Late Palaeolithic cave art and permafrost in the Southern Ural

Shulgan-Tash (also known as Kapova) cave located on the western slope of the Ural Mountains (Russia) is the easternmost European cave art monument of late Palaeolithic age. Radiocarbon dates from cultural layers in the cave suggest an age of about 16.3 to 19.6 ka (cal BP), but dates directly on the paintings were not obtained. In order to constrain the age of this art using an independent method, we performed detailed 230Th-U dating of calcite flowstone underlying and overgrowing the paintings at 22 sites in three halls of the cave. The youngest age for the underlying calcite (i.e., the maximum age of the cave art) is 36.4 ± 0.1 ka, and the oldest overlying calcite (constraining the minimum age of the cave art) is 14.5 ± 0.04 ka. The ca. 21.9 ka-long hiatus in calcite deposition during which the paintings were made is attributed to regional permafrost conditions and sub-zero temperatures inside the cave during Marine Isotope Stage (MIS) 2. This is supported by samples of cryogenic cave calcite, which document seven episodes of freezing and thawing of permafrost associated with stadials and interstadials of MIS 3, respectively.

goal of our study was to reconstruct the environment in the cave during these artistic activities. In our unique approach we studied common speleothems (flowstone, stalagmites) and cryogenic cave calcite (CCC 16 ) to constrain periods of time when liquid water was present in the cave (above-zero temperatures) and when deep cave interiors contained masses of ice (permafrost conditions) respectively.   17 ). The "canvas" ages obtained from paintings in the Domed Hall and Hall of Signs, date to Marine Isotope Stages (MIS) 6 and 5 respectively: 138.6 ± 1.0 ka for painting #6-1 and 96.6 ± 0.5 ka for #12-1 (numeric identifications of paintings, preceded with number sign #, is according to 18 ). Another MIS 5 age was obtained for pre-paint flowstone in the Hall of Chaos (#23-1; 95.7 ± 0.6 ka). Several samples of pre-paint flowstone in the Hall of Chaos indicate that calcite deposition terminated between 36.4 ± 0.1 and 44.1 ± 0.2 ka. Post-paint calcite deposition commenced in the early Holocene (e.g., 10.0 ± 0.5 ka ago for an unnamed red spot and 8.2 ± 0.5 ka ago for painting #15-1) in the Hall of Signs, and at the beginning of the Bølling-Allerød interstadial (14.3-14.5 ka ago; paintings #20-1, -2, -4, and -6) in the Hall of Chaos.
The ages of "canvas" and overgrowth calcite bracket large time windows when the paintings would have been created. The narrowest time window was identified in the Hall of Chaos (Table 1) where the youngest age of "canvas" flowstone is 36.4 ± 0.1 ka (#22-13) and the oldest age of overgrowing flowstone is 14.5 ± 0.06 ka (#20-1, -7). Similar ages were obtained for several other paintings in the Hall of Chaos. The minimum duration of the time window during which the paintings were made in this hall is therefore 21.9 ± 0.06 ka.
Dating speleothems. In order to better constrain past conditions inside the cave we studied two types of speleothems. Formation of "common" speleothems such as flowstone and stalagmites requires the presence of water films and dripping water in the cave. In contrast, cryogenic cave carbonates (CCC) form when cave passages are within the permafrost zone and cave air temperatures are slightly freezing, but overall, permafrost is degrading in response to climate warming 16 . Liquid water from the active layer and epikarst above the cave enters the karst, but forms ice upon reaching the cave chamber. During a long-term warming trend the cave temperature will stabilise at near-zero values, leading to the co-existence of water and ice. As the water in pools on the ice slowly freezes, it becomes progressively enriched in dissolved ions, eventually leading to the precipitation of CCC. 230 Th-U dating of stalagmites and CCC therefore allows the identification of periods of both permafrost-free and permafrost conditions, respectively, in the cave and, by extension, in the study area. To this end, we collected 14 stalagmites from four caves located within 6 km (Shulgan-Tash, Victoria, Grioz, and Kulyurtamak caves 17 Table 1. Selected U and Th concentrations, isotopic activity ratios, and ages of samples from the Hall of Chaos, which provide the best constraint for the age of paintings. More details and additional results can be found in Supplementary Information. All measurements are reported with ±2σ absolute uncertainties. Subscript A denotes activity ratio. *Painting number and names according to. 18

Discussion
The parietal paintings in Shulgan-Tash cave made with red ochre are thought to have broadly similar ages 2,7,9,19 . Our 230 Th-U dating demonstrates that the paintings on the walls of the Hall of Chaos were made between 36.4 ka to 14.5 ka. Four radiocarbon dates reported from the archaeological layer (16.5 to 19.6 ka cal BP) fall within the younger end of this 21.9 ka-long time window, constraining the time of at least some of the artistic activity in the cave to the last glacial termination (LGT), prior to the Bølling-Allerød interstadial (Fig. 2).
Our results indicate that the paintings were made on a dry canvas, i.e. during a depositional hiatus of the flowstone. This hiatus occurred synchronously with breaks in deposition of the stalagmites. Growth stops in such "common" speleothems may be caused by several factors, ranging in scale from local (e.g., changes in the drip site feeding the stalagmite) to intermediate (e.g., affecting the whole cave) to regional (affecting the region as a whole). Because our results were obtained from a number of speleothems collected in four different caves, we rule out local-and intermediate-scale growth controls and attribute the long hiatus during MIS 3 and 2 to cold and dry conditions prevailing in Southern Ural at that time.
The lack of stalagmite growth in Siberian caves during certain periods of the Pleistocene has been used as an indicator of past permafrost conditions 20 . Although this interpretation is warranted for areas that are presently located in the permafrost zone (such as Eastern Siberia), applying it to caves located outside of today's permafrost zone, such as Southern Ural, requires additional justification. This is provided by CCC found in Shulgan-Tash and Victoria caves whose ages indicate that permafrost was temporarily present in the study area during MIS 3. Because CCC form under conditions of degrading permafrost, the spread in CCC ages suggests that the local permafrost was rather unstable during this time interval. At the depth of the studied caves (ca. 50-80 m below the surface), at least seven freeze-thaw cycles are recorded. Permafrost thawing episodes occurred in association with Greenland Interstadials 21 (GI) 16, 15.1, 14, 12a, 11, 7 and 6 (Fig. 2). The relative timing with respect to the GIs varies: some CCC formed at or immediately after the peak warming (e.g., GI 15.1 and 12a); more commonly, however, CCC formed towards the end of interstadials and during the transition into stadials (GI 16, 11, 7 and 6). There might have been more cycles of permafrost freezing and thawing during MIS 3, but not all of them necessarily left traces in the form of CCC in the studied caves. Note that the very thin (<2 mm) pre-paint flowstone in the Hall of Chaos indicates short (1.5-2.5 ka-long) pulses of small-scale seepage flow (i.e. above-zero temperatures in the cave) associated with warmings during GI 8, 9, 10 and possibly GI 12 (Fig. 2).
Our results indicate that the but highly variable climate during MIS 3 was sufficiently cold to support permafrost in Southern Ural. The permafrost dynamically responded (thawed) to short-lived warmings of interstadials, as reflected by formation of CCC during these times. As MIS3 gave way to the significantly colder MIS 2, it is reasonable to assume that pervasive and continuous permafrost developed in the study area and persisted until the major warming at the onset of the Bølling-Allerød interstadial (Fig. 2). Apparently, the MIS 2 temperatures were so low that the short-term warmings associated with GI4 and 3 were not sufficient to cause formation of CCC or even 'common' speleothems in caves.
This means that during the time of artistic activity, i.e., between ca. 36.4 and 14.5 ka, Shulgan-Tash cave was an inhospitable place with below-zero air temperatures throughout the year and no dripping or running water. The paintings were made by Palaeolithic artists on cave walls that were dry 19 ; our data indicate that the dryness was due to the freezing temperatures.
Our palaeoclimate and permafrost reconstruction is supported by data from other regional environmental archives. Studies of mammal assemblages suggest that environmental conditions in the Southern Ural during the Last Glacial Maximum (LGM, 28-20.5 ka) and LGT (20.5-15.5 ka) were characterised by severe cold climate and open periglacial landscapes, i.e. tundra and forest-steppe, with forests largely restricted to river valleys 22 Supplementary Information). Stalagmites in the upper galleries of the cave started growing even later, at 6.4 ± 0.3 ka. The period of no growth, therefore, appears to have been of different lengths in different parts of the cave. This can be explained by the 3D geometry of the cave in relation to the shape of the Saryk-Oskan massif, which controls the dynamics of permafrost thawing. Although the Hall of Chaos belongs to the lower level of the cave, it is located underneath the steep eastern slope of the massif incised by the gorge of the Shulgan brook (Fig. 3). This chamber, therefore, is closer to the surface (25-35 m) than the halls of the upper level of the cave (ca. 45-55 m of rock overburden). The Hall of Signs and Domed Hall are at intermediate positions. Since the degradation of permafrost during climate warming proceeds from the surface downward, it stands to reason that the Hall of Chaos, located closer to the surface, thawed first at ca. 14.5 ka thus allowing water to enter the cave and flowstone to form. Thawing of the thicker rock overburden above the halls and galleries of the upper cave level, where stalagmites were collected, took longer and the water reached this part of the cave only at ca. 6.4 ka. Warming associated with the Bølling-Allerød interstadial was sufficient to thaw the upper 25-35 m of permafrozen rock, allowing water to enter the Hall of Chaos (Fig. 3). Cave passages and halls located deeper thawed later, during the final demise of the permafrost in the Holocene. Similarly, wide-spread stalagmite growth in Victoria cave (at a depth of 75 to 90 m) only commenced at ca. 7.8 ka.
SCIEntIfIC REPORTS | (2018) 8:12080 | DOI:10.1038/s41598-018-30049-w Similar logic explains an apparent inconsistency in Fig. 2, which suggests that at around 41-42 ka CCC and flowstone formed simultaneously. The thin layers of flowstone formed in halls on the lower level of the cave (located closer to the surface), whereas CCC formed on the upper level (underneath a thicker rock overburden). These data therefore demonstrate that cryogenic and "common" speleothem growth can occur simultaneously in different parts of the same cave.
Our study indicates that in the Southern Ural (latitude 53° N) permafrost conditions prevailed during most of MIS 3 and the entirety of MIS 2. Stalagmites and flowstones could not form during these times; their formation resumed on a very local scale only during the Bølling-Allerød interstadial. The onset of wide-spread speleothem formation occurred as late as the mid-Holocene.
Because of wide-spread permafrost, conditions in the Central and Northern Ural and in Siberia were likely even more restrictive with respect to speleothem formation during Late Palaeolithic times. Speleothem growth was possible only during the warmest interglacial periods (cf. 20 ). One exception might be very shallow caves that could have been permafrost-free and allowed local growth of speleothems during short warm intervals, such as interstadials of MIS 3. The Upper Palaeolithic speleothem record therefore is expected to be fragmented (in both spatial and temporal senses) in northern Eurasia, and the potential of 230 Th-U dating of speleothems for improving the chronology of Upper Palaeolithic archaeological cave sites in the Ural Mountains and Siberia therefore appears limited.

Methods
Sampling. Samples of flowstone associated with parietal art were collected either as powders in situ, using a hand-held milling device, or as 8 mm-diameter cores. Stalagmites were collected in those parts of the caves where their removal caused minimal visual impact. Sub-samples for 230 Th-U dating (6 to 200 mg) were milled from the collected samples in the laboratory in a clean air laminar flow hood. Th-233 U-236 U tracer prior to chemical separation of U and Th using an Fe co-precipitation procedure followed by chemical purification on anion exchange columns 27,28 . Procedural chemistry blanks were typically less than 100 at for 230 Th and less than 1 fg for 234 U. U and Th isotopic ratios and concentrations were determined using the latest protocols on a ThermoFisher Neptune MC-ICPMS 28 and the half-lives from 29 . At Max Planck Institute for Chemistry (Germany), chemical separation of U and Th isotopes was performed as described in 30 . U and Th isotopes were analysed using a Nu Plasma MC-ICPMS. Analytical details are described in 31 . Details about the calibration of the mixed U-Th spike are given in 32 . All activity ratios were calculated using the half-lives from 33 , and all ages are reported at the 2σ-level. At Xi'an Jiaotong University (China), analyses followed chemistry procedures described in 27 to separate uranium and thorium. U and Th isotopes were analysed individually by using a ThermoFisher Neptune Plus MC-ICPMS as described in 28,29 . Age corrections assume an initial 230 Th/ 232 Th ratio of 4.4 ± 2.2×10 −6 of bulk Earth 34 .