Ushbulak—A New Stratified Upper Paleolithic Site in Northeastern Kazakhstan

А.А. Anoikin1, G.D. Pavlenok1, V.M. Kharevich1, Z.K. Taimagambetov2, A.V. Shalagina1, S.A. Gladyshev1, V.A. Ulyanov1, 3, R.S. Duvanbekov2, and M.V. Shunkov1 1Institute of Archaeology and Ethnography, Siberian Branch, Russian Academy of Sciences, Pr. Akademika Lavrentieva 17, Novosibirsk, 630090, Russia E-mail: anui1@yandex.ru; lukianovagalina@yandex.ru; mihalich84@mail.ru; aliona.shalagina@yandex.ru; gladyshev57@gmail.com; v.a.ulyanov@yandex.ru; shunkov77@gmail.com 2National Museum of the Republic of Kazakhstan, Pr. Tauelsyzdyk 54, Nur-Sultan, Republic of Kazakhstan E-mail: zhaken.taimagambetov@gmail.com; nms_muzei@mail.ru 3Lomonosov Moscow State University, Leninskie Gory 1, Moscow, 119991, Russia

Asia. This region is characterized by specifi c geographic conditions under which an extreme continental and arid climate favors erosion over the accumulation of sediments, which, in turn, makes the detection of stratigraphically intact Paleolithic archaeological sites very diffi cult. In Kazakhstan, such sites are thus very few and the period is mostly represented by surface fi nds. This scenario applies not only to the early stages of the Paleolithic, but also to the Upper Paleolithic, a period when population density increased sevenfold and humans settled in all regions of the continent, including the extreme North (Pitulko et al., 2012). Upper Paleolithic artifacts have been recorded in situ at several localities in Kazakhstan such as Maibulak and Chokan Valikhanov (Taimagambetov, Ozhereliev, 2008;Fitzsimmons et al., 2017), in southern Kazakhstan. In central and northern Kazakhstan, several stratified Upper Paleolithic sites  are known. However, these either remain in the early stages of excavation, or their stratigraphic position indicates cultural heterogeneity (Merz, 1990;Taimagambetov, Ozhereliev, 2009).
Reconnaissance conducted by the joint Russian-Kazakhstan Expedition in 2016 in the Shilikty Valley, northeastern Kazakhstan, revealed a new stratifi ed site of Ushbulak, whose archaeological assemblages represent various stages of the Upper Paleolithic (Shunkov et al., 2016a(Shunkov et al., , 2016bShunkov et al., 2017).

Site description
The Ushbulak site is located in the eastern part of the Shilikty Valley (Zaisansky District of the Eastern Kazakhstan Region) (Fig. 1). The valley is approximately 80 km long and 30 km wide. The transverse profi le of this intermontane depression is roughly symmetric, while the longitudinal profi le is asymmetric. The valley is surrounded by mountain ridges: Manyrak to the north, Saur to the east, and Tarbagatai to the south and west.
In the course of reconnaissance conducted by the Russian-Kazakhstan Expedition in 2016 in Ushbulak, in the upstream area of Vostochny creek (1500 m asl), numerous Upper Paleolithic artifacts (approximately 1.5 thousand specimens) were collected from the waterway's channel. Excavations further revealed several stratifi ed archaeological complexes attributable to the Initial Upper Paleolithic through the Metal Ages.
In 2016, on the left bank of the Vostochny, near its source, a trench was laid down perpendicular to the slope, and several test pits were made on both banks downstream ( Fig. 2) (Shunkov et al., 2016b(Shunkov et al., , 2016c. In 2017, we conducted two excavations totaling 10.5 m 2 in the upper and lower portions of the trench . Roughly eight lithological strata, including seven layers with cultural material, were identifi ed in the sediment profi le, with a total depth of approximately 7 m.

Stratigraphy
The composite stratigraphic profi le represents sequences recorded in excavations 1 and 2 (Fig. 3, A). Excavation 1 is located near a baulk on the left bank of Vostochny creek, its depth being 3.5 m (Fig. 3, B). Excavation 2 was laid down next to the narrow side of excavation 1, near the base of the left side, and was excavated to a depth of 2.7 m, 1.2 m below the water level (Fig. 3, C) .
The following sediments were recorded in the section (from top down): Stratum 1. A humic horizon of modern soil 0.15-0.20 m thick, with a horizon of blackish-brown sandy loam 0.20-0.25 m thick. Roughly ninety percent of the area is damaged by rodent activity.
Stratum 2. Light gray sandy loam abounding in grus (angular, coarse-grained fragments) and gravel. Roughly eighty percent of the area has been disturbed by rodent activity. The stratum is composed of three horizons corresponding to different dynamic phases of deluvial and proluvial processes. Thickness, 1.0-1.2 m.
Stratum 3. Light sandy and clay loams, pale yellow and grayish-brown in color, with grus and sand. The stratum is composed of three horizons corresponding to different dynamic phases of deluvial and proluvial processes. Thickness, 1.2-1.4 m.
Stratum 4. Fine-grained sand and ochroid and yellowish-brown sandy loam overlying a thin layer of fi ne gravel and grus mixed with sandy loam. The stratum contains two horizons of similar proluvial sediments formed by a temporary stream in the area of active sedimentation. Thickness, 0.2-0.5 m.
Stratum 5. Heavy light gray sandy loam abounding in grus, proluvial-slopewash. The stratum contains two horizons. The lower of the two distinguished by a signifi cantly higher frequency of iron staining. Thickness, 0.4-0.6 m.  Stratum 6. On the basis of changes of lithology and the distribution of artifacts, this stratum was subdivided into eight horizons falling in two sequences. The upper sequence (horizons 6.1-6.5) consists mostly of proluvium/ slopewash. It is composed of a heavy, gray sandy loam with admixture of grus. The lower portion of the sediment displays thin and short lens-like inclusions of light, humic clay loam, blackishbrown in color. The upper sequence is 0.4-0.5 m thick. The lower sequence (horizons 6.6-6.8) is formed by alluvial sediments from a shallow, slow-running stream with a relatively stable hydrological regime and low channel erosional activity. The sequence consists of gray clay loam, which becomes plastic when moist. Its lower portion contains lenses and thin layers of coarsegrained sand, reddish-ochroid in color. The lower sequence is 0.3-0.4 m thick.
Strata 7 and 8 are represented by a sequence of coarse proluvial debris.
Stratum 7. Gravel and grus mixed with porous sand and clay loam infi ll, reddish and brown in color. The gravel particles are large or medium sized, randomly oriented, and both densely and regularly distributed. Based on changes in the clay-rich component in the infi ll and the larger size of gravel particles in the bottom section of the stratum, we established two horizons in this layer. Thickness, 0.3-0.5 m.
Stratum 8. Multicolored sediment composed of grus and gravel, with sporadic boulders and an infi ll of sand and clay loam. Coarse gravel prevails. Based on an increase of the clay component in matrix and a color change in this feature, the increase of gravel particle size, and the appearance of boulders towards the bottom of the layer, this stratum is separated into two horizons. Excavated thickness, 0.8 m.

Scientifi c analyses
AMS-dating. Two radiocarbon dates are available for stratum 6. One of them is 36,180 ± 730 BP (AA-111921: 42,100-39,364 cal BP at 95.4 %; date modeled in OxCal v. 4.3.2, using IntCal13 atmospheric curve). This was generated on charcoal from the middle portion of the stratum (horizon 6.5) by the University of Arizona AMS Facility (Tucson, USA). The other one was released by the Center of Cenozoic Geochronology (Novosibirsk, Judging by the analysis of artifacts from the upper complex (strata 1-4), tools were made of various rocks, mostly low-quality, from sources in the immediate vicinity of the site. These consisted of effusive rocks, slates, quartzite, granitoids, coarse-grained sandstone, and aleurite. Siliceous rocks form less than 30 % of the assemblage from these layers.
The petrographic analysis of artifacts indicates marked differences in the criteria used for selecting raw material between the early and later habitation stages. In the assemblage from the lower complex (strata 5.2-7), artifacts made of local siliceous rocks (chert) of high quality form the largest share of objects (95 %). In some cases, silicifi ed alevrolite and tuff were also utilized. Judging by the cortical surfaces of artifacts, most of the items in this layer were fashioned from larger nodules or, less frequently, on pebbles. Modern exposures of similar siliceous rocks represented by large boulders are located 10 km from the site, in the Ak-Su River gorge, running along the southern slope of the Saur ridge. Siliceous pebbles can also be found in the channels of the Chagan-Obo and Uidene rivers, 8-10 km from the site.
Fauna. In 2016-2017, over 300 unidentifiable fragments of bones belonging to middle-sized ungulates (horse/argali size) were collected from various strata of the site. Most fragments are from 1-2 to 2-5 cm long. Rare identifi able remains, represented primarily by teeth or their fragments, were found in strata 2, 3, 6, and 7. Species composition is similar in all the strata. The assemblage includes argali (Ovis ammon), Siberian ibex (Capra sibirica), and kulan (Equus hemionus). Undifferentiated equid remains (Equus sp.) probably also belong to the kulan. No remains of small mammals were found (Shunkov et al., 2016c;Anoikin et al., 2017;. Note. Percentages, indicated in parentheses, were calculated only for assemblages represented by statistically signifi cant samples. The share of each category is % from the total number of well-represented types (without fragments, shatters or chips). The share of each group within a category is % from the total number of artifacts of the respective category.

Archaeological remains
Within the site's stratigraphic sequence, we identifi ed two complexes of artifacts: the upper (strata 1-4) and lower (strata 5.2-7) assemblages (Fig. 3, A). Lithic artifacts from the upper portion total 177 specimens, including debitage pieces-fragments and shatters (Table 1). The majority of artifacts (8153 spec.) were found in the lower portion of the sequence (Table 1).
Stratum 1 contained 28 lithic artifacts, most of them debitage pieces (16 fragments and pieces of shatter, and 12 fl akes), 12 potsherds dating to the Metal Ages (one of them bears incised horizontal lines), and 28 bones of Holocene animals. The lithic industry from strata 2 and 3 (49 and 18 spec., respectively) includes a narrow-fronted core for microblades and bladelets (Table 2). In terms of large detached products, fl akes dominate the assemblage, while the proportion of blades is insignificant. The presence of microblades and bladelets is indicative of the Final Upper Paleolithic. Typological criteria (fl ake, blade with discontinuous lateral retouch, and fl ake with unifacial retouch) also support this attribution (Table 3). Lithic artifacts from stratum 4 (82 spec.) are concentrated mostly in the upper portion of the sediment. Cores consist of parallel cores with a wide flaking surface, and single-platform cores with one fl aking surface for triangular products and bladelets (see Table 2). Flakes are most numerous among the category of detached pieces. The only tool found in the assemblage is a tablet-like implement with a pointed tip formed through discontinuous stepped retouch (see Table 3). Based on stratigraphic position, stratum 4 would appear to be earlier Table 2. Core-like pieces from Ushbulak, spec.   than strata 2 and 3; however, owing to the paucity of fi nds in this level, it is impossible to assess their chronological attribution with higher precision. The Paleolithic assemblage from strata 5.2-7 is represented by materials spanning the complete technological cycle of fl int knapping, including tested nodules, cores, blanks, core-trimming elements, waste, and fi nished tools (see Table 1).
The principal reduction technique employed in this assemblage was the detachment of blades from parallel and subparallel bidirectional volumetric cores. Twothirds of cores are of this type (see Table 2). This category consist of double-platform, subprismatic nuclei with one fl aking surface for bidirectional reduction (Fig. 4,  1-3, 5), including those with a wide fl aking surface and opposing striking-platforms oriented at different angles (semi-tourné). Some single-platform and single-fronted cores in the fi nal stage of reduction could also have been used for bidirectional knapping at earlier stages. Three morphologically distinct cores exhibit knapping from the narrow face. All display the same reduction technique, which included shaping and rejuvenation of strikingplatforms from the fl aking surface. This normally resulted in ridged or half-ridged fl akes, which ensured regularity, convexity of the fl aking surface, and standardization of cores. Some rejected cores were used as hammerstones, as evidenced by zones of microfl aking on their lateral sides.
Blades, including bladelets and microblades, constitute the most representative category of artifacts among the detached pieces (see Table 1). Microblades were probably removed from small cores or core-burins. However, most microblades in the assemblage represent unintended debitage that, most likely, resulted from preparation of working edges of cores. About 80 % of elongated fl akes demonstrate traces of longitudinal and bilongitudinal faceting of dorsal surfaces (in equal proportions).
Most fl akes have longitudinal faceting on dorsal faces (52 %) and plane striking-platforms (56 %). Half of the recovered fl akes show no traces of preparation on the fl aking zones, while the other half were prepared using the same techniques employed for detachment of blades. Correlation analysis of cores suggests that most fl akes were byproducts of core trimming, i.e., preparation and rejuvenation of striking-platforms, lateral or initial ridges. The fact that fl akes with some remnant cortex on their dorsal surfaces are more numerous than blades with a similar feature, also underscores the technical character of the former.
Byproducts of core trimming in this assemblage (see Table 1) include cortical and secondary fl akes, as well as ridged and half-ridged flakes. A small series of rejuvenation core tablets were also found. Plunging fl akes, which are the outcomes of unsuccessful reduction whereby the core base was lost (usually together with the opposite platform), can also be tentatively attributed to the category of byproducts. The lack of large cortical fl akes suggests that most cores were reduced elsewhere, i.e. outside the excavated portion of the site.
Comparison between striking-platforms on blades and on fl akes provides important information. Frequencies of platform overhang and striking point in blades (22 % and 15 %, respectively) and in fl akes (24 % and 17 %, respectively) are virtually the same. Byproducts of core trimming and tool blanks were possibly detached using the same technique, and with similar hammerstones. Without experimentation using local raw material, it is diffi cult to identify the type of hammerstone used in production. However, in experiments reproducing the Note. Percentages, indicated in parentheses, were calculated only for assemblages represented by statistically signifi cant samples. The share of each category and group is % from the total number of typologically distinct items. Initial Upper Paleolithic tools from Mongolia (Tolbor-15, horizons 5-7), similar frequencies resulted from using a hammerstone whose hardness was close to that of the raw material .
The tool kit from strata 5-7 includes 268 implements, most of them fl akes exhibiting irregular or semi-irregular retouch (50 % of all tools) (see Table 3). Typologically distinct tools (133 spec.) demonstrate the developed Upper Paleolithic component (see Table 3). Endscrapers constitute the most numerous category of implements (28 %, hereinafter, percent of all typologically distinct tools). The most frequent are endscrapers on large blades ( Fig. 5, 1, 4, 5), including carinated endscrapers and an endscraper with a trimmed base (Fig. 5, 5), as well as double endscrapers on narrow, medium-sized blades (Fig. 5, 7). Intensively retouched blades (22 %) and implements with retouched notches (13 %) are common fi nds. Truncated and truncated-facetted implements (9 %) (Fig. 5, 2), perforators, and large spur-like tools (9 %) also form a distinct group. Burins with transverse and angular forms are few in number. On angular burins, removal of the burin spall was prepared through retouching of the blank's longitudinal edge (Fig. 5, 10). The assemblage also contains weakly retouched points (Fig. 5, 8). Sidescrapers are scarce and indistinct, and one of these demonstrates bifacial retouch (Fig. 5, 9). Solitary diagnostic artifacts form an important element of the tool kit. This group of tools includes one waisted blade, several implements with ventrally retouched proximal edge including a beveled, heavily retouched point (Fig. 5, 3), an endscraper (Fig. 5, 5), an implement with a sharp tip (Fig. 5, 6), a stemmed endscraper, blades with ventrally retouched distal sections, and blades with distal ends blunted through retouch. The assemblage from stratum 7 contains a burin-core. Two more burin-cores were collected from the surface in close proximity to the excavated area (Fig. 4, 4). Stratum 6 contained a small talc tablet with traces of artifi cial polishing in some places .
The presence of the above mentioned diagnostic tool types and characteristic features of the primary reduction process (such as the absolute predominance of doubleplatform blade cores for bipolar reduction; the prevalence of blades, including those whose length exceeds 20 cm;  Nearly all artifacts collected from the surface can probably be also attributed to this period (see Tables 1-3). Ninety-nine percent of these objects were recovered from within the stream channel. Their nearly-identical technique, tool types, and raw material link them reliably with strata 6 and 7. Additionally, analysis of geomorphological situation nearby the site has demonstrated that the Vostochny in its upper reaches is actively eroding precisely these strata.

Discussion
Finds from the upper strata of Ushbulak are few, so their interpretation and chronological attribution must remain tentative. The assemblage from Holocene sediments of stratum 1 contains a potsherd decorated with incised horizontal lines, which likely dates to the Metal Ages. Based on the presence of microblades and bladelets, strata 2-4 most likely correlate with the late stages of the Upper Paleolithic.
Within Kazakhstan, the Final Paleolithic industries of Ushbulak have the closest parallels in the late complex of Shulbinka (Petrin, Taimagambetov, 2000) and in the upper complex of Angrensor-2 (Taimagambetov, Ozhereliev, 2009). In the Russian Altai, analogs can be found in Late Upper Paleolithic assemblages from Kaminnaya and Iskra caves, as well as from the open-air sites of Ust-Karakol, Ust-Sema, Srostki, and others (Derevianko, Petrin, Zenin et al., 2003;Markin, 2007).
Archaeological remains from the lower strata of Ushbulak are chronologically and typologically related to industries of the Initial Upper Paleolithic. In western Central Asia, the Middle to Upper Paleolithic transition is best represented by the Obi-Rakhmat sequence, which shows an evolving tradition spanning the period 80-35 ka BP (Krivoshapkin, Kuzmin, Jull, 2010;Vandenberghe et al., 2014). In these industries, primary reduction is characterized by mass production of blade blanks, including microblades. The tool kit at Obi-Rakhmat is dominated by retouched blades (including a pointed variant) and sidescrapers. Diagnostic tools include burincores, truncated-facetted implements, and small, heavily retouched points. Burins and endscrapers are relatively rare (Derevianko et al., 2001;Krivoshapkin, 2012).
In the Russian Altai, the formation of the Upper Paleolithic traditions began ca 50 ka BP (Derevianko, 2011;Derevianko, Shunkov, 2004). The earliest manifestation of these traditions was recorded at Denisova Cave. In strata 11.1 and 11.2 in the Eastern Gallery and in the lower portion of stratum 11 in the Central Hall dating to 50-45 ka BP (Douka et al., 2019), semi-volumetric blade cores and Levallois cores were found in association with tools bearing traces of ventral trimming on the proximal ends, beveled points, and numerous personal ornaments made of organic materials and semiprecious stones (Prirodnaya sreda…, 2003;Derevianko, Shunkov, Markin, 2014).
However, the closest parallels to the lower strata of Ushbulak are finds from horizon UP2 of Kara-Bom (44-43 ka BP) (Derevianko et al., 1998;Rybin, 2014). These materials closely match Ushbulak from all relevant criteria, from the raw material and primary reduction technique to the composition of tool kit (which includes highly diagnostic tools) (Rybin, 2014). The main distinction between these two sites is the absence of Levallois technique in Ushbulak assemblage-although some flakes from the lower portion of stratum 7 and from those recovered in surface collection demonstrate certain "Levallois"-like features. The absence of evidence of Levallois technology at Ushbulak can probably be explained by the relatively small size of the excavated area (4.5 m 2 ).
In northern Mongolia, similar industries, dating to 43-35 ka BP, belong to the southern Siberian-Mongolian variant of the Initial Upper Paleolithic (Tolbor-4 and -21, and others) (Derevianko et al., 2007;Rybin, 2014Rybin, , 2015. In these assemblages, primary reduction is characterized by the prevalence of volumetric and semi-volumetric flaking, combined with sporadic use of Levallois technique. In such contexts, mass production of large blades was based on the use of subprismatic doubleplatform cores with single fl aking surfaces and parallel fl aking. Uni-and bidirectional fl at and narrow-fronted cores were less important for blade creation. Upper Paleolithic implements dominate these tool kits, with endscrapers on blades being most numerous. Another important feature in Siberian-Mongolian assemblages is the presence of several diagnostic artifacts, such as burin-cores, beveled points, points with thinned bases, backed point-bladelets, implements with traces of ventral retouching on the distal edge, bifaces, stemmed tools, and personal ornaments (Rybin, 2014).
The channel for the initial eastward spread of Upper Paleolithic traditions to Mongolia and Trans-Baikal region, ca 45 ka BP, was apparently the Russian Altai (Derevianko, Shunkov, Markin, 2014;Rybin, 2014). This dissemination is believed to have followed several routes, one of which passed via the Mongolian Altai and Dzungaria, along the northern boundary of the Gobi Altai and the Great Lakes Depression to the Selenga basin (Rybin, 2014).
In the Dzungarian Basin in northwestern China, Initial Upper Paleolithic elements are known so far only from one artifact assemblage, a surface-collected assemblage from the site of Luotoshi. This material is characterized by combination of Levallois flake production and subprismatic blade technology with utilization of double-platform bipolar and narrow-fronted cores. The tool kit at Luotoshi consists of numerous retouched blades, spur-like implements, sidescrapers, endscrapers, and denticulate and notched tools. It also includes bifaces, beveled points, points with a thinned transverse edge, implements with ventrally retouched distal end, and burin-cores . Finally, this assemblage contains many radial cores, which, alongside diagnostic Middle Paleolithic forms such as sidescrapers and notched-denticulates, suggests a considerable share of Final Middle Paleolithic elements. The presence of Middle Paleolithic sites in this area is further evidenced by fi nds from Tongtiandong Cave, situated 200 km southeast of Luotoshi. The industry from the lower cultural horizon of the cave, AMS-dated to ca 45 ka BP, contains Levallois and radial cores, heavily retouched sidescrapers, and elongated points, including those of the Mousterian variety. This assemblage corresponds to the late stages of the Middle Paleolithic (Xinjiang…, 2018).

Conclusions
Ushbulak, situated between the Russian Altai and Dzungaria, is a stratifi ed Initial Upper Paleolithic site of the southern Siberian-Mongolian variant. Key sites from this period in the Altai (Denisova Cave and Kara-Bom) are situated 400-450 km north of Ushbulak, and the Chinese site of Luotoshi lies 100 km to the southeast. Other known stratifi ed Upper Paleolithic sites in Kazakhstan, dating to 40-30 ka BP (Maibulak and Chokan Valikhanov), are located in the piedmont of the Tian Shan, 800-900 km southwest of Ushbulak. Geographically and typologically, these assemblages resemble those of Kulbulak and Shugnou, and are usually considered to belong with the Upper Paleolithic industries of western Central Asia (Taimagambetov, Ozhereliev, 2009;Ranov, Kolobova, Krivoshapkin, 2012).
The fact that Ushbulak is stratifi ed and its lower units have yielded a rather large samples of lithics makes it the most significant Upper Paleolithic site in eastern Kazakhstan. Future excavations, detailed analysis, and interpretation of fi nds will allow us to reconstruct the principal trends in the evolution of the local Upper Paleolithic, to evaluate and assess its links with specifi c human populations, and to trace their migration routes across Kazakhstan and Central Asia.