Habitats of Pleistocene megaherbivores reconstructed from the frozen fauna remains

703 –––––––––––––––––––––––––––––––––––––––– © 2020 The Authors. Ecography published by John Wiley & Sons Ltd on behalf of Nordic Society Oikos This is an open access article under the terms of the Creative Commons Attribution License, which permits use, distribution and reproduction in any medium, provided the original work is properly cited. Subject Editor: Eric Post Editor-in-Chief: Jens-Christian Svenning Accepted 4 December 2019 43: 703–713, 2020 doi: 10.1111/ecog.04940 43 703–713


Introduction
The Late Pleistocene landscapes in Eurasia and North America were inhabited by a specific mammalian complex represented most notably by woolly mammoths and other species of megafauna (Vereshchagin and Baryshnikov 1991, Markova et al. 1995, Kahlke 2014. At the end of the Pleistocene (around 10-15 kyr BP), most of the megafauna species went extinct (Stuart 2015). The reasons for such extinctions are complex, including a considerable climate and vegetation change coupled with an increasing human impact (Nogués-Bravo et al. 2008, Allen et al. 2010, Stuart 2015, Pavelková Řičánková et al. 2018), but specific processes and regional differences still remain unclear.
To get a better understanding of these extinction events, palaeoecologists are striving to reconstruct the habitat and vegetation types in which the Pleistocene megaherbivores were living using climate-based modelling (Allen et al. 2010, pollen (Tarasov et al. 2000), macrofossils (Sher et al. 2005) and recently also DNA analyses of plant remains  or stable isotopes (Rivals et al. 2010, Schwartz-Narbonne et al. 2019). However, each method of palaeovegetation reconstructions suffers from various constraints that can distort interpretations of the past landscape changes. Since the early 20th century, several well-preserved frozen carcasses, some of them being even around 50 000 years old , b, 2017, have been found scattered across northern Siberia and North America. This gives us a unique opportunity to examine the plant remains found in the gastrointestinal tract of Pleistocene megaherbivores, and to use these data as a proxy in the reconstructions of the Pleistocene vegetation and landscape .
We compiled two databases to obtain new insights into the Pleistocene vegetation of northern Siberia and Beringia. The first database contains information on plant remains identified in frozen megafauna individuals analysed to date, converted into standardized taxonomy and nomenclature, considering different taxonomic levels to which these plant remains could be identified. The second database contains modern records of plant species composition and cover-abundances in vegetation plots sampled across a broad range of the natural and semi-natural vegetation types in various regions in Siberia. When combined, these two databases make it possible to compare the Late Pleistocene megafaunal diet with current vegetation types of Siberia and to link each animal individual with a set of vegetation types in which it most probably lived shortly before its death. To do so, we developed a new analytical approach enabling a similarity analysis of datasets with different levels of taxonomical identification.
Our aim is to compare the composition of palaeobotanical remains identified in the frozen Pleistocene megaherbivores with modern vegetation types from Siberia and to use this information for reconstructing habitats of these animals and for improving the general reconstruction of the Pleistocene palaeovegetation of northern Siberia and Beringia.

Palaeobotanical records from the frozen fauna
We compiled plant records reported from the Pleistocene and early Holocene frozen megaherbivores found in northern Siberia, Alaska and the Yukon Territory (see Fig. 1, Supplementary material Appendix 2 Table A1 for site details). Frozen animals found in the interior parts of North America south of Alaska and the Yukon Territory were not considered in this study due to considerable differences in plant and animal species composition of that region. The palaeobotanical data included pollen + spores, macrofossils and DNA analyses reported from gastrointestinal tracts of frozen fauna or coprolites (for an overview see Supplementary material Appendix 2 Table A2).
Identification of the plant remains found in the frozen carcasses suffers from several constraints, and it is not always possible to identify plants at the species level (for methodological details see the original publications cited in Supplementary material Appendix 2 Table A2; for an overview of the most frequent plant taxa see Supplementary material Appendix 2 Table A3, A4). To make data on different taxonomical levels (species, genus and family) comparable, we included all these levels to the analyses simultaneously. For example, the species Abies sibirica was also included as the genus Abies and as the family Pinaceae. Only presence/absence data were used since relative abundances based on palaeobotanical remains suffer from various biases and in several cases, such data were not available at all. We unified plant taxonomy and nomenclature used in the original studies to follow Cherepanov (1995) for vascular plants and Ignatov and Afonina (1992) for bryophytes.
We analysed only plant records directly found within the frozen animal carcass or coprolite, but we did not consider plant remains sampled in the adjacent sediment. We decided so because 1) the data from the sediment were available only for a limited number of megafauna individuals, and 2) plant records from the sediment can be from a different time than the carcass and potentially affected by layer transpositions.
For all but one carcass, radiocarbon dating was available in the original publications (Table 1), which enabled us to sort them by age. In the case of several published age measurements, we used the mean value per animal. All values were then calibrated to the calendar age relative to 1950 using OxCal (Ramsey 2009;ver. 4.3), with calibration curve IntCal13 (Reimer et al. 2013) and probability interval of 95.4%. Following  and Willerslev et al. (2014), we used the periods of pre-LGM (50-25 kyr BP; called Kargin in northern Siberia or Wisconsinian Interstadial in North America), LGM (25-15 kyr BP; Sartan Ice Age/  Table 1. Radiocarbon dating of the megaherbivore individuals according to the originally published data. Most carcasses were found in Siberia; those from Alaska and Yukon are indicated by an asterisk*. Calendar age was calibrated relative to 1950 using OxCal; calibration curve IntCal13; probability 95.4%, † indicates a mean of more measurements. The dating for Vilyuy rhinoceros is approximate, based on the chronological groups used by Stuart and Lister (2012). The division into the periods follows  and Willerslev et al. (2014).

ID
Published nickname or (short name) Late Wisconsinian) and post-LGM (15 kyr BP-present; Late glacial and early Holocene).

Current vegetation records
We compiled a database of 1658 vegetation-plot records sampled by Masaryk Univ. teams in the years 2003-2014 across Siberia (southern Urals, western Siberian Plain, Salairskii Kryazh, Gornaya Shoriya, Altai, western Sayan and central Yakutia) and Kazakhstan (GIVD code 00-RU-002, Chytrý 2012), and additional 1181 records from various parts of Yakutia (Gogoleva et al. 1987, Mirkin et al. 1992, Telyatnikov et al. 2013 and unpublished data from north-eastern Yakutia by E. Troeva). All the records used in this study represent natural or semi-natural vegetation (anthropogenic vegetation was not included). All the plots were of the size of 100 m 2 except grassland plots in the Yakutian database that were of the size of 25 m 2 . Each plot contained a full record of vascular plants, terricolous bryophytes and macrolichens. All the vegetation-plot records (n = 2839) were assigned to the following broadly defined vegetation groups and types based on their species composition, ecological characteristics and overall physiognomy of the vegetation: group Blackish taiga: vegetation type Blackish taiga (= mixed coniferousdeciduous forest with nemoral species) (75 vegetation plots); Forests: Taiga (204); Hemiboreal forest (278), Alluvial forest and scrub (41), Fen and swamp woodland (51); Mires: Bog (65), Open fen (117); Arctic/alpine tundra: Arctic or alpine deciduous scrub (42), Arctic or alpine heathland (93), Tundra grassland (114) Like the palaeobotanical records, also the taxa in the current vegetation records were included in analyses at three identification levels, as species, genus and family, in order to make a valid comparison with the palaeobotanical data.

Analyses
Searching for the similarities between palaeobotanical records (frozen animal individuals) and current vegetation is difficult for several reasons: 1) single palaeobotanical records can be mixed samples of several vegetation types in unknown proportions, 2) some taxa in the palaeobotanical records are often identified at the genus, family or even higher taxon level, rather than at the species level as is the case of the current vegetation-plot data and 3) taxonomic spectrum of the palaeobotanical records is probably incomplete due to faster decomposition of the remains of some taxa. For these reasons, it is inappropriate to use any of the widely used similarity indices. Therefore, for both the vegetation plots and palaeobotanical records, we prepared presence-absence matrices that contained 1) all the species, 2) all the genera and 3) all the families of recorded taxa. For example, if a vegetation-plot record contained the species Poa pratensis, the presence was also recorded for the genus Poa and the family Poaceae. If a plant remain (pollen, macrofossil, DNA) was assigned only to a genus or family, its presence in the matrix was recorded only within the same and higher taxonomic rank(s). Then we calculated the number of shared plant taxa in each of the current vegetation plots and each individual of frozen megafauna, and interpreted a higher number of co-occurring taxa as a higher similarity. For each vegetation type, we calculated the mean value of the three most similar vegetation plots to the palaeobotanical record from each frozen animal individual. These mean values were used to sort the vegetation types from the most similar (with the highest number of shared taxa) to the least similar to the palaeobotanical record. This approach was repeated for three categories of palaeobotanical data: 1) separate data on pollen + spores, 2) separate data on macrofossils and 3) pollen + spores, macrofossils and DNA records merged into one data matrix.
To validate our new methodological approach, we tested whether the vegetation plots of a given vegetation type would be most similar to the original vegetation type they were previously assigned by experts. Also in this case, we prepared the matrices combining the species, genus and family levels for all the vegetation plots. Then for each plot, we separately counted the shared taxa between this plot and each of the other plots in the vegetation matrix, like in the case of palaeobotanical records. To calculate the mean number of shared taxa for each vegetation type, we used the three plots with the highest number of shared species; the vegetation type with the highest mean number of shared taxa was considered to be the most similar. The number of plots which had the highest similarity to the vegetation type they were originally classified to, was used as a measure of the reliability of the correct assignment. All the calculations were performed using the JUICE program .

Results
In the analysis based on pollen + spores (Fig. 2), the current vegetation types most similar to the palaeobotanical records were forests (especially taiga and hemiboreal forest), followed by several types of tundra vegetation. Steppe vegetation types showed much lower similarity (with a few exceptions). In contrast, in the analysis based on macrofossils only (Fig. 3), tundra vegetation (including both grasslands and shrublands) was among the most similar vegetation types. The similarity of palaeobotanical records to taiga and hemiboreal forests decreased from the pre-LGM phase (Kargin/Wisconsinian Interstadial; 50-25 kyr BP) through the Last Glacial Maximum (LGM/Sartan/Late Wisconsinian; 25-15 kyr BP) to the post-LGM phase (Late glacial and early Holocene; < 15 kyr BP). However, the post-LGM records were mostly limited to the northernmost parts of the study area, which corresponds to the current tundra zone. An analysis combining pollen + spores, macrofossils and DNA (Fig. 4) showed a pattern similar to that based on pollen + spores only. All the analyses indicated rather wet landscape covered predominantly by vegetation types that require mesic or wet soil. Present-day vegetation of Siberian dry habitats showed a low similarity to the composition of plant material obtained from the frozen megaherbivore remains. The lowest similarity was found for dry steppes, semi-deserts and deserts.
Comparison of the plant remains of different age ( Fig. 2-4) did not reveal any substantial difference in the spectrum of most similar/dissimilar vegetation types over time. There were almost no differences between the remains from the LGM (no. 18-22) and the previous and subsequent periods, with the exception of the above-mentioned changes in the similarities to forest vs tundra found in the macrofossil data.
Validation test of our method of similarity calculation revealed that 59% of the vegetation plots were assigned with the highest similarity to the original vegetation type and 78% of the plots had the highest similarity to a vegetation type in the vegetation group which the original type belongs to (Supplementary material Appendix 2 Table A5). As we considered only the highest similarity (not the second or third highest), this test was rather conservative. Most types of forest, tundra and grassland vegetation were classified successfully.
Less satisfactory results were found in the assignment of Fen and swamp woodland, which were often classified as bogs or other forest types, but mostly within the original Forests group. Tall forbs were also frequently misclassified to different forest types. Steppe scrub was most similar to both the original type and Dry steppe, both belonging to the same vegetation group. Finally, Desert steppe and Semi-desert were often misclassified as Dry steppe. Other vegetation types had 40-80% plots correctly assigned to the original vegetation type. The classification success of the vegetation groups ranged between 52 and 92% for all but one group, the Deserts, where only 32% of plots were correctly assigned to the original group.

Late Pleistocene vegetation of northern Siberia and Beringia
Our analyses suggest that nearly all the individuals of Late Pleistocene megaherbivores preserved in the permafrost of northern Siberia and Beringia grazed or browsed predominantly in mesic, moist and wet habitats, in landscapes with vegetation mosaic containing both tundra vegetation and woodlands along with other vegetation types. This pattern  Table 1 for details). Most of the finds are from Siberia, while those from Alaska and the Yukon Territory are indicated by an asterisk. NA indicates the animals for which this type of palaeobotanical records is not available. was relatively stable over the time span from > 50 kyr to 9 kyr BP, i.e. from the pre-LGM period through the LGM to the early Holocene ( Fig. 2-4). Palaeoclimatic data and previous palaeobotanical finds suggest that in the pre-LGM period, the climate of northern Siberia and Beringia oscillated between warmer and colder phases. This affected north/south distribution limits of the zonal vegetation types, i.e. different types of taiga and tundra (Hopkins et al. 1982, Zazula et al. 2006, Troeva et al. 2010. Drier types of meadows or steppes probably developed only under specific mesoclimatic conditions (Troeva et al. 2010. In general, our results are similar to the reconstructions proposed in the previous literature, however, both macrofossils and to a lesser extent also pollen data indicate higher representation of tundra vegetation, while previous reconstructions suggested tundra to be limited to maritime areas (Troeva et al. 2010). It is possible that tundra was not covering the landscape continuously; it may have occurred in discontinuous patches on windexposed elevations or in wet depressions (compare Yurtsev 2001, Chytrý et al. 2019.

Macrofossils
In contrast, we cannot support the previous reconstructions for the LGM. The LGM was a period of cold, arid and relatively stable climate when north-eastern Siberia was connected with North America through the Beringian Land Bridge (Hopkins et al. 1982). This past connection is reflected in the current distribution of flora (Yurtsev 1982, Guthrie 2001) and fauna (Hopkins et al. 1982, Guthrie 2001. Evidence gained mainly from pollen data and mammalian fossil finds led scientists to interpret the LGM landscape as the treeless 'mammoth steppe'  or a mosaic of different treeless habitats known as the 'steppetundra' (Hibbert 1982, Yurtsev 1982, Blinnikov et al. 2011) occurring over a huge area from western Europe to interior Alaska. However, our comparative analysis of pollen data suggests that representation of the main vegetation types did not change substantially between the pre-LGM and LGM periods, indicating forest continuity at least at some topographically suitable mesic or warmer sites (see also Brubaker et al. 2005, Zazula et al. 2006. Vegetation with the highest similarity to our LGM palaeobotanical data included different types of forests, especially taiga (probably light woodlands with Larix and Betula), and tundra. In contrast, the similarity to steppe vegetation types was lower than we could expect if the predominant ecosystem were the steppe-tundra complex. Only recently, Chytrý et al. (2019) proposed the Altai Mountains ecosystem in southern Siberia to be a close modern analogue of the LGM steppetundra biome, supporting their suggestion by multi-proxy evidence based on flora, fauna and long-term climate stability. They described the steppe-tundra as a predominantly treeless landscape mosaic of several habitat types (including woodlands) depending on precipitation and topography-related distribution of moisture across the landscape. However, there are several differences between the mountain systems of southern Siberia and northern Siberia and Beringia. Firstly, northern Siberia and Beringia currently lack many steppe species typical of European and southern Siberian LGM steppe-tundra (Yurtsev 1982, Chytrý et al. 2019. Similarly, many typical species of the LGM mammalian fauna have never occurred in the north (Pavelková Řičánková et al. 2014(Pavelková Řičánková et al. , 2018. Secondly, Beringia is affected by a maritime climate, which probably enabled suitable conditions for mesic plant survival also during the otherwise cold and dry LGM (Hopkins et al. 1982, see also the dynamic vegetation model by Allen et al. 2010). Last but not least, the current landscape of northern Siberia and Beringia is characterized by thermokarst processes and relatively low evapotranspiration, supporting mesic habitats despite low precipitation rates. It seems that despite the expectations of a vast biome of the Pleistocene mammoth steppe ranging from Europe to Alaska , northern Siberian and Beringian landscape might have been different from southern areas with relatively stable dry climate. In agreement with our conclusions, palynological evidence of crater-lake sediments also led Lozhkin and Andersson (2006) to propose a prevalence of tundra with some local woodland refugia in northeastern Siberia and Beringia. Similarly, based on palynological, macrofossil and insect evidence, Elias et al. (1996Elias et al. ( , 1997 reconstructed widespread mesic tundra in central parts of Beringia during the Late Pleistocene, while Goetcheus and Birks (2001) provided evidence for dry steppe-tundra occurrence at other sites in central Beringia during the LGM. All of these findings seem to support the existence of a vegetation mosaic with both wet and dry habitats, depending on the landscape topography.
Palaeoecological reconstructions of the post-LGM period refer to increases in both humidity and temperature, which were reflected in the Early-Holocene spread of forests and establishment of bogs . Nowadays the zonal vegetation of northern Siberia and Beringia is tundra and taiga, whereas steppe can be found only azonally, depending on the landscape topography (Edwards and Armbruster 1989, Walker et al. 1991, Yurtsev 2001, Troeva et al. 2010, Reinecke et al. 2017. With considerable climate and vegetation changes at the Pleistocene/Holocene transition, herbivores were forced either to retreat to areas with more suitable conditions (Pavelková Řičánková et al. 2015) or to change their habitats and diets significantly as was evidenced, e.g. by completely changed dental wear of reindeer and moose (Rivals et al. 2010), and by changes in carbon and nitrogen isotopes of European bison, aurochs (Bocherens et al. 2015), and caribou and saiga (Schwartz-Narbonne et al. 2019).

Methodological considerations
Like in the case of other palaeoecological studies, possible sampling biases and taphonomic issues have to be considered when interpreting our results. Firstly, frozen fauna finds are bound to the permafrost zone of northern Siberia and northern North America. Palaeoclimate studies suggest that this area had specific features that differed from other full-glacial landscapes (Allen et al. 2010). Therefore, conclusions based 710 on our results cannot be extrapolated to the Pleistocene landscapes in Europe or southern Siberia.
Secondly, only specific environmental conditions allow the preservation of soft tissues for more than fifty thousand years. Most of the animals preserved as frozen carcasses probably died near water bodies where they looked for water or fresh and productive vegetation (Haynes 1991, Kaczensky et al. 2008, Zhang et al. 2015, and thermokarst processes enabled their preservation (Vereshchagin and Tomirdiaro 1999). Therefore, the high similarity of our palaeobotanical data with wet to mesic vegetation types can be partly caused by the fact that shortly before they died, the studied animals grazed in wet areas, while the surrounding landscape could have been much drier.
Thirdly, although different types of plant remains indicated the same general pattern (high similarity to wet and mesic habitats), there were some differences. Taxon spectra from pollen analyses were more similar to forest vegetation, especially to taiga and hemiboreal forest (Fig. 2, 4), while macrofossils mostly indicated treeless vegetation (Fig. 3). Pollen analysis reflects the presence of trees in the wide surroundings of the study site due to wind transport of pollen, and consequently, it can suggest a more forested landscape than it actually was. However, repeated findings of Larix macrofossils can also support our conclusions about woodland refugia. Both macrofossils and also DNA represent mainly local plants eaten by the animals. However, as the animals may have been concentrated in more productive habitats in the landscape or around water sources (Hopkins et al. 1982), the reconstruction based on the macrofossils and DNA may provide incomplete picture of the vegetation types in the landscape.
Finally, the interpretation of vegetation from plant remains in frozen animals can be biased by feeding preferences of these animals. Deducing from the number of megafauna finds , it is supposed that northern Siberia and Beringia offered suitable habitats to large populations of megafauna over most of the Pleistocene , Zimov et al. 1995. Although low-productive habitats might have been common in the landscape, we can expect that animals preferentially selected more productive vegetation bound to mesic or wet sites (see also Chytrý et al. 2019, or recent meta-analysis of stable isotopes from mammoth bone collagen by Schwartz-Narbonne et al. 2019). In general, the biology and feeding ecology of extinct megaherbivores can be reconstructed based on the morphology of particular species, biology and ecology of their closest living relatives, although with some degree of uncertainty. The traditional assumption that megaherbivores, especially woolly mammoths, were only grazers and their diet consisted mainly of sedges and grasses (Guthrie , 2001, was already questioned by several authors, who confirmed the surprisingly high abundance of digested twigs ) and suggested a dietary shift to mixed feeding when the supply of forbs and graminoids was limited (Rivals et al. 2010, Tiunov and. Considering our data set, extant reindeer, the last wild horse Equus przewalskii and possibly feral horses E. caballus could partly corroborate our results. The habitat and food preferences of extant reindeer fit well to our results (Heggberget et al. 2002). Horses are predominantly grazers with a variable proportion of trees and shrubs in their diet (Putman 1986, Kuiters et al. 2006, Kaczensky et al. 2017, Cromsigt et al. 2018. The wild horse Equus przewalskii is considered to be a grazer within desert-steppe habitats, but the isotope evidence suggested a dietary shift from a mixed feeder in winter and grazer in summer to a year-round grazer due to the human persecution (Kaczensky et al. 2017). When its known Pleistocene distribution (Deng 2006) is compared with an LGM vegetation model (Wang et al. 2017), it occupied many various habitats including forested landscapes. The ecological versatility of horses seems to be also supported by their Holocene distribution (Groves 1974, Kuzmina 1997, Naundrup and Svenning 2015. To conclude, our results are fully compatible with the current knowledge about the diet and habitat preferences of reindeer and horses, suggesting that we provide realistic estimations of the megaherbivore habitat preferences.

Conclusions
Our study presents the vegetation mosaic of the Pleistocene landscape of northern Siberia and Beringia from the megaherbivore point of view. Although the plant data from frozen animal carcasses cannot be used to describe the whole picture of the mammoth steppe, we summarize and complement previous finds and interpretations based on pollen, macrofossil and DNA data, palaeoclimate modelling and other proxies. In contrast to more southern areas covered by the mammoth steppe in the Late Pleistocene, we suggest that northern Siberia and Beringia were characterized by a significant representation of mesic and wet habitats (including tundra, mires and woodlands) at that time.

Content
Appendix 1 Brief characterization of the vegetation types used in this study.

Appendix 2 Table A1
Localities of the frozen fauna finds.

Table A2
Overview of the palaeobotanical data from frozen fauna.

Table A3
Most frequent taxa recorded in the frozen fauna: pollen + spores.

Table A4
Most frequent taxa recorded in the frozen fauna: macrofossils.

Table A5
Verification test of the robustness of the similarity calculation method used in this study.

Appendix 1.
Brief characterization of the vegetation types used in this study with the lists of diagnostic, constant and dominant species. Some descriptions are partly or entirely taken from . All descriptions are based on the expert knowledge of the authors and the references listed below. Species names follow Cherepanov (1995). Species were considered diagnostic for a given vegetation type when their fidelity value measured as the phi coefficient of association was higher than 0.25 and corresponding Fisher's exact test indicated a significant concentration of species in the vegetation type at the level of 0.05. Highly diagnostic species (highlighted in bold) were those with phi coefficient higher than 0.40. As the fidelity values might be affected by differences in species numbers between vegetation types, we used standardization to equal sizes of all groups before calculation (Chytrý et al. 2002). Constant species or highly constant species (highlighted in bold) are those with a frequency in the dataset over 40% or 50%, respectively. Dominant species were those that reached higher cover values than 25% in at least 5% of plots. All analyses were performed in Juice software  Blackish (in Russian chernevaya) taiga is a forest composed of a mixture of coniferous and deciduous trees and both nemoral and boreal species in the herb and shrub layers. They occur especially in the southern subtaiga and southern-taiga zone of western Siberia, but they also occur in the precipitation-rich northern front ranges of the southern Siberian mountain systems such as the Altai and smaller mountain ranges north of the Altai. These forests are dominated by Abies sibirica, Betula pendula, B. pubescens, Populus tremula, with an occasional admixture of Picea obovata, especially on the valley bottoms. Unlike in the boreal forests, herb layer is often rather dense and dominated by herbs and grasses, whereas dwarf shrubs, as well as mosses and lichens, are less abundant or even absent. Peatland forests occur on the valley bottoms or in shallow depressions that are saturated with water from lateral groundwater flow, as in the case of minerotrophic fens. Dominant trees include Betula pubescens, Picea obovata and Pinus sylvestris. Herb layer is rich in grasses (e.g. Calamagrostis canescens and C. langsdorffii), sedges (e.g. Carex cespitosa and C. juncella) and bryophytes including some species of Sphagnum.

Bog
(65 vegetation plots) These are rainwater-fed mires occurring mainly in the lowlands of the boreal zone of Siberia, especially on the West Siberian Plain. They are dominated by peat mosses (Sphagnum spp.) and contain a significant amount of dwarf shrubs such as Chamaedaphne calyculata, Ledum palustre, Rubus chamaemorus, Oxycoccus spp. and Vaccinium uliginosum). They are either open or covered by sparse stands of Pinus sylvestris. In the permafrost zone of northern Siberia, they often form small elevations with ice cores (palsas).

Open fen
(117 vegetation plots) This type of open mire, occurring in valleys, shallow depressions or around springs, is saturated by groundwater. It is dominated by sedges (e.g. Carex chordorrhiza, C. diandra, C. lasiocarpa and C. rostrata), herbs (e.g. Comarum palustre and Menyanthes trifoliata) and mosses, although the species of the genus Sphagnum are less abundant than in ombrotrophic bogs or even absent. Arctic and alpine heathland occurs especially in the tundra zone of northern Siberia as well as in the areas above the timberline of the Siberian mountain systems. It is dominated by dwarf shrubs (e.g. Empetrum spp., Vaccinium myrtillus and V. vitis-idaea) associated with perennial herbs, graminoids, and with significant participation of bryophytes and lichens.

Tundra grassland
(114 vegetation plots) Mesic meadows typically occur on river terraces which are rarely flooded. Their species composition consists of both graminoids and dicot herbs. Productivity is lower than in the wet meadows, but the swards are often dense and tall and used for grazing or haymaking by local farmers.