned browsers, such as Stephanocemas, Listriodon, Chalicotherium, Anchitherium, Palaeotragus, Platybelodon, and Serridentinus. There are, however, moderately high-crowned forms, such as Turcoceros and rhinos. This is in sharp contrasts to the high-crowned ungulates in the late Miocene Baodean faunas. Colbert (1939b: 7) envisioned the Tunggur environmental setting as follows: “It would seem probable that the Tung Gur fauna is actually a borderland assemblage, containing forest forms and plains forms intermingled. Evidently these mammals were living on a broad flood-plain, traversed by numerous tree-bordered rivers.”

In summary, the following types of habitats were present in the Tunggur region as indicated by various lines of evidences: mixed forests of conifers and broadleaves (pollens and mammals), shrub or grass plains (pollens and mammals), shallow, meandering rivers (channels, aquatic mammals, fishes, and bivalves), and high-alkaline ponds during periods of dry climates (marls). Such a mixture of diverse habitats is quite different from the present day eastern Inner Mongolia, which is dominated by short-grass prairies, deserts, sand dunes, and high-alkaline ponds under a very dry climate.