Preserved lipid signatures in palaeosols help to distinguish the impacts of palaeoclimate and indigenous peoples on palaeovegetation in northwest Siberia

Arctic reindeer herders demonstrate resilience to climate fluctuations by adjusting their pastoral practices to changing environments. The multiple phases of occupation at one of the oldest identified reindeer-herding sites, ̂ IA rte 6 on the ̂ IA mal peninsula, northwest Siberia, are thought to be linked to its local vegetation cover. Here we provide information on local palaeovegetation and climate shifts which occurred between the 7th and the 11th century CE based on lipid biomarkers. Aliphatic compounds, pentacyclic triterpenoids, branched glycerol dialkyl glycerol tetraethers (brGDGTs) and other lipids were analysed in four separate palaeosols from two loess-palaeosol sequences at the site. Based on different indices, the impact of human-or root-derived lipids and post-depositional microbial degradation on the signature of the studied pedosequences seem limited, which indicates that palaeobiomarkers are well preserved and representative of the analysed compounds. n -Alkanes, n -carboxylic acids, n -aldehydes and pentacyclic triterpenoids point to the progressive colonization of grasses, sedges and herbs at the site, which can be attributed to the regional decrease in temperature suggested by bacterial-derived brGDGTs. During the last phase of occupation, however, when proxies point to increasing temperatures, and shrubs would normally be expected to return, the shrub cover continued to decrease, probably due to the impact of camping and holding reindeer at the site. The decoupling trends observed for particular pentacyclic triterpenoids as potential dwarf birch biomarkers could suggest a preferential shift of species within shrubs, probably influenced by human activities. Multi-family lipid biomarker analysis therefore made it possible to distinguish anthropogenic impacts on the local vegetation cover from regional climatic changes, and show how significant the impacts of humans on local vegetation can be, even in extreme environments where such activities are limited.


Introduction
In tundra ecosystems, permafrost thawing caused by global warming is having dramatic impacts on flora and carbon release (Zimov et al., 2006).In addition, warmer temperatures and increased humidity due to more rapid evaporation of Arctic seas can induce dramatic climatic events such as the recent rain-on-snow events encountered in the ÎAmal Peninsula of northwestern Siberia (Forbes et al., 2016).Rain-on-snow, which rapidly freezes to form an impenetrable ice cover, has considerable consequences for wildlife and domesticated reindeer, and consequently for the local indigenous Nenetses.Their reindeer-herding lifestyle is dependent on accessible winter food sources (lichen) for deer (Forbes et al., 2016).As such, understanding how Nenetses and their predecessors responded to climate variations over time is of primary importance for predicting the future of Arctic indigenous peoples.At the same time, herding reindeer (Rangifer tarandus) can have significant impacts on the local landscape, for example due to wood-cutting for fuel, shrub grazing, soil trampling and inputs of urine and faeces by reindeer (Olofsson, 2006;Skarin et al., 2020).As a result, it can be very complex to determine the causes of palaeoenvironmental changes on sites with a long human-animal-environment relationship history.
ÎArte 6, an archaeological site on the ÎAmal Peninsula, through the analysis of loess-palaeosol sequences in its immediate hinterland.This site, located in a region still used by local Nenetses, has been occupied several times since at least the 3th century CE.The combined anthropological, palynological, geoarchaeological and faecal lipid biomarker data led Anderson et al. (2019) to suggest that the presence of humanreindeer groups began at the site at a modest scale, aimed mainly at hunting wild herds with just a few decoy deer.Over time, as the wild deer hunting and carcass-processing activities at the site intensified, ultimately resulting in the deposition of tens of thousands of reindeer bones and artefacts (Plekhanov, 2014;Nomokonova et al., 2018), reindeer-pulled sled transport became crucial for larger groups and their goods.The faecal lipid biomarker evidence around the site indicated an increase in the size of the domestic herds held there.This intensive phase of human and reindeer activity ca.1000 years ago is thought to be responsible for the unusually lush grass that still covers the site today, which stands out sharply in a tundra landscape otherwise characterised by a sparse cover of tundra herbs, mosses, lichen and dwarf shrubs (Plekhanov, 2014).
One of the key unanswered questions about this site is why this particular location was repeatedly selected as a place to camp and eventually process large numbers of reindeer carcasses.Based on ethnographic analogies with modern reindeer hunters and herders, Anderson et al. (2019) suggested that the microtopography and vegetation cover at the site probably played major roles, including the availability of shrub wood for fires, shrub cover for hunting hides, and suitable pasture for reindeer.However, the palaeovegetation cover at this site is very difficult to reconstruct based on plant macro-and microfossils.The nearby peat bog, which was sampled and analysed for pollen in 2013, does not date as far back as the first millennium CE, Fig. 1.Location of: (a) the ÎAmal Peninsula; (b) the ÎArte complex; (c) ÎArte 6 elevation model, including the locations of the loess-palaeosol sequences in test pits 10/80 and 30/60, the location of the different excavations conducted, modern vegetation community analysis (Bruce Forbes, personal data, quadrat numbers as in Supplementary Table S1) and the location of the soil monolith analysed for pollen (Anderson et al., 2019).while the lakes in the region, which do have sufficiently long sequences, provide only a broad regional picture with unclear palaeoclimatic trends (Anderson et al., 2019; Lake Three).Moreover, the next nearest lakebased Holocene palaeoenvironmental reconstructions were conducted further than 2000 km east of the ÎAmal Peninsula, and at a kyr time scale resolution, preventing any comparison with the ÎArte archaeological context (Andreev et al., 2004, Müller et al., 2009).Previous to the current study, only pollen data obtained from a monolith collected at ÎArte 6 from the excavation itself were available to interpret the past vegetation cover, but this sequence was disturbed and covered a restricted timeframe (Panova, 2008;Anderson et al., 2019).Therefore, the macro-and microplant fossil records available to reconstruct the palaeovegetation at ÎArte 6 is incomplete and requires additional corroborating evidence from the palaeosol sequences.
Compared to pollen, which can travel greater distances, biochemicals from plants, especially lipids, can be associated with very localized inputs of soil organic matter (SOM) by a progressive incorporation and physical degradation of local plant remains after decay (Davis, 2000).Lipid-based palaeoenvironmental reconstructions using soils or sediments are often based on n-alkanes, which can allow the identification of the main plant group sources based on the distribution of long-chain homologues.However, the identification of palaeovegetation using only n-alkane distributions in environmental archives can be complex because of; (a) inter-and intra-plant species chemotaxonomic differences, (b) other sources of n-alkanes in soil organic matter (SOM) such as plant roots, and (c) microbially driven selective degradation of n-alkanes (Jansen and Wiesenberg, 2017;Thomas et al., 2021 and references therein).Complementary information can be obtained from the study of other lipid classes, and the potential effects of other sources of biomarkers (roots, human-derived compounds, etc.) or post-depositional processes (e.g., microbial degradation) have to be taken into account (Jansen and Wiesenberg, 2017).
In addition to palaeovegetation reconstruction, lipid biomarkers and related proxies can also be used to infer past temperature change (Jansen and Wiesenberg, 2017).Among these compounds, branched glycerol dialkyl glycerol tetraethers (brGDGTs) are a group of membrane lipids produced by some bacteria, which are ubiquitously present in terrestrial and aquatic environments.Their distribution in soils has been shown to be mainly dependent on air temperature and soil pH (Weijers et al., 2007;De Jonge et al., 2014;Véquaud et al., 2022), and they are now widely used to reconstruct past mean annual air temperatures (MAT) in terrestrial environments (e.g., Rees-Owen et al., 2018).
In this study, various lipid biomarkers (n-alkyl compounds, pentacyclic triterpenoids and brGDGTs) were analysed to reconstruct palaeovegetation and climate changes from two loess-palaeosol sequences at the ÎArte 6 site (NW Siberia).The palaeoenvironmental changes derived from lipid-based proxies were then compared to other local and regional palynological and dendroclimatological records to disentangle humanfrom climate-induced palaeovegetation shifts.Even though lipid biomarkers have previously been applied to soil-sediment archives (Wiesenberg et al., 2015) and loess-palaeosol sequences (e.g., Zech et al., 2012;Tarasov et al., 2013) to reconstruct Quaternary palaeoenvironments in Siberia, the present study is the first one to investigate the potential human-animal impacts on the environment at a Siberian archaeological site.

Site description and sampling strategy
Detailed descriptions of the ÎArte 6 site, its surrounding landscape and associated palaeosols can be found in Anderson et al. (2019) and Harrault et al. (2019).The site is located on the western edge of a high and open terrace overlooking the southern floodplain of the ÎUribei River in the ÎAmal Peninsula, northwest Siberia (Fig. 1).The modern mean annual temperatures range between − 5 • C and − 10 • C, with snow and ice cover up to ten months a year.

Modern vegetation
The site is located ca. 250 km north of the current tree line.The contemporary landscape at ÎArte 6 displays different types of vegetation over a span of a few dozen metres (Fig. 1), a level of variation consistent with that observed in the wider vicinity (Skarin et al., 2020;Verdonen et al., 2020).The modern vegetation cover was formally surveyed by Bruce Forbes during the summer of 1996, and has remained largely unchanged since then.It consists of an unusually lush grassy sward largely dominated by Poceae (quadrats 1 and 2 in Supplementary Table S1) on the western edge of the terrace, where the archaeological site is located, and on the slope below the site.Shrubs are almost entirely absent from this zone.East of the archaeological site is a disturbed sandy tundra meadow composed of lichens, mosses, grasses and very low shrubs (quadrat 4).Even further east is an ombrotrophic bog dominated by lichens (quadrat 5).Large patches of deciduous shrubs (Betula nana, Salix spp., etc.) are found on the lower part of the terrace slope (quadrat 6) as well as further west towards the mire (no quadrat number) and nearby oxbow lakes in the floodplain.North and south of the ÎArte 6 terrace is a typical tundra dwarf shrub habitat dominated by dwarf shrubs (Betula nana, Ledum decumbens, Carex bigelowii) and stunted shrubs (Salix spp).

Pedosequence description
The focus of the present study is the sandy tundra meadow east of the archaeological site, which contains discontinuous loess-palaeosol sequences (LPS) with buried A horizons (Ab, palaeosols) separated by aeolian (windblown) sandy silt.These were recorded and sampled using shovel test pits on a 10 m grid (subsequently reduced in one zone to 5 m for higher resolution), the full details of which are in Anderson et al. (2019).At several locations, the Ab horizons and/or their associated humic horizons (Ahb) contained charcoal, ceramic fragments, bone fragments and/or burnt bone fragments, which were associated with temporary human occupation (Anderson et al., 2019).Due to frostheave and incomplete or irregular sealing of palaeosols, many test pits presented only two or three Ab horizons, while the highest number of buried palaeosols found was four.These were located at two locations: 10/80 and 30/60, and the samples from the Ab horizons in these test pits were used for the present study (Figs. 1 and 2).
The test pit located at 30/60 on the local grid contained a clear pedosequence with a depth of 41 cm.It consisted of a succession of sandy silt or silty sand palaeosols and aeolian deposits, all of which contained living roots (Fig. 2, Supplementary Table S2).Due to the high surface magnetic susceptibility measurements, abundant pottery pieces, and wood charcoal fragments, 30/60 was interpreted as the location of an annex campsite used before and during the occupation of the main ÎArte 6 site.As described in Anderson et al. (2019), the calibrated radiocarbon dates on the largest charcoal pieces (Betula sp. and Salix sp.roundwood) from each palaeosol (Fig. 2) spanned the 7th to 11th century CE.
In the 10/80 test pit, the four palaeosol horizons were distinguishable on the basis of colour and boundaries, but there was no loess separating the upper two palaeosols (Ab1 and Ab2) or the lower two (Ab3 and Ab4).This could be the result of wind erosion, but elevated concentrations of phosphates and faecal lipid biomarkers (5β-stanols) at this location also suggest that the soil could have been eroded by the localized congregation of reindeer, as was observed at Nenets campsites throughout Siberia (Anderson et al., 2014;Harrault et al., 2019).Living roots from grasses and dwarf shrubs were present in the four palaeosols (Supplementary Table S2).No charcoal or artefacts were found in this pedosequence, making it difficult to date independently.

Sampling
After photographing the freshly exposed soil profiles in each test pit and recording the depth, Munsell soil colour (Munsell Color Co., Inc., 1975), texture, structure and inclusions for each soil horizon (Supplementary Table S2), 50-500 g of fresh soil was collected from the buried A horizons, depending on their thickness (Fig. 2).The loess between the buried soils was not sampled from each test pit because the focus of the original sampling design was the identification of anthropogenic deposits/inclusions and faecal lipid biomarkers associated with the stable land surfaces (palaeosols).Soil samples were collected directly from the exposed soil profile on the edge of the test pit with a metal trowel or a spatula wiped with a clean paper tissue and rinsed with deionised water.The exposed surface of each palaeosol was sampled by scraping and bagging a few mm of soil from the whole thickness of the palaeohorizons identified as buried soils on the basis of their homogeneous colour, structure and texture (Fig. 2).Thus, each palaeosol sample was considered to be an integrated snapshot of palaeoenvironmental conditions while the whole of the horizon was forming.No field replicates were sampled.Soil samples were stored in ziplocked polyethylene bags for a few hours, before being taken to a field lab tent, where they were placed into hand-made foil basket for 48 h of air drying.Air-dried samples were sieved at 2 mm to remove bigger roots and pebbles and to retrieve charcoal and bone pieces, then sieved samples where repacked into plastic bags.After the month-long field season, samples were re-dried at the lab in an oven at 50 • C overnight before further treatment for lipid biomarker extraction.

Lipid analysis
Samples from each palaeosol (Ab1 to Ab4) from both test pits (30/60 and 10/80) were split into three sub-samples, which were each extracted and analysed independently, and then treated as triplicates (n = 3).

Sample preparation
Lipid extraction and separation procedures were adapted from Harrault et al. (2019).Approximately 4 g of dried, crushed and 250 µmsieved soil samples were extracted three times by sonication with a dichloromethane (DCM)/methanol (MeOH) mixture (2:1, v/v), then centrifuged and filtered.Pooled and concentrated extracts were separated into apolar and polar fractions by solid-liquid chromatography in glass columns filled with silica: apolar fractions were eluted with nheptane and a n-heptane/DCM mixture (2:1, v/v), then polar fractions were eluted with a DCM/MeOH mixture (2:1, v/v).Before analysis by GC-MS, deuterated n-alkane (d 50 -C 24 ) and 5α-cholestane were added as internal standard (IS) to aliquots of concentrated apolar and polar fractions, respectively.The latter was derivatised with a mixture of N,Obis-(trimethylsilyl)trifluoroacetamide and trimethylchlorosilane (BSTFA + TMCS, 99:1, v/v) at 70 • C for 30 min.For brGDGT analysis, a second aliquot of the polar fraction was centrifuged, dissolved in nheptane and a known amount of a C 46 -synthesized internal standard was added prior to analysis (Huguet et al., 2013).

Apolar fraction analysis
Apolar fractions were analysed by a gas chromatograph coupled with a mass spectrometer (GC-MS), Agilent GC6890N-MS5973N equipped with an Rxi-5Sil MS capillary column (Restek, 30 m × 0.25 mm i.d., 0.5 µm film thickness). 1 µl samples were injected in splitless mode at 280 • C. Helium was used as carrier gas and flow was set at 1 ml/min.The GC oven temperature program started at 50 • C, increased to 320 • C at 4 • C/min, and held for 15 min.The transfer line temperature was set at 310 • C and EI voltage at 70 eV.Analyses were carried in full scan mode (m/z 35-700).Biomarkers were identified from mass spectra and GC retention times.Semi-quantification was achieved by comparison of compound total ion current (TIC) peak area with those of the internal standard and expressed as relative abundance (%) compared to the sum of all integrated compounds within the fraction.Within each compound family, compound variations were expressed as their relative abundance compared to the sum of compound within the family considered.

Polar fraction analyses
Polar compounds were analysed with the same GC-MS instrument and capillary column as the apolar fractions.1 µl of derivatised fractions were injected at 280 • C in split mode (10 ml/min) with a constant 1 ml/ min flow of helium as the carrier gas.The GC oven temperature program started at 60 • C, increased to 150 • C at 15 • C/min, then increased to 310 • C at 3 • C/min and was held for 20 min.The transfer line temperature was set at 310 • C and EI voltage at 70 eV.Analyses were conducted in full scan mode (m/z 35-700).Compounds were identified from mass spectra and GC retention times.Semi-quantification was achieved by comparison of their TIC peak area with those of the IS (5α-cholestane) and expressed as the relative abundance of compound families and compounds within each family, as for apolar fractions.
GDGT analyses were performed using high performance liquid chromatography-atmospheric pressure chemical ionisation-mass spectrometry (HPLC-APCI-MS) with a Shimadzu LCMS-2020, as described by Huguet et al. (2019).Briefly, separation was achieved with two Hypersil Gold silica columns (Thermo Scientific, 150 mm × 2.1 mm, 1.9 µm) mounted in series at 40 • C, using different proportions of hexane and hexane/isopropanol (9:1, v/v) at 0.2 ml/min.This method allows the separation of 5-methyl and 6-methyl brGDGT isomers.The injection volume was 30 µl.Detection was performed in selected ion monitoring mode, targeting the protonated molecules at m/z 744,1018,1020,1022,1032,1034,1036,1046,1048 and 1050 (Supplementary Fig. S1).Semiquantification of the GDGTs was performed by comparing the integrated signal of the respective compound with the signal of the internal standard previously added, as described by Huguet et al. (2013).

Lipid proxies used for palaeoenvironmental reconstruction
n-Alkane odd-over-even predominance (OEP) was calculated as follows (Zech et al., 2009): In Eq. (1) and following, C i represents the relative contribution of the homologue with an i-long carbon chain compared to the sum of the homologous series.
Palaeovegetation indices based on n-alkanoic acid distributions were calculated according to Schäfer et al. (2016): Deciduous trees, Grasses, The ratio of saturated over unsaturated C 16 and C 18 fatty acids was used as a proxy for microbial degradation of native SOM (Wiesenberg et al., 2015): With ΣC 16:1 and ΣC 18:1 representing respectively the sum of the different monounsaturated homologues of C 16:1 and C 18:1 .
The n-aldehyde C 26 /C 28 ratio was calculated as a possible proxy for palaeovegetation based on the results in van Bergen et al. (1997) displaying the presence of the C 28 homologues in some tree and shrub leave species and its absence in grasses and herbs.
As the concomitant occurrence of betulin, betulinic acid, lupeol and lupenone are used as markers for Betulaceae plant species in soil archive or archaeological material (Modugno et al., 2006), the sum of their respective abundances in SOM was used in the present study as a proxy for Betulaceae contributions to palaeovegetation.
Local plant (aerial parts and roots) lipids were not analysed in the present study.We compared our SOM lipid data to those of Ronkainen et al. (2015), which analysed the lipid composition (mainly n-alkanes) of above-and below-ground plant parts from a site located in a similar biome with a similar plant species as ÎArte 6 ′ s, ca.300 km to the southwest.Some n-alkane-derives proxies such as CDI or OEP, not originally presented in Ronkainen et al. (2015), were calculated here.
Additional lipid distributions are available in Supplementary Figs.S2 and S5 and Supplementary Material S1.

n-Alkanoic acids
Fatty acids are largely dominated by long chain saturated homologues in both soil sequences (more than 90% of total FAs) while unsaturated straight chain (mainly C 16:1 and C 18:1 homologues, data not shown) and branched homologues (odd numbered-carbon chain homologues from C 13 to C 19 ) are less abundant (respectively less than 8% and 2% of total FAs).Aliphatic saturated FA distributions range from C 6 to C 34 and show strong EOP predominance with values higher than 3.4 and unimodal distributions mostly centered at C 24 (Supplementary Fig. S2).In both pedosequences, EOP generally decrease with depth with values ranging from 4.11 ± 0.09 (Ab1) to 3.79 ± 0.08 (Ab4) along the 10/80 profile and from 4.31 ± 0.04 (Ab1) to 3.41 ± 0.02 (Ab4) along the 30/60 one (Fig. 5a,b).

Sources of organic matter in the tundra meadow palaeosols adjacent
to ÎArte 6

Natural vs anthropogenic OM
The composition of the SOM derived from lipid indices and ratios in buried A horizons from both 10/80 and 30/60 sequences showed overall similar variations with depth (e.g., ratios/proxies based on n-alkanes; saturated fatty acids and n-aldehydes; functionalised triterpenoids; brGDGTs, Figs.4-7, respectively).These geochemical similarities occur despite the fact that they probably represent two different human and animal activity areas (a campsite vs an animal congregating area), as suggested by the presence of artefacts and micro-refuse (charcoal, ceramics, etc.) in the 30/60 sequence alone (Anderson et al., 2019).This suggests that the anthropogenic impacts on the relative abundances of the different lipid biomarkers were limited despite the contrasting occupation histories of the two pedosequences, and that the environmental proxies derived from these lipid families can be used for palaeoenvironmental reconstruction.This also implies that even though it was not possible to stratigraphically link the horizons in the two distant soil profiles, the dates of their buried A horizons are likely to be roughly correlated.

Plant vs microbial OM
The diversity of n-alkyl families found in SOM, the large predominance of their long-chain homologues and their distinctive EOP (nalkanoic acids, n-alkanols, n-aldehydes α,ω-hydroxy alkanoic acid) or OEP (n-alkanes, n-alkan-2-ones) strongly suggest that lipids from ÎArte 6 palaeosols mainly derive from plant material (Jansen and Wiesenberg, 2017; Supplementary Fig. S2).Similarly, the high diversity of hydrocarbons and functionalised triterpenoids, and the predominance of phytosterols (Supplementary Fig. S3) suggest a major input of plantderived OM in soils (Volkman, 2005).Nevertheless, the presence of OM derived from microbial sources and activities to SOM cannot be excluded as microorganisms (including mycorrhizal fungi) can produce long-chain n-alkyl compounds (Nguyen Tu et al., 2011) and a nonnegligible amount of certain functionalised triterpenoids to SOM (Grandmougin-Ferjani et al., 1999).However, their contribution to SOM is generally considered negligible compared to plant-derived straight chain n-alkyl lipids and triterpenoids (Bai et al., 2009).Also, the presence at low abundance of branched and short chain fatty acids (iso and anteiso homologues, C 13 to C 19 , data not shown), cholesterol as well as bacterial brGDGTs in SOM suggest the occurrence, but limited, of microorganism-derived lipids in palaeosols.
The occurrence of some post-depositional biodegradation of palaeovegetation-derived lipids by soil microbial communities is suggested by the decrease with depth of the FA EOP (Schäfer et al., 2016;Fig. 4), the sitosterol/sitostanol ratio (Gaskell and Eglinton, 1976;Fig. 6) and the increase in the FA C 16+18 sat./unsat.ratio (Wiesenberg et al., 2015;Fig. 5), mainly observed between Ab1 and Ab2.However, this microbial degradation of organic matter seems limited, as the values of the FA C 16+18 sat./unsat.ratio remain lower than 10 and decreasing or similar values with depth from Ab2 to Ab4 suggest OM preservation (Wiesenberg et al., 2015).
The limited impact of microbial degradation is also suggested by the values of FA EOP, which are higher or similar (for Ab4 from the 30/60 profile, EOP = 3.41 ± 0.06; Fig. 5) to the values found by Schäfer et al. (2016) in the second topsoil under a deciduous litter (EOP = 3.46).In addition, the relatively high n-alkane OEP values found along both soil profiles (>5), although far lower than fresh plant material analysed by Zech et al. (2009; OEP = 15.0 for grass and herbs and OEP = 17.9 for deciduous trees), remain higher that the OEP values analysed in the soil samples under grassland (OEP = 4.5) and deciduous forest (OEP = 5.0) from the same study.Their increasing trend with depth support a low bacterial biodegradation of plant wax-derived n-alkanes along the ÎArte 6 palaeosols (Zech et al., 2009).Also, the n-alkane CPI values lower than 10 (Fig. 4) suggest some microbial degradation of the native plant lipids, but, as their values remain much higher than 1, lipid degradation seems limited (Cranwell, 1981).Moreover, one would expect the n-alkane CPI values to decrease with depth, as microbial degradation should increase with time in environmental archives.The fact that opposite trends were observed in both pedosequences strongly suggests that this degradation was low, or at least that the sources/processes responsible for this proxy overprinted that of microbial reworking.

Palaeovegetation vs modern root-derived plant lipids
As living roots from modern plant cover were present throughout both soil profiles (Fig. 2), their contribution to the SOM lipid pool has to be assessed to disentangle their signature from that of actual palaeovegetation-derived biomarkers (Jansen and Wiesenberg, 2017).The OEP and CDI values of n-alkanes in the soil samples (Fig. 4) are in the range of those of plant leaves analysed by Ronkainen et al. (2015) and lower than those of roots (Supplementary Fig. S6), suggesting that the impact of root-derived n-alkanes from modern plants on the SOM of the analysed palaeosols is low.Regarding the other lipids, the comparison of fatty acid, α,ω-hydroxy alkanoic acid and α,ω-alkanedioic acid composition of local plants (aerial and sub-surface parts) could not be carried out and is highly dependent on the context (Mendez-Millan et al., 2011;Gocke et al., 2013).
Overall, the analysis of the different lipids suggests that SOM in ÎArte 6 meadow palaeosols is mainly derived from above-ground plant material, probably incorporated during the development of each A horizon.Although minor microbial and root-derived lipid inputs to palaeosol OM is likely, their lipid signatures do not seem to have been overprinted by that of aerial plant material.In addition, although some proxy trends suggest a possible early degradation of certain lipids (sitosterol/sitostanol and saturated/unsaturated C 16+18 FA ratios), the absence of clear and shared trends between these proxies (n-alkane OEP, FA EOP) suggest a relatively good preservation of the lipids throughout the pedosequences, most probably facilitated by the local cold conditions (Zimov et al., 2006), which argue in favour of a fairly good representation of lipid signatures as palaeovegetation biomarkers.

Reconstruction of the local palaeovegetation during the occupation of ÎArte 6
Due to the incomplete pollen records from the soil monoliths taken from the sections of the archaeological excavation area and the lake sediment core (Panova, 2008;Anderson et al., 2019), the goal of this The high n-alkane OEP values of the oldest (Ab4; Fig. 4) palaeosols of ÎArte 6 meadow are closer to those of the fresh dwarf shrubs leaves analysed by Ronkainen et al. (2015;Supplementary Fig. S6; Rubus chameamorous and Vaccinum uliginossum) than to those of grasses (Carex aqualitis and Eriophorum sp.).The progressive decrease in n-alkane OEP values from the oldest to the youngest palaeosol towards intermediate values between those of shrubs and those of grasses suggests a decreasing contribution of shrub-derived n-alkanes to SOM.This trend might be the result of a growing contribution of grass-derived n-alkanes, as OEP values remained higher than those of fresh grasses analysed by Ronkainen et al. (2015, 6.0 for Eriophorum).The contribution of mossand root-derived lipids is likely limited as the OEP values of SOM remain largely higher than theirs.Similarly, the high values of the n-alkane C 31 / C 27 and C 33 /C 27 ratios found in the oldest palaeosol of the 10/80 pedosequence (8.3 and 2.0, respectively; Fig. 4) are more comparable to those of the dwarf shrubs E. nigrum and L. palustris than any other plant species or part, and their decrease over time toward lower values higher up the sequence suggests a progressive and concomitant decrease of the 6.Depth variations of ursolic acid, suggested Betulaceae biomarkers (sum of betulin, betulinic acid, lupeol and lupenone) and sitosterol/sitostanol ratio in palaeosols sequences (a) 10/80 and (b) 30/60.shrub cover along an increase in the grass cover.In addition, the CDI values encountered in the Ab3 and Ab4 palaeosols from the 10/80 pedosequence (2. 1 and 1.9, respectively; Fig. 4) are as low as those found in dwarf shrub leaves from Ronkainen et al. (2015, CDI ≤ 2.1 for B. nana, E. nigrum, L. palustris and V. uliginossum), while the higher CDI values of the youngest palaeosols tend towards CDI values encountered in grasses, mosses and roots.
In our soil sequences, both OEP and CPI of n-alkanes increase with depth (Fig. 4), which likely results from vegetation shifts rather than post-depositional degradation of the native biomarkers.Indeed, the preferential degradation of n-alkanes occurring in environmental archives is generally highlighted by a decrease in OEP and CPI with depth, even though such a decrease can also result from palaeovegetation changes (Thomas et al., 2021).Altogether, the trends of the abovementioned n-alkane-based proxies (OEP, C 31 /C 27 and C 33 /C 27 ratios and CDI) suggest a gradual decrease in the abundance of dwarf shrub species relative to grasses over time.
The overall increase with time in the fatty Index Grass along with the decrease in the Index Deciduous (Schäfer et al., 2016) in both pedosequences also suggest a shift in the palaeovegetation with a decreasing shrub cover for the benefit of grasses (Fig. 5).The increase in the aldehyde C 26 /C 28 ratio values observed in the two soil sequences from the bottom palaeosol to the top one might also indicate an increased contribution in grass lipids to SOM over time (Fig. 5), as grasses and herbs can contain higher proportions of C 26 compared to shrubs and trees, richer in C 28 and longer homologues (van Bergen et al., 1997).
The overall decrease in ursolic acid abundance in the two pedosequences from the oldest palaeosol to the youngest (Fig. 6) could also be attributed to a decreasing contribution of shrubs to palaeovegetation cover (Fisher et al., 2003).Finally, a decreasing trend in shrub cover over time was also observed in the pollen diagram from the ÎArte 6 onsite monolith (Supplementary material 8 in Anderson et al., 2019).As mosses can display significantly higher n-alkane C 23 /C 25 and C 23 /C 27 ratio values compared to other plant species (Ronkainen et al., 2015;Fig. 4; Supplementary Fig. S6), the increasing values of these proxies from the bottom to the top of our pedosequences argue for an increase in the contribution of mosses to palaeovegetation over time.However, as the values of these ratios are low in our soils compared to those of fresh mosses, the contribution of mosses to the palaeovegetation of the meadow studied might be limited in comparison to that of shrubs and grasses.
Altogether, these results suggest that the dwarf shrub species dominated the palaeovegetation at ÎArte 6 when the Ab4 palaeosol developed, and then that the plant cover evolved toward an increasing and higher contribution of grasses and mosses.
The concomitant occurrence of betulin, betulinic acid, lupeol and lupenone in environmental and archaeological samples has been used as an indicator for the presence of species from the Betulaceae family, which is dominated by shrubs and trees (Colombini and Modugno, 2009).Indeed, these lipids are found in high abundance in different birch subspecies compared to subspecies such as Alnus, Corylus and Carpinus and other families such as Salicaceae or Ericaceae (Supplementary material in Diefendorf et al., 2012).In the present study, the sum of the relative abundance of betulin, betulinic acid, lupeol and lupenone followed a trend different to the one of n-alkane proxies reflecting the overall shrub contribution (i.e.n-alkane-derived OEP, C 31 /C 27 and C 33 /C 27 ratios), suggesting that these compounds could record the specific variations of Betulaceae species.The fact that the variations of this specific Betulaceae proxy are not visible in the trends of more general shrub proxies suggests that the contribution of Betulaceae species to the global shrub lipid signature was probably low.This is in accordance with the current local predominance of shrubs like Vaccinium sp., Rubus sp. or Salix sp.(Supplementary Table S1), although the balance in shrub species distribution probably changed over time.The highest abundance of Betulaceae markers recorded in the third and the second palaeosols (Ab3 and Ab2; Fig. 6) could correspond to the high peak of Betula nana pollen displayed at 60 and 80 cm depth in the ÎArte 6 soil monolith (Supplementary Fig. S7a), all the more as the other shrubderived pollen (Alnus, Salix and different Ericaceae) did not show such variation.

Regional palaeoclimate reconstruction
Several global calibrations based on brGDGT distribution in soils (Weijers et al., 2007;Peterse et al., 2012;De Jonge et al., 2014) were applied to reconstruct mean annual temperature (MAT).The most recent brGDGT calibrations (e.g., De Jonge et al., 2014;Véquaud et al., 2022) rely on improved analytical methods allowing the separation of 5-and 6methyl brGDGTs.The exclusion of 6-methyl brGDGTs was shown to improve the accuracy of MAT reconstructions compared to previous global calibrations where the two types of homologues were not separated (cf. De Jonge et al., 2014).Nevertheless, even though 6-methyl brGDGTs are present in most soil samples discussed in the literature, including those previously analysed in our laboratory (e.g., Huguet et al., 2019), they are sometimes not detected (e.g., in 17 out of 239 soils reported by De Jonge et al., 2014), as was the case in the samples from the palaeosol sequences at ÎArte 6.Therefore, in the present case, whatever the calibration, the MAT reconstructions are based on 5methyl brGDGTs (Fig. 7).
The tentative MAT reconstructions based on the different calibrations showed similar trends from the oldest to youngest palaeosols in the ÎArte 6 meadow sequences (Fig. 7) and followed roughly the regional temperature anomalies reconstructed by Briffa et al. (2013) based on high-resolution dendroclimatological data (see Fig. 11 in Anderson et al., 2019).As with the brGDGTs (Fig. 7), in the 10/80 pedosequence, the variations of soil magnetic susceptibility, a palaeoclimate proxy often used in LPS contexts (Kravchinsky et al., 2008;Zeng and Yang, 2019), suggest that regional temperatures were slightly higher in the oldest and the youngest palaeosols than in the second and the third palaeosols developing in the ÎArte 6 meadow (Fig. 7).The trends observed in the 10/80 palaeosol sequence are also supported by pollen analysis in a regional record from a lake located a few kilometres southwest of ÎArte 6 (Lake Three; Anderson et al., 2019), where a decrease in grasses relative to shrubs occurred around 800 CE.This was interpreted as a result of an increase in air temperature (Supplementary Fig. S7b), which can be correlated to the transition between the Ab2 palaeosol (718-916 CE) and the Ab1 (838-1037 CE), and the corresponding MAT variations (Fig. 7).
Overall, the trends in palaeoclimate recorded by brGDGTs in both pedosequences suggest that a decrease in air temperature occurred between the 6th and 7th centuries CE, followed by warmer temperatures between the 8th and the 10th centuries CE (Fig. 7).
However, whatever the calibration, the absolute values of MAT inferred from brGDGTs in the palaeosols were too high compared to the modern MAT measured at the Mare-Sale weather station, which are ca.
-8 • C (150 km NW from the site, Vikhamar-Schuler et al., 2010).These differences between MAT estimates and actual temperatures could be the result of; (a) a lack of brGDGT-based calibrations for high latitude soil environments, (b) the impacts of various vegetation assemblages, suggested by other lipid biomarkers, on the calibrations of brGDGTbased temperature proxies (Liang et al., 2019), (c) a minor contribution of brGDGTs derived from rhizomicrobial association (Huguet et al., 2012; but see discussion in Section 4.1), and/or (d) brGDGT distribution being more representative of summer environmental conditions, when bacteria are more active (Huguet et al., 2013).Bearing this in mind, when a high latitude lacustrine calibration (Foster et al., 2016) allowing the reconstruction of mean summer temperature (MST) is applied to our soils, the reconstructed temperatures (increasing with depth from 4.4 • C to 8.9 • C in the 10/80 sequence, data not shown) fit better with the actual local summer temperatures (ca.7 • C at Mare-Sale, Vikhamar-Schuler et al., 2010).This stresses the need for the development of regional brGDGT calibrations based on modern datasets from comparable, high-latitude environments, and for the combination of brGDGTs with other temperature proxies, such as those based on chironomids (Langdon et al., 2011).However, the current results highlight the potential of GDGT-based proxies to provide information on past regional temperature variations in loess-palaeosol sequences, which has significant implications for palaeoclimate reconstructions and archaeological studies elsewhere.

ÎArte 6 palaeoreconstruction and history
Finally, the different proxies used in this study show a good agreement between the molecular signatures in the palaeosols and the regional palaeoenvironmental reconstructions (Panova, 2008;Briffa et al., 2013;Anderson et al., 2019).During the time covered by the ÎArte 6 meadow LPS, both the local lipid and pollen records show an increase in grasses relative to shrubs .This vegetation shift might be due to the cooling climate observed until the 9th century (Fig. 7; Schubert et al., 2019), but it may also have been promoted by the gradual intensification of human and reindeer presence at the site (Anderson et al., 2019).Although they may originally have used the shrubs of the ÎArte 6 headland as hunting hides (Anderson et al., 2019), the presence of willow and birch charcoal in buried hearths found at the site and in the palaeosol sequences adjacent at the site (especially at 30/ 60), and the trends observed for specific pentacyclic triterpenoids as potential dwarf birch biomarkers, indicate that the shrub cover was harvested for firewood.At the same time, the presence of congregating reindeer at different periods testified by the faecal lipid biomarker data (Anderson et al., 2019) may have decreased the shrub cover by repetitive trampling and selective grazing on dwarf shrubs, and by promoting grass growth over shrubs via urine and faeces fertilization as it is observed in modern contexts (Olofsson, 2006;Skarin et al., 2020).The late increase in air temperature inferred from brGDGTs (Fig. 7) would have been expected to promote shrubification in a tundra environment (Forbes et al., 2010).As an opposite trend was observed, this suggests that the effects of the impacts of reindeer activity at the site on the local vegetation cover, and especially shrubs, were most probably predominant over of the possible positive impacts of climate on the promotion of shrub species (Fig. 8).

Conclusions
In this study aiming to reconstruct the palaeoenvironment of a Siberian archaeological site, the similar lipid signatures of two distinct palaeosol profiles containing a dominance of plant-derived lipids (longchain n-alkyl compounds, phytosterols and pentacyclic triterpenoids) showed that human and animal activities had a greater impact on the local vegetation than regional climate change.This multi-proxy study included a rigorous evaluation of the plant-derived lipid preservation in the soil sequences studied.Although post-depositional bacterial degradation of plant lipids did occur, the values and trends with depth of the sitosterol/sitostanol ratio, the FA C 16+18 sat./unsat.ratio, FA EOP and nalkane OEP and CPI proxies suggested that this was limited.n-Alkane OEP and CDI proxy values also argue in favour of a limited impact of modern roots on the SOM, implying that SOM lipid signatures were representative of the palaeovegetation at the site.
The values and trends of different lipid-based palaeovegetation proxies (n-alkane CDI, OEP, C 23 /C 25 and C 23 /C 27 , C 31 /C 27 and C 33 /C 27 ratios, FA Index and n-aldehyde C 26 /C 28 ratio) from the oldest to the youngest palaeosols support the shift from an early predominance of dwarf shrub species to an increasing contribution of grasses and mosses, as also suggested by pollen analysis and the modern vegetation cover, although the different trends of some triterpene compounds (betulin, betulinic acid, lupeol and lupenone) imply independent temporal variation of Betulaceae shrub.
The trends in palaeoclimate recorded by brGDGT-based MAT reconstructions suggest that a decrease in air temperature occurred between the 6th and 7th centuries CE, followed by warmer temperatures between the 8th and 10th centuries CE.Temperature increases in the Arctic are normally associated with the promotion of shrub species over grasses, but in this case, the palaeovegetation proxies suggest the opposite.This bifurcation of local vegetation and regional climate trends is most likely to be the result of the occupation of the site by reindeer herders and their herds and their impacts on local vegetation, including grazing, trampling, the harvesting of shrubs for firewood, and the input of dung.
This study demonstrates that palaeovegetation and palaeoclimate reconstructions based on multi-proxy lipid biomarker analysis, when applied to archaeological contexts, are an important complement to routine archaeological and palynological studies, as they enable the distinction between climatic and anthropogenic impacts on local vegetation.Understanding past impacts of human occupation and reindeer herding on Arctic vegetation will facilitate the management of Arctic ecosystems in the wake of ongoing climate change.

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

Fig. 2 .
Fig. 2. Pedosequence description of soil test pits located at 10/80 and 30/60 (Fig. 1).Soil horizon notation corresponds to Jahn et al. (2006), A for buried palaeosol and C for wind-deposited sediment.Soil sampling for lipid analyses is described in detail in the Material and Method section.Radiocarbon dates are from Anderson et al. (2019).

Fig. 3 .
Fig. 3. Total ion current (TIC) chromatograms of the: (a) apolar and (b) polar fractions of a soil methodological replicate from the Ab4 horizon of sequence 30/60.C# shows the length of the n-alkyl chain.Compound names, formulae and other structural and spectral details can be found in Supplementary TableS3.

Fig. 5 .
Fig. 5. Depth variations of fatty acid-based proxies and the C 26 /C 28 n-aldehyde ratio in palaeosols from sequences: (a) 10/80 and (b) 30/60.Proxy calculations are detailed in the Materials and method section.