Late Pleistocene to Holocene vegetation and climate changes in northwestern Chukotka (Far East Russia) deduced from lakes Ilirney and Rauchuagytgyn pollen records

This paper presents two new pollen records and quantitative climate reconstructions from northern Chukotka documenting environmental changes over the last 27.9 ka. Open tundra‐ and steppe‐like habitats dominated between 27.9 and 18.7 cal. ka BP. Betula and Alnus shrubs might have grown in sheltered microhabitats but disappeared after 18.7 cal. ka BP. Although the climate was rather harsh, local herb‐dominated communities supported herbivores as is evident by the presence of coprophilous spores in the sediments. The increase in Salix and Cyperaceae ~16.1 cal. ka BP suggests climate amelioration. Shrub Betula appeared ~15.9 cal. ka BP, and became dominant after ~15.52 cal. ka BP, whilst typical steppe communities drastically reduced. Very high presence of Botryococcus in the Lateglacial sediments reflects widespread shallow habitats, probably due to lake level increase. Shrub Alnus became common after ~13 cal. ka BP reflecting further climate amelioration. Simultaneously, herb communities gradually decreased in the vegetation reaching a minimum ~11.8 cal. ka BP. A gradual decrease of algae remains suggests a reduction of shallow‐water habitats. Shrubby and graminoid tundra was dominant ~11.8–11.1 cal. ka BP, later Salix stands significantly decreased. The forest‐tundra ecotone established in the Early Holocene, shortly after 11.1 cal. ka BP. Low contents of green algae in the Early Holocene sediments likely reflect deeper aquatic conditions. The most favourable climate conditions were between ~10.6 and 7 cal. ka BP. Vegetation became similar to the modern after ~7 cal. ka BP but Pinus pumila came to the Ilirney area at about 1.2 cal. ka BP. It is important to emphasize that the study area provided refugia for Betula and Alnus during MIS 2. It is also notable that our records do not reflect evidence of Younger Dryas cooling, which is inconsistent with some regional environmental records but in good accordance with some others.

Continuous and well-dated palaeoenvironmental records for the Late Pleistocene are rare in arctic Chukotka. Lacustrine sediments are valuable archives as they contain various palynomorphs, which can be used for reconstruction of regional past climate and vegetation changes. Unfortunately, the accumulation rates in large and deep arctic lakes are often very low and these records often do not permit high-resolution reconstructions of vegetation and climate changes (e.g. Melles et al. 2012;Biskaborn et al. 2016). However, such records are required to address the major open palaeoenvironmental questions in arctic Chukotka, especially for the time intervals older than the Holocene.
One of these open questions is the Younger Dryas (YD) cooling in Siberia. This cooling is well reflected in the Lateglacial (c. 15-11.8 cal. ka BP) records from western parts of East Siberia, but in many areas in northeastern Siberia (West Beringia) this cooling seems to be much weaker or even negligible (e.g. Kokorowski et al. 2008a, b;Lozhkin et al. 2018 and references therein). Kokorowski et al. (2008a, b) reported spatially heterogeneous climate conditions during the YD in Beringia: cooling in southern Alaska, East Siberia, and some parts of northeastern Siberia; and no cooling in different locations from a number of sites from northern Siberia and Far East Russia. Moreover, the Late Pleistocene and Holocene vegetation and climate history of the region is of particular interest as detailed reconstructions of the environmental changes, such as tree line dynamics and the identification of the Lateglacial refugia, is important for the reliable reconstruction of  The northern study site, Lake Rauchuagytgyn (67°49 0 N, 168°44 0 E, 619 m a.s.l.), is about 4.35 km long with a maximum water depth of~36 m. Bathymetry varies with a deep main basin giving way to shallower water (<6 m) close to the lake outflow. The lake area is about 6.24 km 2 with a catchment area of about 215 km 2 (Biskaborn et al. 2019a). The lake is located in the northernmost part of the forest-tundra ecotone, in the Anadyr Mountains. The surrounding mountainous terrain has elevations up to 1500 m a.s.l. with a catchment geology dominated almost exclusively by silicicintermediate extrusive volcanic lithologies including tuffs and lavas. Rare outcrops of carbonate sandstone occur within the lake vicinity.
The lake vicinity is dominated by prostrate dwarf shrubs, herb-and graminoid (mostly Dryas octopetala accompanied by different taxa from Poaceae, Fabaceae and Asteraceae) tundra that transits to barren land at higher elevations whilst forest tundra dominates at lower elevations along the river valleys and favourable habitats.
The region is characterized by extremely harsh climate with mean annual air temperature about À11.8°C, mean July temperatures of +13°C and mean January temperatures of À30°C, short growing season (100 days per year) and low annual  Zhuravlev et al. (1999), and inferences by Melles et al. (2005). The bathymetric model (lower panel) was created in Arcmap using inverse-distance based interpolation. [Colour figure can be viewed at www.boreas.dk] precipitation of~200 mm (Menne et al. 2012). The study area belongs to the continuous permafrost zone and evidence for permafrost processes is indicated by the presence of thermokarst lakes and icewedge polygons in the inflow regions at both lake sites.

Coring
Coring was carried out during a joint Russian-German expedition to the Chukotka region in summer 2016. Prior to coring, the water depth was measured from a boat using a handheld echo-sounder. The coring positions were selected at the deepest parts of the lake basins: for Lake Ilirney it was placed at 67°33.7953'N, 168°29.5433'E and for Lake Rauchuagytgyn at 67°78.876'N, 168°73.837'E. Sediment cores were retrieved by a modified hammer UWITEC gravity corer deployed from a catamaran. Precise positioning of the coring device was supported by a cable winch fixed to the catamaran. The sediment coreswere collectedwithin 3 m PVC plastic liners, cut into pieces of maximum 1 m and stored in dark and cool thermoboxes until further analyses at AWI laboratories in Potsdam, Germany. Under clean-laboratory conditions sediment cores were opened, halved and subsampled for pollen analysis every 2 cm.

Radiocarbon chronology
No dateable macrofossil remains were found in either of the analysed cores during subsampling. The Ilirney core (16-KP-01-L02) was 14 C dated using 17 bulk sediment samples (Table 1). Ten samples from 16-KP-01-L02 were initially sent to the Pozna n radiocarbon laboratory (Poland). These samples revealed very low carbon contents (nine samples <1 mg C), accompanied by unrealistically old radiocarbon ages, which may originate from an unsuitable carbon extraction targeting only the oldest carbon fraction. Therefore, these samples were discarded but are listed in Table 1. Eight larger samples from the 16-KP-01-L02 core were subsequently sent to the Mini Carbon Dating System (MICADAS) laboratory at AWI Bremerhaven (Germany) including an additional surface sample (EN18214 0-0.5 cm) retrieved during a follow-up expedition in 2018. To test the integrity of the age model for Lake Ilirney we used the high-resolution age model of core EN18208 from the same lake published by Vyse et al. (2020). As the acoustic profiles show very homogenous sediment patterns, we transferred modelled calibrated ages from EN18208 to the nearby (~250 m) 16-KP-01-L02 core using XRF-derived and additive log ratio transformed elemental K/Ti and Si/ Al data (Vyse et al. 2020). We correlated between the two cores using robust tie-points between distinct peaks (Fig. 3C) and reported the median ages and 2 sigma ranges from EN18208 onto 16-KP-01-L02 for comparison (Fig. 3A).
Pozna n AMS radiocarbon ages were based on the alkali residuals from bulk total organic carbon (TOC) samples. MICADAS ages were also based on bulk TOC following acidification and graphitization. 14 C ages were calibrated using the IntCal13 calibration curve and modelled in the package 'Rbacon' version 2.4.1 in Rstudio, version 1.2.5019 (Fig. 3;Blaauw 2010;Reimer et al. 2013). The Rauchuagytgyn core (16-KP-04-L19) was 14 C dated using eight bulk sediment samples sent to Pozna n including one surface sample of correction for the 'old carbon' effect (Table 1, Fig. 3).

Pollen
A standard HF technique was used for pollen preparation (Berglund & Ralska-Jasiewiczowa 1986). A tablet of Lycopodium marker spores was added to each sample to calculate total pollen and spore concentrations, following Stockmarr (1971). Water-free glycerol was used for sample storage and preparation of the microscopic slides. Pollen and spores were identified at magnifications of 4009 with the aid of published pollen keys and atlases (Kupriyanova & Alyoshina 1972, 1978Bobrov et al. 1983;Reille 1992Reille , 1995Reille , 1998. In addition to pollen and spores, a number of non-pollen palynomorphs (namely fungal spores, remains of algae and invertebrates) were also identified when possible according to van Geel (2001).
At least 250-300 pollen grains were counted in each sample. Only 100 pollen grains were counted in a few samples with extremely low pollen concentration from the Lateglacial part of the Ilirney core. The relative frequencies of pollen taxa were calculated from the sum of the terrestrial pollen taxa. Spore percentages are based on the sum of pollen and spores. The percentages of fungal spores are based on the sum of the pollen and fungal spores, and the percentages of algae are based on the sum of pollen and algae. TGView software version 1.7.16 (Grimm 2004) was used for the calculation of percentages and for drawing the diagrams (Figs 4, 5). The diagrams were zoned by a qualitative inspection of significant changes in pollen associations, pollen concentrations and occurrence of particularly indicative taxa.

Principal component analysis
A principal component analysis (PCA) was applied to the square-root transformed relative proportions for both pollen records using the RDA function from the 'vegan' package version 2.5.6 (Oksanen et al. 2019) in order to portray the major structure in the multivariate data set. The samples are pooled using the same zones as BOREAS those used for the pollen diagrams for both sites. Principal component (PC) 1 and 2 axis scores were extracted and visualized in biplots. For better visibility, only the names of the taxa that explained most variance in the PCAs are presented (Figs 6, 7).

Climate reconstructions methods
Climate reconstructions were based on a modern pollen training data set that was selected from sites within a 2000-km radius around Lake Ilirney (1037 modern sites) from various sources (Whitmore et al. 2005;Davis et al. 2020; Fig. 1). The area from which modern pollen analogues were taken is restricted because same pollen taxa can represent different plant taxa in different regions; however, the area should also be not too small so as to cover a reasonably large gradient (for details see Cao et al. 2017;Herzschuh et al. 2019). Thus, the selected 2000-km radius area covering East Siberia (West Beringia) as well as Alaska (East Beringia), which were connected during the glacial period, is a reasonable compromise. The created data set was taxonomically harmonized with the fossil taxa (Table S1). For all modern pollen sites the corresponding mean July temperature (T July ) and annual precipitation (P ann ) were extracted from WorldClim 2 data (Fick & Hijmans 2017; https://www.worldclim.org/) covering a T July range from 0.9 to 18.7°C and P ann range from 109 to 1058 mm. The climate reconstructions were performed using the modern analogue technique (MAT) transfer function (Overpeck et al. 1985), taking seven analogues in the modern pollen training data set into account by using the MAT-function in the rioja package (version 0.9-21, Juggins 2019) for R (R Development Core Team 2010). In addition, a statistical significance test according to Telford & Birks (2011) was performed for the reconstruction using the randomTF-function in the palaeoSig package (version 2.0-3; Telford 2019). T July and P ann were tested as single variables. Furthermore, the significance of TJuly reconstruction was tested with taking Pann as conditional variable (and vice versa) in the inherent constrained ordination.
Estimation of the root mean square error of prediction (RMSEP) is derived from cross-validation of the calibration set. It yielded a RMSEP of 2.04°C for T July and of 94.32 mm for P ann . These prediction errors point to the potential range of the absolute biases; however, the relative changes in climate have likely much lower biases. Table 1. Radiocarbon and Intcal13 calibrated ages derived from bulk TOC samples from cores 16-KP-01-L02_L3 (Lake Ilirney) and 16-KP-04-L19_L2 (Lake Rauchuagytgyn). Samples with the prefix 'Poz' were dated at the Pozna n radiocarbon laboratory, Poland, while samples with prefix 'AWI' were dated at the MICADAS laboratory, Germany. Ages marked with * were used for age modelling. Pozna n dates from Lake Ilirney were discarded because of too low carbon contents.

Sample ID
Depth (   This means that maybe the climate time series needs to be shifted to the left or right by about the RMSEP. Trends between sections rather than fluctuations between single samples should be interpreted because the MAT technique is sensitive to the availability of analogues. The inclusion or exclusion of single modern samples during the MAT routine may artificially cause some fluctuations, which may increase the already low signal-to-noise ratio of pollen data e.g. originating from the abundance changes of high pollen producers and the fact that pollen data are a 'closed' data set (i.e. percentage data where changes of one taxon affects all other taxa). See Birks & Sepp€ a (2004) for further discussion of the concept and potential biases of pollen-based climate reconstructions.
For plotting of the reconstructions we applied a Gaussian smoother with 2000 years smoothing length on irregularly sampled time series using the CorIrregTimser-function from the corit R software package, version 0.0.0.9000 (Reschke et al. 2019). The smoothed time series were plotted on top of the reconstructed time series in red colour for T July and in blue colour for P ann .

Results
Ilirney core (16-KP-01-L02) Core lithology. -The core is 235 cm in length and is dominated mainly by silty sediments containing irregularly distributed, millimetre to centimetre size vivianite aggregations. This mineral was identified visually by its distinctive change from white to blue coloration, and may have formed diagenetically below the water-sediment interface due to reducing conditions at low sedimentation rates (Biskaborn et al. 2013b). Clayey silt forms the dominant lithology between 235 and 120 cm. The upper 120 cm consists of grey-brownish silty sediments of coarser grain size and is characterized by irregularly spaced laminations and higher total organic content. Age/depth model. -To establish an age-depth model for core 16-KP-01-L02 0-1 cm from Lake Ilirney we used the radiocarbon dating results of eight samples from MICADAS ( Table 1). Dating of surface sediment samples from Lake Ilirney yielded a 14 C age of 1.721 ka (EN18214 0-0.5 cm). We used this value to represent the old carbon effect at Lake Ilirney. This date was subtracted from all 16-KP-01-L02 radiocarbon dates prior to calibration (Table 1) under the mandatory assumption of a temporally constant reservoir effect (Biskaborn et al. 2012;Philippsen 2013). Though the reservoir effect may not be uniform with age, without the application of alternative comparative dating techniques, this cannot be effectively constrained. We found that the distributions of reservoir corrected ages represent realistic agedepth relationships without reversals. However, we treated the sample at 6-8 cm depth (5950AE25 14 C a BP) as an outlier because of too low radiocarbon concentra-tions. XRF derived elemental data allowed a robust correlation between the published age model from the nearby (~250 m) core EN18208 (Vyse et al. 2020) and core 16-KP-01-L02 used here (Fig. 3C). The approach of correlation is valid considering the generally homogenous pattern of sedimentation revealed by acoustic studies at Lake Ilirney (Vyse et al. 2020). Eight robust tie-points allowed transfer of median ages and respective 2 sigma ranges from EN18208 onto the age model established for 16-KP-01-L02. All EN18208 median ages fit into the 2 sigma ranges modelled for 16-KP-01-L02 and thus validate the developed chronology. The minor change in sedimentation rate as observed at around 20 cm is probably related to the overall errorof the dating strategy and the necessary assumption of a constant old carbon effect over time. It is possible that around 20 cm depth true ages can be several hundred years younger, which should be taken into account when interpreting the data. The old carbon corrected and calibrated samples show a maximum age at the 235 cm depth of 27.9 cal. ka BP. Mean accumulation rate calculated from the age/depth model was 0.015 cm a -1 with predominantly low values (0.006-0.011 cm a -1 ) between~190 and 101 cm (26.03-11.36 cal. ka BP). Nanae (up to 12%) and Alnus fruticosa (up to 8%) pollen are also characteristic for IL I. Spores of Selaginella rupestris and remains of green algae colonies (Pediastrum and Botryococcus) are also very common in the spectra. Pollen concentration is rather low (up to 4350 grain g -1 ). IL I can be subdivided into two subzones: IL Ib (187-153 cm, c. 25.45-18.65 cal. ka BP) differs from IL Ia by higher percentages of Selaginella rupestris (up to 27%) spores and remains of Botryococcus (up to 18%).
IL II (153-136 cm, c. 18.65-15.90 cal. ka BP) is notable by significantly lower percentages of Betula sect. Nanae (up to 3%) and Alnus fruticosa (up to 2%) pollen as well as spores of Selaginella rupestris (up to 17%). The uppermost pollen spectra demonstrate a decrease in Poaceae pollen percentage (up to 26%), while percentages of Salix (up to 13%) and Cyperaceae (up to 38%) pollen and Selaginella spores (up to 20%) as well as remains of Botryococcus (up to 36%) increased.
IL III (136-116 cm, c. 15.9-13.2 cal. ka BP) is dominated by Betula sect. Nanae (up to 63%) pollen, while percentages of herb pollen (especially Poaceaeup to 18%, Cyperaceaeup to 15% and Artemisiaup to 8%) are reduced. Numerous remains of Botryococcus colonies (up to 47%) are also very characteristic for the zone. Pollen concentration is higher (up to 12 910 grain g -1 ) in IL III in comparison to the lower zones.
IL IV (116-106 cm, c. 13.20-11.84 cal. ka BP) is notable by an increase in Alnus fruticosa percentages (up to 42%) and high percentages of Betula sect. Nanae (up to 43%) and Salix (up to 13%), while percentages of herb pollen and remains of Botryococcus colonies decreased. Pollen concentration is higher than in IL III (up to 34 170 grain g -1 ).
ILV (106-99 cm, c. 11.84-11.19 cal. ka BP) is notable by a further increase in Alnus fruticosa percentages and high percentages of Betula sect. Nanae and Salix, while percentages of herb pollen and Selaginella spores decreased. Pollen concentration is significantly higher than in IL IV (up to 198 630 grain g -1 ).
The pollen concentration is the highest (up to 219 560 grain g -1 ) in IL VI (99-43 cm, c. 11.19-7.53 cal. ka BP), which is also notable for high amounts of Alnus fruticosa and Betula sect. Nanae pollen. Single pollen grains of Larix and Pinus are also characteristic for this zone.
The uppermost zone, IL VII (43-0 cm, c. 7.53-0 cal. ka BP), demonstrates the higher percentages of Pinus and herb taxa pollen as well as remains of Botryococcus colonies; however, Alnus and Betula pollen dominate the pollen assemblages. The pollen concentration is lower than in IL VI (up to 93 240 grain g -1 ).
PCA results. -The PCA biplot of the first two axes (Fig. 6)  Climate reconstructions. -Climate reconstructions were performed for the Ilirney core. Comparison of fossil pollen data with the modern training data revealed poor analogues before the Holocene but very good analogues for the Holocene period (Fig. 8). The reconstructed T July fluctuated between 7.7 and 9.4°C from 27.90 to 15.72 cal. ka BP (Fig. 9A) (Fig. 9B). A stronger rise in P ann (up to 282 mm) occurred around 14.7 cal. ka BP. Climate conditions gradually became more humid during the Lateglacial as indicated by increasing P ann values. The highest precipitation was reconstructed for the HTM, when P ann values reached 343.5 mm around 8.9 cal. ka BP. After the HTM the climate became drier again, but with strong fluctuations in P ann , reflecting wetter and drier periods.
According to cross-validation results of the calibration set, the T July reconstructions (r 2 = 0.56) are more reliable compared with the P ann reconstructions (r 2 = 0.31). The T July reconstruction was significant (according to Telford & Birks 2011) when included as the single variable in the significance test (p < 0.001) as well as when including the precipitation reconstruction as a conditional variable (p < 0.001). P ann was only significant when included as the single variable (p < 0.04) but not when taking the T July reconstruction as a conditional variable (p < 0.1). Core lithology. -The core is characterized by homogenous brown-coloured laminated silts similar to those observed in the Holocene part of the Ilirney core (16-KP-01-L02), which also contain irregularly distributed vivianite aggregations. The core sediments show a slight darkening towards the sediment surface reflecting increasing total organic carbon content. Age/depth model. -A surface sediment age of 1.03 cal. ka BP found for Lake Rauchuagytgyn was subtracted from all radiocarbon dates, applying an identical procedure to 16-KP-01-L02. Due to the higher carbon content of dated radiocarbon samples (six samples >1 mg C), samples were not dated additionally at the MICADAS laboratory and the final age model was constructed from Pozna n dated samples. The old carbon corrected and calibrated samples show an age at the 16-KP-04-L19 core base of 9.47 cal. ka BP (Fig. 3B). Mean accumulation rate for core 16-KP-04-L19 was 0.044 cm a -1 and thus approximately three times the accumulation rate observed for core 16-KP-01-L02.
Pollen stratigraphy. -A total of 53 samples were analysed for pollen and NPPs. The pollen spectra can be subdivided into two main pollen zones (Fig. 5). R I (290-193 cm, c. 9.47-6.89 cal. ka BP) is dominated by Alnus fruticosa and Betula sect. Nanae pollen with some Poaceae and Cyperaceae. Pollen concentration is very high (up to 107 010 grain g -1 ) in the zone.
R II (193-0 cm, c. 6.89-0 cal. ka BP) is also dominated by Alnus and Betula pollen, but increased amounts of Pinus and Salix pollen are characteristic for this zone. Percentages of Poaceae, Cyperaceae and Artemisia pollen as well as spores of Polypodiaceae and Sphagnum are higher than in R I. Pollen concentration is significantly lower (up to 32 120 grain g -1 ) than in R I. R II can also be subdivided into two subzones: R IIa (193-12 cm, c. 6.89-0.32 cal. ka BP) differs from R IIb by slightly higher percentages of Pinus pollen and spores of coprophilous fungi.
PCA results. -The PCA biplot of the first two axes (Fig. 7) explains 50.29% of the variance in the data set. Samples from R I are spread across PC2 and are located mostly in the right quadrants. R I is characterized by the presence of Alnus fruticosa. The R II samples, located in the left quadrants, are spread across PC2 and represented by Salix, Pinus s/g Haploxylon, Cyperaceae, Artemisia and Poaceae pollen. The presence of pollen types from Rosaceae or Betula sect. Nanae does not discriminate between the two zones.
Climate reconstructions. -Comparison of fossil pollen data with the modern training data revealed very good analogues (Fig. 8). The reconstructed T July values for the Rauchuagytgyn core range between 11.1 and 13.3°C with a mean value of 12.3°C for the whole time interval (Fig. 10A) (Fig. 10B). Thereafter until 1.66 cal. ka BP, P ann became more stable showing lower fluctuations within the range of 239.1 to 281.3 mm around the mean value of 260.5 mm, with a significantly higher peak of 313.1 mm at 4.39 cal. ka BP. Subsequently, climate conditions became slightly wetter again, reaching its highest level of 297.9 mm at 0.4 cal. ka BP. For Lake Rauchuagytgyn only the reconstructed P ann is significant at p-value level <0.09 when P ann reconstruction was included as a single variable.

Chronological limitations
The 14 C-based age/depth models presented for lakes Rauchuagytgyn and Ilirney are among the first to be developed for northern Chukotkan glacial lakes and as such can provide valuable information regarding the timing of major climatic changes. The lack of agereversals within both age/depth models from both lakes suggests continuous lacustrine sedimentation over the studied time periods. The broadly consistent timing of major vegetation and climate changes at both lake sites as well as of palynological data from other Chukotkan sites provides support to the consistency of major regional vegetation changes (e.g. Lozhkin & Anderson 2015). In addition, the agreement between XRF tie-point ages and 2 sigma-range overlaps from the age/depth model of core EN18208 give (Fig. 3C) support to the presented age/ depth model for core 16-KP-01-L02 (Vyse et al. 2020). Chronologies of arctic lakes should however be treated with caution as radiocarbon dating of lacustrine sediments from high latitude sites situated within the continuous permafrost zone have been shown to be limited by numerous factors that increase uncertainty associated with age/depth modelling (e.g. Biskaborn et al. 2012Biskaborn et al. , 2013aBiskaborn et al. , 2019bBouchard et al. 2016). Long landscape residence times of organic matter within lake catchments situated within permafrost regions followed by erosion, entrainment and redeposition by glacial and/ or fluvial pathways to lake basins can act to produce sediment ages that are unrealistically old (Nelson et al. 1988;Zimov et al. 1997;Oswald et al. 2005). This is especially problematic within lake systems where datable macrofossil remains are sparse or absent and hence dating is restricted to sediment bulk TOC as is the case at both lake sites discussed here (Kaufman et al. 2004). Recent studies have shown that the usage of bulk sediment 14 C ages may result in age offsets when compared with dated macrofossils in particular within sediments of low organic carbon content deposited underglacial conditionswhere sediments are particularly susceptible to contamination by old carbon (Oswald et al. 2005;Gaglioti et al. 2014;Strunk et al. 2020). A comparison however between ages derived from macrofossils and bulk sediment could not be applied to the lakes considered within this study, due to the absence of appreciable quantities of macrofossils for 14 C dating. The low amount of organic carbon availability for 14 C dating and small sample sizes of 14 C samples, particularly those dated at Pozna n, leading to their exclusion from age/ depth modelling here, can also contribute to age/depth model uncertainty (Baumer et al. 2021). Quasi-permanent ice cover during cold phases may also impact uncertainty by leading to variations in lake reservoir age as has been suggested to explain reservoir effects affecting 14 C chronologies at Lake El'gygytgyn (Melles et al. 2007;Swann et al. 2010). The influence of detrital carbonate from catchment carbonates is likely of minimal influence at both lake sites due to the minor presence of carbonate lithologies (Zhuravlev et al. 1999). Limitations associated with 14 C dating of lacustrine and other terrestrial sites have been widely discussed for Chukotkan sites as well as records from Arctic regions of North America , 2015Kaufman et al. 2004;. Despite uncertainties pertaining to the age/depth models presented within this study, support is garnered from regional vegetation comparisons and a within-lake correlation that supports the presented chronologies.  2)) indicate that open tundra-and steppe-like landscapes were dominant. Relatively high percentages of Betula sect. Nanae (up to 12%) and Alnus fruticosa (up to 8%) pollen in these MIS 2 sediments are comparable with modern pollen assemblages from high arctic Chukotka sites (e.g. Lozhkin et al. 2001). We assume that Betula and Alnus shrubs might have survived in small refugia in the Lake Ilirney vicinity between c. 27.90 and 18.65 cal. ka BP. Sheltered microhabitats in deep valleys and protected lower slopes might provide conditionswhere summer temperatures, effective moisture and snow accumulation were enhanced as compared to the regional climate conditions (Lozhkin et al. 2018) The presence of their DNA in the lake sediments confirms this suggestion (Huang et al. 2020). The probability that these taxa might survive in the lake vicinity is in line with other records from West Beringia including from the Kankaren Range (southern Chukotka (sites 2-3 on Fig. 1; Anderson & Lozhkin 2015)) and the Ledovy Obryv sediments (southwestern Chukotka, site 4 on Fig. 1; Lozhkin et al. 2000Lozhkin et al. , 2018. Our reconstructions show that T July in the lake area were~4-5°C below the modern values during this MIS 2 interval (Fig. 9A). Although P ann reconstructions are less reliable than those for T July they show that P ann varied between 200 and 250 mm in this time (Fig. 9B). The high amounts of Selaginella rupestris spores (indicator of dry and cold climate conditions) found in all records from Chukotka reflect severe environmental conditions. Rather high percentages of coprophilous fungi spores (mostly Sporormiella) represent indirect evidence for the presence of rather numerous herbivores at the lake vicinity during this time. It is also indirect evidence that the bioproductivity of local plant communities was sufficient to support these herbivores. Only single pollen grains of Betula and Alnus were found, which probably reflect the decrease or disappearance of shrubs in the study area during this time interval. However, some dwarf Betula probably survived in more protected habitats as is evident from DNA data (Huang et al. 2020) but they may not have produced much pollen due to extremely unfavourable environmental conditions (Lozhkin et al. 2018 and references therein).
The uppermost IL II pollen assemblage demonstrates that shortly before 16.1 cal. ka BP Salix shrubs and Cyperaceae significantly increased their presence around the lake. These changes suggest the beginning of Lateglacial climate amelioration in the area. T July increased up to 1.6°C shortly after 16 cal. ka BP compared to the previous interval (Fig. 9A).
c. 15.90-13.34 cal. ka BP. -The increased percentages of Betula sect. Nanae pollen in the IL III (Fig. 4) spectra reflect the fact that dwarf Betula started to grow in the lake vicinity at about 15.9 cal. ka BP. After 15.52 cal. ka BP, dwarf Betula and Salix communities became dominant in the Ilirney area. The rapid increase of Betula pollen percentages around this time is well known from Beringian pollen records (e.g. Brubaker et al. 2005;Edwards et al. 2005 and references therein) and is associated with the onset of Lateglacial climate amelioration. The revealed increase started earlier than in the majority of West Beringia records reflecting that shrub Betula tundra became dominant c. 14.7-13.8 cal. ka BP. However, the Ilirney record coincides well with the southern Chukotka records, where Betula percentages started to increase as early as~18-19 cal. ka BP (Anderson & Lozhkin 2015).
Small increases in Ericales and Cyperaceae pollen percentages suggest the spread of wetter habitats. Spores of Selaginella rupestris decreased in the pollen assemblages, but remained relatively high until c. 13.34 cal. ka BP suggesting still rather harsh climate conditions in the study area. T July reached 10.1-10.4°C between c. 15.70 and 13.38 cal. ka BP (Fig. 9A). P ann also gradually increased up to 280 mm (Fig. 9B).
Similar changes in pollen assemblages were documented in the southern Chukotka records for an equivocal time interval (Anderson & Lozhkin 2015). Thus, the Lateglacial pollen records suggest gradual climate amelioration in Chukotka between c. 15.52 and 13.34 cal. ka BP.
The slightly increased amounts of Alnus pollen may suggest the northward migration of shrub alder confirming climate amelioration during the IL III interval; however, Alnus probably did not grow yet in the study area. Although the Kankaren Range (sites 2-3 on Fig. 1; Anderson & Lozhkin 2015) and the Anadyr Lowland (sites 5-7 on Fig. 1; Lozhkin & Anderson 2013) records document that shrub alder was already widespread in southern and central Chukotka around 13.8 cal. ka BP, it is unlikely that shrub Alnus grew in the Ilirney area earlier than c. 13.34 cal. ka BP.
Very high presence of Botryococcus remains in the IL III sediments reflects widespread shallow habitats favourable for these green algae. One possible explanation of the observed Botryococcus peak could be shortterm lake level increases resulting in flooding of flat shore areas within the lake outflow area. These lake level pulses may reflect melting of local glaciers due to higher air temperatures (Vyse et al. 2020) and may also be associated with increasing mean annual precipitation during the Lateglacial.
c. 13.34-11.84 cal. ka BP. -The gradual increase in Alnus fruticosa pollen percentages starting at the beginning of IL IV (Fig. 4) documents that alder stands became common in the study area after c. 13 cal. ka BP. Herbs and Selaginella rupestris gradually decreased further within the vegetation cover, reaching a minimum at the Pleistocene/Holocene boundary. Changes in pollen assemblages and a significant increase in pollen concentration reflect gradual climate amelioration with T July rising slightly above 12°C (Fig. 9A) and increasing P ann (Fig. 9B).
At the end of the interval, the pollen spectra do not reflect any significant fluctuations, which can unequivocally be interpreted as YD cooling. The only pollen assemblage showing some increase in Artemisia and Poaceae pollen percentages (IL IV) may indicate cooling around c. 12.5 cal. ka BP. This is inconsistent with the El'gygytgyn environmental records (Andreev et al. 2012;Melles et al. 2012), where the YD cooling is well pronounced, andwith records from the Anadyr Lowland  and from a number of sites across West Beringia (Anderson & Lozhkin 2002 and references therein), which also show a significant decrease of shrub pollen percentages at the Lateglacial/ Holocene transition coinciding with the YD cooling. The well-dated Lateglacial sediments from Wrangel Island (northern Chukotka, site 8 on Fig. 1; Lozhkin et al. 2001) with drastic increases in percentages of Selaginella rupestris spores, a good indicator of dry and cold environmental conditions, and a simultaneous decrease in Betula pollen percentages also reflect climate deterioration at the Lateglacial/Holocene boundary coinciding with the YD, although Lozhkin et al. (2001) did not interpret this event as the YD. Palynological records located west of the Ilirney Lake (Lena and Indigirka drainages) also demonstrate the cooler climate reversal at the Lateglacial/Holocene boundary (Andreev et al. 2011;Lozhkin et al. 2018 and references therein).
Although the Ilirney pollen assemblages of YD age, which do not show any significant cooling, differ from other northern records they show a good similarity with the contemporaneous pollen records from southern Chukotka (sites 2-3 on Fig. 1; Anderson & Lozhkin 2015), which also do not show evidence of any significant cooling. The relatively warm climate conditions during the YD time were also inferred from peat records from other locations in the northern regions of Far East Russia (Lozhkin et al. 2011 and references therein). Generally spatially heterogeneous climate conditions during the YD were reported by Kokorowski et al. (2008a, b). Based on 75 lake and peat records, they revealed coherent but spatially complex patterns: cooling in southern Alaska, eastern Siberia and some parts of northeastern Siberia; and uniform to warmer-than-present conditions in different locations from a number of sites from central and northern Alaska, northern Siberia and Far East Russia. As a possible explanation for the observed geographical differences it was suggested that large-scale climatic forcing was further modified by local nonclimatic factors, such as topography, soil substrates, and disturbance of sediments (Kokorowski et al. 2008a, b). However, it is difficult to explain why rather closely situated Ilirney and El'gygytgyn sites demonstrate such different environments during the YD time. Pollen records from southern and southeastern Chukotka (the relatively closely located Kankaren and Anadyr Lowland sites) also demonstrate a similar discrepancy Anderson & Lozhkin 2015). The local topography, soils, as well as the disturbance of sediments during deposition and age/ depth model uncertainty may play important roles resulting in such discrepancies at both northern and southern locations. These closely situated, but spatially heterogeneous records have to be additionally studied with high-resolution sampling and supported by reliable age control, which may reveal possible hiatuses and/or extremely low sedimentation rates. Unfortunately, the lack of organic material in the Lateglacial, and especially within YD sediments often prevents such control making direct comparison of the sediment records problematic.
Holocene. -The Holocene sediments from both study sites (IL V-VII of Fig. 4 and R I-II of Fig. 5) are dominated by Alnus and Betula pollen with some Salix, Poaceae and Cyperaceae. These pollen assemblages reflect that shrub and graminoid tundra communities were dominant around Lake Ilirney between c. 11.84 and 11.10 cal. ka BP. Salix stands significantly decreased after c. 11.1 cal. ka BP. Single pollen grains of Larix occurring in the Early Holocene sediments probably reflect that first larch stands established in the region. This suggestion is in good agreement with the pollen and macrofossil data from the El'gygytgyn Crater where larch seeds and needles found in the permafrost sediments were dated between 11.2 and 9.0 cal. ka BP (Shilo et al. 2008;Andreev et al. 2012). Thus, a foresttundra ecotone seems to be established in the study area already in the Early Holocene, shortly after 11.1 cal. ka BP.
The Early Holocene sediments show the highest pollen concentration reflecting the most favourable climate conditions during the HTM (c. 10.6-7.0 cal. ka BP). T July reached 13.3°C at the end of the HTM in the Ilirney area and slightly decreased after c. 7 cal. ka BP (Fig. 9A). P ann reached 343.5 mm c. 8.9 cal. ka BP, but climate conditions became drier after 7 cal. ka BP (Fig. 9B). However, there are several increases in P ann , reflecting wetter intervals. T July and P ann reconstructed from the Rauchuagytgyn pollen assemblages show similar trends (Fig. 10A, B). The small differences most likely are associated with local topography as both lakes are situated in mountainous terrain.
Our data are in good agreement with other existing environmental records from Chukotka and the adjacent arctic areas (e.g. Lozhkin & Anderson 2006, 2013Andreev et al. 2011Andreev et al. , 2012Lozhkin et al. , 2018Anderson & Lozhkin, 2015 and references therein). The pollen-based biome and climate reconstructions from the Lake El'gygytgyn record also show that the most favourable environmental conditions occurred during the Early Holocene, c. 11.7-8.0 cal. ka BP (Melles et al. 2012;Tarasov et al. 2013). However, T July and P ann values reconstructed from the El'gygytgyn record are lower than at our sites, which is most likely related to a more easterly location of the lake and/or local topography. The difference might be also connected with different modern pollen sample data sets. Direct support for warmer-than-present temperatures is obtained from 14 C-dated macrofossil remains of trees and shrubs found beyond their modern distribution limits (e.g. Binney et al. 2009Binney et al. , 2017Lozhkin et al. 2011 and references therein). The Early Holocene (HTM) was the warmest time in high arctic Siberia in comparison to the continental regions as the coastline at the time was 200-500 km farther north, causing more continental conditions and lessened influences of the cool arctic seas (Kokorowski et al. 2008b;Biskaborn et al. 2016 and references therein).
The contents of green algae colony remains in the Early Holocene sediments are significantly lower than in the Lateglacial ones. This most likely reflects further melting of local glaciers and higher precipitation at the Holocene onset resulting in increased lake levels with deeper aquatic conditions, less favourable for Botryococcus and Pediastrum as they need shallow environments where water-masses can be warmed up quickly (Jankovsk a & Kom arek 2000 and references therein).
Amounts of Pinus pollen started to increase after c. 7.5 cal. ka BP in the Ilirney sediments and after c. 6.8 cal. ka BP in the Rauchuagytgyn sediments. These slight increases in Pinus pollen percentages do not confirm that dwarf stone pine grew close to the study areas at this time, but it may reflect the appearance of shrub stone pine in southerly locations. The southern Chukotka pollen records (Anderson & Lozhkin 2015) show that Pinus pumila was probably present there as early as 7.4 cal. ka BP, which coincides well with a slight increase of Pinus pollen percentages in our studied sediments starting c. 7.5 cal. ka BP. The uppermost pollen zones from both studied cores (ILVII and R II) demonstrate the highest percentages of Pinus. It is important to note that Pinus percentages are higher (up to 20%) in the Rauchuagytgyn sediments than in the Ilirney sediments, although Pinus pumila is absent close to Lake Rauchuagytgyn. The El'gygytgyn records also demonstrate increases in Pinus pollen percentages, but later, after 5 cal. ka BP (e.g. Melles et al. 2012). However, dwarf stone pine was also found similarly to not grow close to Lake El'gygytgyn. Thus, the most northern pollen records (Rauchuagytgyn and El'gygytgyn) demonstrate rather high Pinus percentages, although Pinus pumila is not currently growing around these lakes. Surprisingly, the Ilirney sediments do not show significant Pinus percentages even in the uppermost samples, whilst Pinus pumila stands are rather common in the lake vicinity. The DNA of Pinus was also not found in the Ilirney sediments.
The overrepresentation of Alnus pollen is also characteristic for both studied sediment cores. Very high percentages (40-60%) of Alnus pollen is documented in the Rauchuagytgyn sediments, whilst the nearest alder stands can be found only at about 5 km from the lake. Although shrub alder stands were found in the Lake Ilirney vicinity, they are very rare in the study area; however, the Alnus pollen percentages reach up to 40-50% in many samples. Very low presence of Alnus was found in the sedimentary ancient DNA in the uppermost sediments of Lake Ilirney (Huang et al. 2020). Pollen assemblages including modern assemblages from the El'gygytgyn crater also demonstrate high percentages of Alnus (Matrosova 2006(Matrosova , 2009Andreev et al. 2012 and references therein); however, no shrub alder currently grows in the lake vicinities. These examples demonstrate that considerable input of pollen grains from distant sources of several hundred kilometres must be considered when interpreting high arctic pollen records.
The uppermost pollen zones of both study sites demonstrate higher percentages of herb pollen taxa suggesting an increase of open herb-dominated habitats in both study areas. The slightly increased percentages of Ericales pollen and Sphagnum spores reflect that wetter habitats became more common suggesting increased humidity in both study areas around c. 7 cal. ka BP. Generally, other available records also show that vegetation cover in northern Chukotka became similar to modern c. 7 cal. ka BP. However, it remains unclear when stone pine (Pinus pumila) inhabited the Ilirney area.

Conclusions
The continuous lacustrine pollen records from Lake Ilirney and Lake Rauchuagytgyn offer valuable insights into the Late Pleistocene and Holocene vegetation and climate history of northwestern Chukotka: • Predominantly open steppe-and tundra-like vegetation dominated the areabetween 27.9 and 18.7 cal. ka BP producing sufficient biomass to support herbivores.