Vegetation on mesic loamy and sandy soils along a 1700‐km maritime Eurasia Arctic Transect

Abstract Questions How do plant communities on zonal loamy vs. sandy soils vary across the full maritime Arctic bioclimate gradient? How are plant communities of these areas related to existing vegetation units of the European Vegetation Classification? What are the main environmental factors controlling transitions of vegetation along the bioclimate gradient? Location 1700‐km Eurasia Arctic Transect (EAT), Yamal Peninsula and Franz Josef Land (FJL), Russia. Methods The Braun‐Blanquet approach was used to sample mesic loamy and sandy plots on 14 total study sites at six locations, one in each of the five Arctic bioclimate subzones and the forest–tundra transition. Trends in soil factors, cover of plant growth forms (PGFs) and species diversity were examined along the summer warmth index (SWI) gradient and on loamy and sandy soils. Classification and ordination were used to group the plots and to test relationships between vegetation and environmental factors. Results Clear, mostly non‐linear, trends occurred for soil factors, vegetation structure and species diversity along the climate gradient. Cluster analysis revealed seven groups with clear relationships to subzone and soil texture. Clusters at the ends of the bioclimate gradient (forest–tundra and polar desert) had many highly diagnostic taxa, whereas clusters from the Yamal Peninsula had only a few. Axis 1 of a DCA was strongly correlated with latitude and summer warmth; Axis 2 was strongly correlated with soil moisture, percentage sand and landscape age. Conclusions Summer temperature and soil texture have clear effects on tundra canopy structure and species composition, with consequences for ecosystem properties. Each layer of the plant canopy has a distinct region of peak abundance along the bioclimate gradient. The major vegetation types are weakly aligned with described classes of the European Vegetation Checklist, indicating a continuous floristic gradient rather than distinct subzone regions. The study provides ground‐based vegetation data for satellite‐based interpretations of the western maritime Eurasian Arctic, and the first vegetation data from Hayes Island, Franz Josef Land, which is strongly separated geographically and floristically from the rest of the gradient and most susceptible to on‐going climate change.

Here we describe the vegetation along the 1700-km Eurasia Arctic Transect (EAT) that includes the Yamal Peninsula and Franz Josef Land ( Figure 1). The aim is to characterize vegetation on zonal loamy and sandy soils along the complete maritime Arctic climate gradient in western arctic Russia to aid in remote-sensing interpretations of land-cover and land-use change . The zonal patterns, geological conditions, permafrost and summer thaw depth (active layer) conditions are generally well described along the length of the peninsula. We analyse the variations in plant growth forms and species richness in each layer of the plant canopy with respect to summer temperature and soil texture, present a preliminary numerical classification and use indirect ordination methods to analyse the relationship of the plots and species to a suite of measured environmental factors. Management (BOEM); Slovak Academy of Science, Grant/Award Number: 2/0135/1; U.S. National Aeronautics and Space Administration, Land-Cover Land-Use Change Program, Grant/Award Number: NNG6GE00A, NNX09AK56J and NNX14AD90G Co-ordinating Editor: Borja Jiménez-Alfaro Methods: The Braun-Blanquet approach was used to sample mesic loamy and sandy plots on 14 total study sites at six locations, one in each of the five Arctic bioclimate subzones and the forest-tundra transition. Trends in soil factors, cover of plant growth forms (PGFs) and species diversity were examined along the summer warmth index (SWI) gradient and on loamy and sandy soils. Classification and ordination were used to group the plots and to test relationships between vegetation and environmental factors.
Results: Clear, mostly non-linear, trends occurred for soil factors, vegetation structure and species diversity along the climate gradient. Cluster analysis revealed seven groups with clear relationships to subzone and soil texture. Clusters at the ends of the bioclimate gradient (forest-tundra and polar desert) had many highly diagnostic taxa, whereas clusters from the Yamal Peninsula had only a few. Axis 1 of a DCA was strongly correlated with latitude and summer warmth; Axis 2 was strongly correlated with soil moisture, percentage sand and landscape age.
Conclusions: Summer temperature and soil texture have clear effects on tundra canopy structure and species composition, with consequences for ecosystem properties.
Each layer of the plant canopy has a distinct region of peak abundance along the bioclimate gradient. The major vegetation types are weakly aligned with described classes of the European Vegetation Checklist, indicating a continuous floristic gradient rather than distinct subzone regions. The study provides ground-based vegetation data for satellite-based interpretations of the western maritime Eurasian Arctic, and the first vegetation data from Hayes Island, Franz Josef Land, which is strongly separated geographically and floristically from the rest of the gradient and most susceptible to on-going climate change.

K E Y W O R D S
above-ground biomass ordination, Arctic, bioclimate subzones, Braun-Blanquet classification, DCA ordination, Normalized Difference Vegetation Index, plant growth forms, remote sensing, soil texture, summer warmth index, tundra biome WALKER Et AL.

| Site selection and sampling
We established the EAT during four expeditions in the summers Mean July temperatures range from 1°C at the northern end of the transect to 15.8°C at the southern end. Six study locations were selected along the EAT to represent zonal (Razzhivin, 1999;Walter, 1954Walter, , 1973 vegetation conditions in each of the five Arctic bioclimate subzones and the forest-tundra transition, as mapped on the Circumpolar Arctic Vegetation Map (Walker et al., 2005;Yurtsev, 1994b; Table 1). At each location we chose at least two study sites -one on mesic loamy soils and one on mesic sandy soils (see Supporting Information Appendix S1 for geological setting in relationship to soils).
We used the Braun-Blanquet approach (Westhoff & Van der Maarel, 1978) to sample mesic loamy and sandy sites at each location. At most study sites there was adequate space for a large relatively homogeneous 50 m × 50 m sample site that corresponded approximately to the 30-m to 70-m pixel size of the Landsat satellite sensors. Sample plots and transects were arranged in the pattern shown in Supporting Information Appendix S2. Here we describe the data mainly from 5 m × 5 m (25 m 2 ) plots, except at the Nadym forest site, where 10 m × 10 m (100 m 2 ) plots were used, and the Nadym tundra site, where 1 m × 1 m (1 m 2 ) plots were used to sample homogeneous areas of vegetation on patterned ground features (earth hummocks). We sampled 79 plots, but eliminated three Nadym wetland plots, resulting in a final data set of 76 plots, distributed among the six EAT locations: Krenkel (KR, ten plots), Ostrov Belyy (BO, 20 plots), Laborovaya (LA, ten plots), Kharasavey (KH, ten plots), Vaskiny Dachi (VD, 15 plots) and Nadym (ND, 11 plots) (see Supporting Information Appendix S3 for descriptions and photographs of the study sites.) Each vascular plant, bryophyte and lichen species occurring within a plot was recorded and a sample taken as a voucher.
The environmental data from each plot include 107 variables, including site, soil, biomass, spectral data, NDVI and canopy structure variables. (see details in, Supporting Information Appendices S5.1 and S5.2, and the project data reports; Walker, Carlson, et al., 2011;Walker, Orekhov, et al., 2009). Soils samples were collected from the uppermost mineral soil horizons at a point just outside the southwest corner of each vegetation plot. Larger soil pits were dug just outside the southwest corner of the 50 m × 50 m grid to fully describe vertical and horizontal variation in the soil profiles. The pits were described by Dr.
Georgy Matyshak according the Russian approach and translated into descriptions corresponding to the US Soil Taxonomy approach (Soil Survey Staff, 1999) and are included with photographs in the data reports cited above.

| Climate
The Arctic bioclimate zonation patterns portrayed on the Circumpolar Arctic Vegetation Map (CAVM Team et al., 2003) are based primarily on summer temperature regimes and structure of the vegetation (Yurtsev, Tolmachev, & Rebristaya, 1978;Yurtsev, 1994a). We use the summer warmth index (SWI), which is the sum of monthly mean temperatures above 0°C, measured in °C month "thawing degree months". The SWI is calculated from monthly mean temperature data and is very strongly correlated with thawing degree days, which require daily mean temperature to calculate. SWI is equivalent to the warmth index, a, used by Steve Young for the vascular plant flora of St. Lawrence Island, Alaska (Young, 1971).
Four of the six EAT locations have long-term climate station data; for these locations, we calculated the SWI for air temperatures (SWI a ) at the standard 2 m height of weather station observations. To obtain consistent summer temperature data for all study locations over the same length of record, we used data from the thermal infrared channels of satellite-based Advanced Very High Resolution Radiometers (AVHRR, years 1982(AVHRR, years -2003Comiso, 2003Comiso, , 2006 to calculate SWI g , the ground surface summer warmth index (SWI g ) within 12.5-km pixels containing the study locations (Bhatt et al., 2010). Consistent data for other climate factors, such as precipitation and wind, were not available across all study locations.

| Cluster analysis
We used a hierarchical dendrogram approach, available in PC-ORD to group the plots into clusters based on the similarity of their species compositions (MjM Software, Gleneden Beach, OR, US) via the JUICE 7.0 software (Tichý, 2002). The most meaningful separation of the 76 plots was achieved with the flexible beta group linkage method (β = −0.25) with the Sørensen distance measure and square root data transformation. We included species-level taxonomic determinations in the analyses, and we excluded taxa that were identified only to the genus level. To determine the optimal number of clusters providing the highest 'separation power' for the data set, we used the Crispness of Classification approach (Botta-Dukát, Chytrý, & Hájková, 2005) available through the Optimclass function in JUICE (Tichý, 2002). A synoptic table was prepared using the combined synoptic table function in JUICE. Taxa with high fidelity (modified phi coefficients ≥ 0.5) were interpreted as diagnostic for the group; taxa with very high fidelity (modified phi coefficients ≥ 0.8) were interpreted as highly diagnostic.

| Analysis of vegetation and environmental variables
We compared the trends of plant growth form (PGF) cover along the bioclimate gradient (SWI g ) for each layer of the plant canopy (tree and shrub layer, herb layer and cryptogam layer); and the species richness within groups of dominant PGFs (deciduous shrubs, evergreen shrubs, graminoids, forbs, mosses, lichens). We also examined trends of soil properties along the bioclimate gradient.

| Ordination
We explored several ordination methods available in the R pro-

| Descriptions of the EAT locations and study sites
An overview of the study sites (Table 1)  Clay, silt and sand percentages for loamy and sandy sites are shown using the US Department of Agriculture soil texture triangle ( Figure 2a). Loamy sites had 19%-61% sand and 31%-62% silt.
Sandy sites generally had >80% sand, and <20% silt. Clay percentages were low (<25%) at all sites. On the loamy sites, silt and clay percentage were somewhat higher in the central part of the summer temperature gradient. Sand percentages were higher at both ends of the gradient (Figure 2b).

| Classification and syntaxonomic interpretation
The cluster analysis dendrogram shows the progressive linkage of plots according to their floristic similarity (Figure 3). Clusters with TA B L E 2 Temperature and precipitation along the Eurasia Arctic Transect. Mean (1961Mean ( -1990 July temperature and precipitation data (columns 3 and 4) are from the nearest relevant climate stations. Summer Warmth Index (SWI) is the sum of the monthly mean temperatures above freezing. The mean atmospheric SWI (SWI a ) (column 5) is calculated from the mean ) station data, where available. Ground Summer Warmth Indices (SWIg) (column 6) are calculated from AVHRR thermal bands for the 12.5-km pixels containing the EAT study locations. Value for SWI g in the circumpolar Arctic subzones (column 7) are calculated using all circumpolar pixels within each subzone  Bioclimate subzone EAT study location   (Tichý, 2002), which resulted in the six optimal clusters (red numbers). The red line is where the line was adjusted to separate out cluster 6, which based on field observations was distinct from cluster 5. Background colours correspond to the bioclimate subzones (A to Forest-tundra). Also shown are loamy and sandy groups of plots (black Roman labels), and micro-topographic groups of plots in patterned ground complexes (italics) higher levels of similarity are toward the left side of the diagram.
Crispness of Classification identified two clusters with the highest level of separability (dissimilarity). One cluster contained all of the Yamal plots (subzones B, C, D and E) and the other contained all the plots of FJL (subzone A) and Nadym (FT transition). The next highest level of dissimilarity was achieved with six clusters, separated at the level of the red dashed line in Figure 3. At this level, clusters 5 and 6 in Figure 3 were joined, forming one large cluster containing most of the plots on the Yamal Peninsula, including the subzone D loamy plots, all subzone C plots and the subzone B loamy plots. Based on our knowledge of the rather unique floristic character of the loamy subzone B site, which has characteristics similar to the moist non-

| Soils, vegetation structure and species richness
Trends of key soil and key vegetation canopy factors (canopy layer height, litter, standing dead, LAI, NDVI, total phytomass) vs. SWI g are in Supporting Information Appendix S8.
Soil properties that increase with higher SWI g include percentage sand (on sandy sites), thickness of organic horizons, percentage soil carbon (on loamy sites) and active layer thickness (Supporting Information Appendix S8, Figure S8-1). Soil properties that tend to

| Ordination
The DCA plot ordination (Figure 5a

| Mesic vegetation transitions along the EAT summer temperature gradient
A primary motivation for this study was to develop a baseline of ground-based vegetation information along the complete Arctic summer temperature gradient in the maritime Arctic portion of western Russia to support remote sensing interpretations. We sampled and analysed plant communities on homogeneous mesic sites with loamy and sandy soils along the summer temperature gradient of the EAT. Satellite-derived summer land-surface temperatures (Comiso, 2006;Raynolds, Comiso, Walker, & Verbyla, 2008) provided a consistent spatial record of mean summer ground-surface temperatures (SWI g ) across the full length of the EAT, including locations where station data were unavailable.
The EAT analysis focused on mesic tundra areas where climate is the primary factor controlling the character of the vegetation.
Although we initially considered these mesic sites to be zonal habitats, it soon became clear that the tundra over nearly the entire Yamal Peninsula is strongly influenced by a long history of reindeer grazing. The only locations that were free of recent reindeer foraging were Krenkel and Nadym at the extreme northern and southern ends of the bioclimate gradient. Both of these sites had high cover of lichens, indicating that reindeer at the other sites have greatly reduced the lichen cover. Reindeer herds graze heavily on lichens particularly during the snow-covered months of winter and spring. The results of our study and others (Pajunen, 2009;Pajunen, Virtanen, & Roininen, 2008;Vowles, Lovehav, Molau, & Björk, 2017;Yu, Epstein, Walker, Frost, & Forbes, 2011) and comparison with results from a similar transect in North America where there are relatively low Rangifer densities  indicate that the reindeer have had a long-term major impact on the shrub, graminoid and moss layers on the Yamal (Forbes et al., 2009). Quantifying this effect is difficult because of lack of reindeer exclusion areas.  (Elmendorf et al., 2012;Matveyeva, 1998), (b) increases in vascular plant cover and diversity along the summer temperature gradient (Daniëls et al., 2013;Rannie, 1986;Young, 1971), and (c) exclusion of woody plants, sedges and Sphagnum peat from the northernmost subzone A (Yurtsev, 1994b). While cover and species richness of evergreen and deciduous shrubs generally increased with higher SWI g , cover of lichens and forbs declined. Graminoid cover and species richness of lichen and bryophyte species richness showed parabolic trends with maximum values in the central part of the temperature gradient.
Much recent research regarding productivity patterns in the Arctic has focused on the increased abundance of shrubs associated with warming temperatures, which are thought to be a primary cause of the recent increases in NDVI observed in satellite data (Myers-Smith et al., 2011). Our study documented strong, mostly positive, exponential trends with SWI g for deciduous and evergreen shrub cover, shrub layer height, herb layer height, litter cover, LAI, NDVI and above-ground phytomass. The study also documented the dominance of shrubs in the Low Arctic (subzones E and D), dwarf shrubs, graminoids and bryophytes in the Middle Arctic (subzones C and B), and forbs and crustose lichens in the extreme High Arctic.

| The role of soil texture
The floristic contrast between the loamy and sandy sites varies considerably between locations across the EAT, a result of much greater site-factor heterogeneity of the sandy sites. The Nadym and Ostrov Belyy locations illustrate rather extreme contrasts in ecosystem structure that can occur on loamy vs. sandy soils. At Nadym, the site on the sandy, relatively young surface at ND-1 is relatively well drained, has no permafrost and is forested; whereas the ND-2 site on older, more fine-grained soils is ice-rich, relatively poorly drained, and covered with hummocky tundra vegetation (Supporting Information Appendix S3, Figure S3-6). A host of site factors interact to affect the vegetation structure and composition at this site, including much thicker soil organic layers, thin active layers, relatively cold soils and very low CECs on the older loamy soils. A similar contrast occurred at Ostrov Belly (Supporting Information Appendix S3, Figure S3-2) and is illustrated in the numerical classification and DCA ordination, where the sandy and loamy plots are placed in separate clusters (Figure 3, clusters 6 and 7) and are widely separated along Axis 2 of the ordination (Figures 3 and 5). The sandy sites at Ostrov Belyy are much drier than the loamy sites at this location and have many other site factor differences that separate them.
The opposite situation occurs at Krenkel (subzone A; Supporting Information Appendix S3, Figure  Part of the explanation for much larger variation in the sandy sites is that during site selection, it was relatively easy to find large sites to sample vegetation on mesic silt loam to sandy loam soils, whereas the availability of mesic very sandy sites was more limited.
The relatively young sandy sites are also more susceptible to disturbance by reindeer and strong winds, whereas the older loamy sites have tended to stabilize toward the regional zonal conditions.

| Special importance of subzone A
A major accomplishment of this study was the first detailed veg- unlike any other site along the EAT. Sites not exposed to excessive wind erosion had unexpectedly high hand-held NDVI (0.44-0.48), most likely caused by the high cover of wet biological soil crusts, which covered 50%-85% of the soil surface and comprised 33%-86% of the total biomass . Rich fruticose lichen communities occurred on the most favourable zonal sites on Hayes Island, a result of the absence of reindeer (Supporting Information Appendix S12).
Numerous other studies have also noted the unique vegetation in subzone A (Chernov & Matveyeva, 1997;Daniëls et al., 2016) and its extreme susceptibility to climate change (Walker, Raynolds, & Gould, 2008). It is interesting that the total species richness of the coldest, most northern zonal location (Krenkel, KR-1, 37 species) is higher than that of the warmest most southern zonal location (Nadym, ND-1, 20 species; Supporting Information Appendix S12). The relatively high species richness at Krenkel is due to the large number of cryptogam species (24-27.8 species). Other arctic researchers have also noted high plot-scale cryptogam species richness at cold temperatures (Bültmann, 2005;Lünterbusch & Daniëls, 2004;Matveyeva,1998;Timling et al., 2012). In studies of Arctic lichen floras from subzone E to subzone A, the number of vascular plant species declines by approximately 95%, whereas the number of lichen species declines by only approximately 15% (Dahlberg, Bültmann, & Meltofte, 2013). The same authors note that the relatively small decline in lichen species at higher latitudes is due mainly to reductions in the number of lichens that normally grow on woody plants, which are greatly reduced toward the north. Increased availability of light due to reduced competition from herbs and shrubs is a major cause of high moss and lichen richness at the more northern sites (Marshall & Baltzer, 2015;Walker et al., 2006). Further competition for light occurs within very dense cryptogam layers in the southern locations, where a few reindeer lichen species with erect fruticose lichen growth forms (e.g. Cladonia stellaris, C. stygia, C. rangiferina, C. arbuscular and C. mitis) densely cover the ground of lichen woodlands and out-compete other species.

| Implications for Arctic climate change and ecosystem studies
Ground-based documentation of existing patterns of vegetation is a critical element of space-based monitoring of changes to terrestrial ecosystems during a time of rapid climate and land-use change in the Arctic (Stow et al., 2004). The patterns of vegetation greenness (NDVI) change have not been spatially or temporally consistent across the Arctic, due in part to the constantly changing patterns of sea ice in the Arctic basin (Bhatt et al., 2013) and changes in the growing season and productivity patterns ( (Park et al., 2016).