Functional diversity and trait composition of vascular plant and Sphagnum moss communities during peatland succession across land uplift regions

Most of the carbon accumulated into peatlands is derived from Sphagnum mosses. During peatland development, the relative share of vascular plants and Sphagnum mosses in the plant community changes, which impacts ecosystem functions. Little is known on the successional development of functional plant traits or functional diversity in peatlands, although this could be a key for understanding the mechanisms behind peatland resistance to climate change. Here we aim to assess how functionality of successive plant communities change along the autogenic peatland development and the associated environmental gradients, namely peat thickness and pH, and to determine whether trait trade‐offs during peatland succession are analogous between vascular plant and moss communities. We collected plant community and trait data on successional peatland gradients from post‐glacial rebound areas in coastal Finland, Sweden and Russia, altogether from 47 peatlands. This allowed us to analyse the changes in community‐weighted mean trait values and functional diversity (diversity of traits) during peatland development. Our results show comparative trait trade‐offs from acquisitive species to conservative species in both vascular plant and Sphagnum moss communities during peatland development. However, mosses had higher resistance to environmental change than vascular plant communities. This was seen in the larger proportion of intraspecific trait variation than species turnover in moss traits, while the proportions were opposite for vascular plants. Similarly, the functional diversity of Sphagnum communities increased during the peatland development, while the opposite occurred for vascular plants. Most of the measured traits showed a phylogenetic signal. More so, the species common to old successional stages, namely Ericacae and Sphagna from subgroup Acutifolia were detected as most similar to their phylogenetic neighbours. Synthesis. During peatland development, vegetation succession leads to the dominance of conservative plant species accustomed to high stress. At the same time, the autogenic succession and ecological engineering of Sphagna leads to higher functional diversity and intraspecific variability, which together indicate higher resistance towards environmental perturbations.


| INTRODUC TI ON
In peatlands, accumulation of organic matter drives autogenic succession towards drier, more acidic and less fertile conditions as increasing peat thickness reduces the flow of mineral-rich waters to the peatland surface (Gignac & Vitt, 1990). Succession typically proceeds from fens dependent on catchment hydrology towards drier, nutrient-poor bogs (e.g. Bauer et al., 2003;Hughes & Barber, 2003;Kuhry & Turunen, 2006). The change in the vegetation and environmental conditions during the peatland succession modifies ecosystem functions, and the rate and vulnerability of the carbon (C) accumulation (e.g. Tuittila et al., 2013). Younger stages have been found to respond more readily to changing weather conditions while the annual variation in CH 4 and CO 2 dynamics is smaller in the older stages (Leppälä, Laine, et al., 2011;Leppälä, Oksanen, et al., 2011).
Recent research from boreal peatlands implies that vegetation composition of the earlier successional stages that are dominated by vascular plants is more sensitive to environmental change, such as lower water availably, than later, Sphagnum moss-dominated stages (Kokkonen et al., 2019). However, the mechanism behind the resistance of later stages is so far unknown.
Trait-based ecology, with focus on quantitative plant properties, offers tools for mechanistic understanding on how vegetation responds to the changes in environmental conditions, and how vegetation change impacts ecosystem functions (Cornelissen & Makoto, 2014;Moor et al., 2017;Suding et al., 2008). Generally, decreasing resources are associated with environmental filtering of plant properties (i.e. functional traits) towards a combination that enhances resource-conservative growth strategy over resourceexploitative strategy (Reich, 2014). Over such gradients, properties that allow quick returns on resource investments, that is, high resource acquisition rates and high growth rates, such as high SLA and photosynthesis, are replaced with properties related to slow growth such as high leaf dry matter content (Eskelinen & Harrison, 2015).
Although the combinations of nutrient and moisture conditions in peatlands complicate the bases of simple resource gradient (Moor et al., 2017), a pressure towards resource-conservative growth strategy over the peatland hydrosere succession from the relatively high resource habitats, such as rich fens, to the extremely resource-poor habitats, such as oligotrophic fens and bogs, can be expected.
During peatland succession, plant community dominance is known to shift from sedges and forbs to Sphagnum mosses and sedges to finally Sphagnum mosses and shrubs. While plant traits are yet little studied in peatlands, there is a strong evidence that the functional plant groups differ in their productivity, photosynthetic capacity (e.g. Hájek et al., 2011;Laine et al., 2016;Goud et al., 2017), decomposability (Bengtsson et al., 2018;Dorrepaal et al., 2005;Lang et al., 2009;Straková et al., 2012) and impact on methane emissions (Bubier et al., 1995;Gray et al., 2013). Therefore, successional changes in the functional group dominance are likely to impact the community trait composition and lead in the trait trade-offs from the high resource acquisition rates and rapid growth (typical to forbs and grasses) towards greater resource conservation (typical to evergreen shrubs), as has been found for vascular plant communities in different types of ecosystems (Chapin III et al., 1993;Diaz et al., 2004;Reich, 2014). While in peatlands the trait trade-offs have not been tested over large environmental gradients, the few existing studies imply that in bogs an artificial N enrichment has an ability to increase community-weighted leaf size, SLA and plant height, that is, cause a shift towards the more resource acquisitive trait composition (Lin et al., 2020), although a study by Bubier et al. (2011) found weak species-specific responses.
The Sphagnum moss traits are yet little studied, but there are findings indicating the trait trade-offs from acquisitive to conservative strategy also within Sphagna: Laine et al. (2011) found lower rates of photosynthesis and growth from species present at the bogs than from the younger peatlands, and Mazziotta et al. (2019) found a fast-slow economic spectrum among Sphagnum mosses growing along water level and shading gradients. Altogether, the successional changes in traits related to productivity and decomposability could be a way forward in understanding the differences in the rates of C accumulation between peatland types (e.g. Tuittila et al., 2013).
In addition to the direction of change in the plant properties, the trait-based approach can be used to evaluate the resistance of a plant community to an environmental change, that is, the ability of its functions to remain essentially unchanged despite species Ericacae and Sphagna from subgroup Acutifolia were detected as most similar to their phylogenetic neighbours.

4.
Synthesis. During peatland development, vegetation succession leads to the dominance of conservative plant species accustomed to high stress. At the same time, the autogenic succession and ecological engineering of Sphagna leads to higher functional diversity and intraspecific variability, which together indicate higher resistance towards environmental perturbations.

K E Y W O R D S
functional diversity, functional traits, intraspecific variability, peatland development, phylogenetic signal, plant economic spectrum, Sphagnum turnover in response to the environmental change (see e.g. Robroek et al., 2017). One tool is to observe the relative contributions of species turnover (interspecific traits variability) versus intraspecific trait variation. A larger relative contribution of intraspecific variation implies greater community resistance as species are able to adapt to changing conditions by modifying their traits according to the new environmental requirements (Albert et al., 2011;Lepš et al., 2011;Violle et al., 2012). In addition, functional diversity can be used in evaluating resistance, as greater trait dispersion within a community implies higher resistance (Dıaz & Cabido, 2001). Consequently, the recently observed lower community resistance in the fens than in the bogs (Kokkonen et al., 2019) gives us a reason to expect increased intraspecific variation in traits and/or increased functional diversity during peatland succession.
Primary paludification and the following peatland development occur in the areas of ongoing post-glacial rebound, namely along the coast of Bothnian Bay in Finland and Sweden, the Hudson Bay lowlands in Canada and along the White Sea in Russia, regions where the ice sheet was thickest during the last glaciation. The phenomenon creates chronosequences that can be used in addressing both the compositional and the functional change over long-term succession (e.g. Tuittila et al., 2013), which is not possible with experimental or palaeoecological approach.
We take advantage on the successional gradients found in the rebound areas in Finland, Sweden and Russia to (a) quantify if succession leads in a trait trade-off from acquisitive to conservative strategy that relates to low decomposition rates, a prerequisite of C accumulation in peatlands, at both vascular plant and moss commu-

| Study areas
We sampled four study areas located in the regions with ongoing post-glacial rebound: Sävar on the west coast of Bothnian Bay (63°50′N, 20°40′E, Sweden), Siikajoki (64°45′N,24°43′,Finland) and Hailuoto island (65°07′N, 24°71′E, Finland) on the east coast of Bothnian Bay, and Belomorsk-Virma (63°90′N, 36°50′E, Russia) on the coast of the White Sea ( Figure 1). Within each area, we selected peatlands with differing elevation and consequently different age and developmental stage. The peatlands are assumed to have initiated via primary paludification immediately after the uplift exposed new coastal land, except in Sävar where most of the sites emanate from lake ingrowth. The number of sampled sites varied from 17 in Siikajoki to 7 in Hailuoto, with 47 sites in total. The climatic conditions were quite uniform across the study areas, with mean annual temperature ranging from 2.8 to 4.6°C, mean annual precipitation from 542 to 579 mm, coldest month temperatures ranging from −8.2 to −9.6°C and warmest month from 15.5 to 16.2°C.

| Plant community and environmental measures
At each site, we determined the plant community composition by visually estimating the projection cover of each species separately for field (vascular plants) and moss layer using the scale 0.1%, 0.25%, 0.5%, 1%, 2%, 3%, etc (e.g. Sottocornola et al., 2009)

Peat thickness was measured with a Russian corer except in
Hailuoto where an iron rod with a rounded head was used. pH was measured from peat water samples with a YSI Professional Plus (Pro Plus; YSI Incorporated). A small hole was dug to peat and water was allowed to collect at the hole and stabilize for at least half an hour before pH measurement. Site location and elevation were recorded with a GPSMAP 64s (Garmin Ltd.; precision ± 3.65 m). To estimate the average moisture conditions in the sites, we calculated a moisture index for each site as a weighted average based on the species Ellenberg moisture index values and average species cover. Ellenberg moisture indices were attained from the studies of Hill et al. (1999Hill et al. ( , 2007 and Ellenberg et al. (1991). In our dataset, the species-vice index ranged from 5 (Vaccinium vitis-idaea, Pleurozium shreberii) to 12 (Utricularia intermedia), that is, from a moist-site indicator, mainly on fresh soils of average dampness (5) to a submerged plant, permanently or almost constantly under water (12).

| Functional trait measurements
Functional traits were measured during the plant community measurement campaigns. At each site, we selected the most common (on average 80% coverage) vascular and moss species for trait measurements (Tables S1.1). Plant height was measured from 10 individuals per vascular species per site, while other traits were measured from half of these plants, and three replicates per species per site were used for mosses.
We measured the following vascular plant traits: plant height, leaf size (LS), SLA and leaf carbon (C) and nitrogen (N) contents. These traits were selected as they are known to correlate with ecosystem functions, that is, productivity and decomposability, and with plant strategies. Increase in plant height has been linked to switch from nutrient competition to competition from light, while the selected leaf traits are associated with leaf economy. The measurements were made according to the study by Pérez-Harguindeguy et al. (2016), and we sampled only fully grown plants and fresh and undamaged leaves. Plant height (cm) was measured as the shortest distance between the upper boundary of the main photosynthetic tissues (excluding inflorescences) and the ground level. To measure SLA (i.e. the one-sided area of a fresh leaf divided by its oven-dry mass, cm 2 /g), the freshly picked leaf or a sample of three leaves in case of shrubs with small leaves was pressed flat between a board and a glass, and a standardized photograph was taken. The LS (cm 2 ) was analysed from the photographs with ImageJ software (Schindelin et al., 2015;Schneider et al., 2012). The leaf samples were stored in paper bags and dried at 60°C for a minimum of 48 hr. The dried samples were weighed, and SLA was calculated. The SLA samples were used for C and N content analysis. Leaves from each species from each site were pooled into one sample, which was milled (Retsch MM301 mill) and analysed for C and N concentrations and for C:N ratio on a CHNS/O Elemental analyzer (EA1110; University of Oulu).
Sphagnum moss samples were collected with a corer (7 cm diameter, area 38 cm 2 , height at least 8 cm) to maintain the natural density of the stand. The collection was targeted to single species stands.
Samples were stored in plastic bags in a cool room or frozen until measured. Unlike for the vascular plants, there are no established set of traits for Sphagnum mosses. Therefore, we selected such properties that our recent study linked to the capacity to remain moist and, consequently, to photosynthesize (Gong et al., 2020). We defined the following Sphagnum traits: stand density (number of shoots cm −2 ), capitulum width (cap_width, mm) and dry weight (cap_dw, mg), fascicle density (number cm −1 ), capitulum dry matter content (CDMC, mg/g), capitulum water content (cap_wc, g/g) and capitulum C and N contents and C:N ratio. Stand density was measured as the number of mosses in the sample. From 10 individuals we measured the width of the capitula and counted the number of fascicles from a 5-cm segment below capitulum. We separated the 10 moss individuals into capitulum and stem (5 cm below capitula), wetted them and allowed to dry on top of tissue paper for 2 min before weighing them for water-filled fresh weight. Samples were placed on paper bags and dried at 60°C for at least 48 hr after which the dry mass of capitula and stems were measured. CDMC and cap_wc were calculated from the fresh and dry weight. We used the capitulum samples for analyses of C and N concentrations and for C:N ratio, and treated them similar to vascular plant samples.

| Plant community data
To visualize the variation in peatland vegetation along the successional gradients in the four areas we applied non-metric multidimensional scaling (NMDS) ordination via the metaMDS function in the r package vegan (Oksanen et al., 2015). Our ordination was based on the Bray-Curtis dissimilarity matrix of the vegetation datasets with 47 sites. Significance of the effect of peat thickness, pH and moisture index on vegetation composition was tested with a permutational multivariate analysis of variance (PERMANOVA, adonis in vegan) on the same dissimilarity matrix (999 permutations). Envfit function (also from the vegan package) was used to visualize the correlation of peat thickness, pH and moisture index with NMDS ordination. To evaluate how the studied sites fitted within previous interpretation on primary peatland succession from the intensively studied Siikajoki area (e.g. Laine et al., 2011;Leppälä et al., 2008;Tuittila et al., 2013), we used a combination of CLUSTER analysis with Bray-Curtis similarities matrix and group average algorithm, and 'Similarity profile' (SIMPROF) tests, using the clustsig package (Whitaker et al., 2014) in the R environment (R Development Core Team, 2011). The most common species and average environmental conditions of each cluster are given in Tables S1.2. and S1.3. respectively. Finally, Generalized Additive Models (GAM) were used to evaluate the dependency of moss and vascular plant cover on the peat thickness, pH and moisture gradient.
Vascular plants were reduced to three groups, namely shrubs, sedges and grasses plus forbs, and mosses to two groups, Sphagnum mosses and 'others'. The analyses were performed with Canoco 5.11.
In all analyses we considered each site as a unit by averaging the sample plot data. Rare species that were present in less than three (of 47) sites were removed from the analysis.

| Functional trait data
Vascular plant and Sphagnum moss data were analysed separately as mostly different traits were used for each community. Communityweighted mean (CWM) trait values were calculated for each site using two differing trait datasets. The first set, called 'site-specific', included site-specific average trait values for those common species, generally covering ~80% of the vegetation, that were measured from the particular site. For the remaining rare species, present at the plant community but not measured from the site, we used an average trait value calculated from the full dataset covering all sites. In the second dataset called 'fixed', the average trait values calculated from the full data were used for all species. Community-weighted mean traits were calculated using the fd package (Laliberté et al., 2014) in the R environment (R Development Core Team, 2011). We used principal component analysis (PCA) to examine the covariation between site-specific CWM traits with centred and standardized trait values. To examine how main variation in traits relate to environmental variation, peat thickness and pH were included as supplementary variables with the envfit function. The analyses were performed using package vegan in r. We complemented the analyses with Horn's Test of Principal Components with Package paran in r environment. To relate individual traits with peat thickness and pH, we used linear mixed effect models with study area as a random factor with the nlme package in the R environment.
To assess vegetation resistance to environmental gradients, peat thickness and pH, associated with peatland succession, we evaluated the relative contributions of species turnover and intraspecific trait variation along the gradients for both vascular plant and Sphagnum moss communities using a method presented by Lepš et al. (2011). In this method, between-site differences in CWM trait values based on fixed trait values can only be caused by differences in species composition (i.e. species turnover). The between-site differences in CWM trait values based on site-specific trait values can be caused either by differences in species composition or/and intraspecific trait variation, while the difference between site-specific and fixed averages can be caused solely by intraspecific trait variation. For this we ran three generalized linear model regressions for each trait, using either 'sitespecific', 'fixed' or 'intraspecific' CWM site averages with peat thickness or pH as an explanatory variable. The 'intraspecific' being the difference between 'site-specific' and 'fixed' estimates. From each model, the sum of squares (SS specific, SS fixed and SS intraspecific respectively) was extracted and finally, the SS cov component, which is the effect of covariation between inter-and intraspecific trait variability, was calculated by subtracting SS fixed and SS intraspecific from SS specific. Models were run in the R environment.
To complement the analysis of the vegetation resistance to environmental gradients, we evaluated the dependence of functional diversity on pH and peat thickness gradients with PCA including peat thickness and pH as supplementary variables with the envfit function. We complemented the analyses with Horn's Test of Principal Components with Package paran in r environment. We used Rao's quadratic entropy from fd package in r (Botta-Dukát, 2005) as a measure of functional diversity and calculated it for both vascular plant and Sphagnum moss communities using (a) all measured traits and (b) each trait separately.

| Testing for phylogenetic signal in vascular and sphagnum traits
We reconstructed the phylogenetic tree of vascular plants using the v.phylomaker package in r (Jin & Qian, 2019) that generates phylogenies for vascular plants based on two recent large phylogenetic trees (Smith & Brown, 2018;Zanne et al., 2014). The resulting mega-tree includes 10,587 genera and 74,533 species of vascular plants. When the species are missing from the mega-tree, V.Phylomaker uses species that are sister to or closely related with some species or genera in the mega-tree to accurately attached species to their relatives in the phylogenetic tree. As bryophytes are missing from V.Phylomaker, we used gene sequence from the ribosomal non-coding protein trnL intron and trnL-F intergenic spacers to build Sphagnum phylogeny (Piatkowski & Shaw, 2019). We selected these loci as they were the only available sequences for all Sphagnum species in our species list in the GeneBank database. Each locus was aligned using MAFFT for multiple alignments (Katoh & Standley, 2013). Then, phylogenetic inference was performed using NGphylogeny.fr (Lemoine et al., 2019).
The resulting Sphagnum phylogenetic tree was supported by the high-resolution tree recently built for Sphagnum mosses and based on 16 loci (Piatkowski & Shaw, 2019).

We assessed the phylogenetic signal within vascular and
Sphagnum traits under the Brownian Motion model (BM). BM predicts that the variance in trait increases at a constant rate proportionate to evolutionary distance, with more closely related species having more similar values, indicating that the trait has a phylogenetic signal (Felsenstein, 1985). We used three indices to evaluate the phylogenetic signal within plant traits: Abouheif's C mean , Pagel's λ and Blomberg's K (Münkemüller et al., 2012). These indices build on the concept of autocorrelation. Each index has been tested for the null hypothesis of absence of the phylogenetic signal, that is, trait values are randomly distributed in the phylogenetic tree. This is achieved by randomization for K and C mean (1,000 permutations), and by likelihood ratio test for λ. While these indices give precious information about the general presence of a phylogenetic signal within a phylogenetic tree, these approaches assume that traits evolve similarly across the phylogeny. However, the phylogenetic signal likely varies across clades, and additional local statistics are needed to highlight local trait patterns within the phylogenetic tree. To do so, we used the Local Indicators of Phylogenetic Association (LIPA) index that allow to detect hotspot of autocorrelation within the phylogenetic trees between species and their neighbours. The phylosignal r package was used to run these analyses (Keck et al., 2016).

| Vegetation composition
The transects covered the successional trend from mesotrophic fen mosses and herbs (e.g. Warnstorfia exannulata, Sphagnum squarrosum, Potentilla palustris) to mosses and shrubs typical for ombrotrophic bogs (e.g. Sphagnum fuscum, Calluna vulgaris) seen in the variation in the species composition along the first NMDS ordination F I G U R E 2 Ordination based on nonmetric multidimensional scaling (NMDS) with plant species composition of the 47 mires and peat thickness and pH as supplementary environmental variables. Ordination has two dimensions and the Stress value is 0.16. Non-metric fit, R2 0.976, linear fit, R2 0.893. (a) Ordination of species with the largest contribution to the analysis, and (b) peatland sites clustered, based on their vegetation, into 10 groups (different symbols) with SIMPROF. The most common vascular and moss species per group are given in Tables S1.2. Five of the sites from Siikajoki chronosequence, which were previously divided into different successional stages, namely wet meadows, mesotrophic fens, oligotrophic fens, fen/bog transition and bog , are marked with arrows In here we estimated peatland moisture conditions based on the Ellenberg's indicator values and therefore the response of different plant groups to moisture is a sum of species expected affinity towards moisture. All peatland species have high (>6) moisture index, but shrubs and many Sphagnum mosses are accustomed to drier conditions than sedges and grasses (Figure 3c).  Table S1.4. Similar to vascular plant traits, CWM traits of Sphagnum mosses formed a successional gradient. Along this gradient capitulum width, N content and water content decreased and capitulum C content, capitulum C:N, moss stand density and the amount of fascicles per stem and per stand increased from the youngest to the oldest peatlands. Along this gradient, peat thickness increased and pH F I G U R E 3 Fitted Generalized Additive Models (using a Gaussian distribution with smoothness term of 2.0 df) for different plant functional types with (a) peat thickness (Sphagnum, sedge, shrub and other moss models had r 2 of 0.37, 0.48, 0.46 and 0.31, respectively, and p < 0.0005; grass + forb model was insignificant), (b) pH (Sphagnum, shrub and grass + forb models had r 2 of 0.29, 0.24 and 0.26, respectively, and p < 0.005; sedge and other moss models were insignificant) and (c) moisture index (Sphagnum, sedge, shrub and grass + forb models had r 2 of 0-28, 0.30, 0.56 and 0.26, respectively, and p < 0.005; other moss model was insignificant). Functional typewise plots including the measured data points are shown in Figure S1 decreased (Figures 4b and 6). Average traits values per Sphagnum species are given in Table S1.5.

| Resistance of successive plant communities
The importance of species turnover versus intraspecific CWM traits variability along the peat thickness and pH gradients differed between vascular plant and Sphagnum communities (Table 1).
For vascular plants, the changes in CWM traits were due to species turn-over (with exception of plant height along pH gradient), while for Sphagna also the intraspecific variability was significant for most traits. As an example, for fascicle density and C content the intraspecific variability was equally strong as variance caused by species turnover. For Sphagna, the covariation was always positive, indicating that sites dominated by species with larger value of a trait were also dominated by individuals of these species having larger than average value of the trait.
The response of functional diversity, expressed as Rao's quadratic entropy (RaoQ), to peat thickness and pH gradients was contrasting for vascular plant and moss communities (Figure 7a,b). For vascular plant communities, RaoQ increased along with pH ( Figure 7a), that is, towards the early succession sites, while for the Sphagnum moss communities, RaoQ increased along with peat thickness, that is, towards the older stages (Figure 7b). The pattern was similar for RaoQ calculated based on all traits (FD(Rao)) and based on the individual traits. The only exceptions were RaoQ for C:N and C of vascular plant communities, which increased along the peat thickness gradient.

| Phylogenetic signal in plant traits
Vascular plant traits showed a strong phylogenetic signal (Table 2), plant height being the only trait without the phylogenetic signal. Sphagnum mosses showed more contrasted results as phylogenetic signal was absent for Cap_wc, Cap_dw, C and N contents (Table 2). LIPA analyses on vascular plants revealed a significant local phylogenetic signal within Ericacae for all traits (Figure 8a), while among Sphagnum traits phylogenetic signal was local and significant within subgenus Acutifolia (Figure 8b). These two species groups were particularly associated with older successional stages ( Figure 2), and therefore the sites with higher proportion of species with significant phylogenetic signal were more common at conditions with acidic pH and high peat thickness (Figure 4a,b).

| D ISCUSS I ON
Our study assessed the change in plant traits and their plasticity during peatland succession. The sampling that encompassed a wide primary successional gradient of peatlands ranging from the coastal wet meadows through the sedge fens to the ombrotrophic bogs agreed with earlier literature on peatland succession (e.g. Bauer et al., 2003;Elina & Kuznetsov, 1996;Hughes & Barber, 2003;Klinger & Short, 1996). Peatland plant community dominance shifted from forb and graminoid to shrub and Sphagnum moss, following the gradient formed by increasing peat thickness and decreasing pH, that is, development towards the less fertile habitats (Gignac & Vitt, 1990). A turning point in peatland succession occurs when Sphagnum mosses become an important component of the vegetation (Magyari et al., 2001). In our study Sphagnum moss projection cover exceeded that of vascular plants already at sites with shallow peat (50 cm) and at pH of 4.8.

| Trait trade-offs during peatland succession
The trait trade-off from the acquisitive to the conservative resource use strategy is generally linked with ecosystem functions so that the communities with acquisitive species have higher productivity, rapid litter decomposition, high N turnover and availability, while high litter accumulation, resistance to herbivory and soil C accumulation are related characteristics of conservative strategy Lavorel & Grigulis, 2012). In peatlands, lower productivity but nevertheless also greater net C sequestration is connected to slow-growing species with low litter decomposability, commonly occurring in nutrient-poor peatlands (Freschet et al., 2012;Moor et al., 2015;Turetsky et al., 2008). Our trait-based results support this view as we found a similar vascular plant trait trade-off from acquisitive to conservative resource use strategy along the successional peatland gradient, as has been described from fertility and successional gradients in grasslands and forests (e.g. De Deyn et al., 2008;Garnier et al., 2004;Reich, 2014;Sterck et al., 2006;Wright et al., 2004) and along an experimental N gradient in an ombrotrophic peatland (Lin et al., 2020). To the best of our knowledge, this is the first study focusing on the trait trade-offs along peatland succession, the community-weighted SLA and LS decreased with increasing peat thickness and decreasing pH (decreasing nutrient status), agreeing with the study by Lin et al. (2020) who found SLA and LS to increase along with increasing N loading (increasing nutrient status). In general, high values of these leaf-level traits enhance light capture and leaf cooling and enable high photosynthetic capacity and growth rates (Pérez-Harguindeguy et al., 2016;Poorter & Evans, 1998;Walker et al., 2014). However, leaf N content, which is integral to the proteins of photosynthetic machinery, especially Rubisco (Lambers et al., 2008), showed a less clear association with the successional gradient. It has been shown that in low SLA leaves (i.e. plants with a conservative strategy), a larger share of N is bound to cell wall proteins per unit leaf area, leading to a smaller allocation of N to the photosynthetic machinery (Onoda et al., 2004;Takashima et al., 2004). Therefore, it is not self-evident that high N content should be an indicator of high acquisition rates but may, as well, be related to longer leaf life span, supporting a conservative strategy (Hikosaka, 2004). Plant height affects plants ability to capture light (Westoby et al., 2002) and varies independently from the leaf-level traits (Diaz et al., 2004). In our study, the plant height showed strong negative dependency on the peat thickness. This pattern occurred although all the sites were open or only sparsely treed peatlands, so it was not related to strong shading from tree stand. The CWM plant height was overall low, and varied relatively little (average 28 cm, range 12-52 cm). The only measured vascular plant traits positively associated with the increasing peat thickness were leaf C content and C:N ratio, which high values are good predictors for slow litter decomposition rates (e.g. Pérez-Harguindeguy et al., 2000).
In mosses, there was a trait trade-off from acquisitive to conservative species that can be considered analogous to that of vascular plants. In earlier studies, the high acquisition strategy of Sphagnum species has been associated with large capitula (apical growing part), low stand density, higher N content and greater rates of photosynthesis, growth and litter mass loss, while species with small capitula and densely packed stands coupled with slow ecosystem processes are considered conservative (Bengtsson et al., 2016;Laine et al., 2011;Laing et al., 2014;Mazziotta et al., 2019;Moor et al., 2015;Rice et al., 2008). In most of these studies, the trait trade-off has been related to a water-table and/or shade gradient, but here we show a strong correlation with peatland succession associated with peat thickness and pH gradients. We found a clear trend from early successional fens to ombrotrophic bogs. Early successional peatlands support capitula with high N content and loose stands that allow deeper light penetration and likely have greater vertical distribution of photosynthesis, metabolically inexpensive tissues that photosynthesize and elongate rapidly (Rice et al., 2008).
Ombrotrophic bogs, on the contrary, are dominated by Sphagnum species that tolerate environmental stress by high investment in water holding capacity, high carotenoid concentrations and larger C allocation to main structural carbohydrate polymers (Elumeeva et al., 2011;Limpens et al., 2017;Rice et al., 2008). In our study, capitulum size was an exception: similar to older bogs, early succession wet meadows that have unstable hydrology (Laine et al., 2016) support hummock-forming Sphagnum species with small capitula, such as Sphagnum fimbriatum. Therefore, the species with the largest capitula occurred at fens at the middle of the successional gradient.
At the early successional sites with a thin layer of peat, mineral nutrients from the underlying soil and incoming flow of water and nutrients enable the 'acquisitive strategy' that involves high use of resources (nutrients and water) to rapidly acquire C (Wright et al., 2004). At the other end of the successional gradient, thick peat layer creates an extremely restrained growth environment, poor in mineral nutrients and often with either limited (hummock) or excess (hollow) water availability (Rydin & Jeglum, 2013). In addition, these sites are usually exposed to direct sunlight, as vascular plant cover is low. Under such conditions, resource conservation enhances survival and plants with a conservative strategy thrive.

| Resistance of peatland plant communities
With our trait dataset we were able evaluate if the resistance of peatland plant communities agrees with the finding of higher resistance of the later successional stages (bog) than earlier stages (fens) by analysing the relative contributions of inter-versus intraspecific trait variation and the rate of functional diversity along the successional gradient. We observed different responses for vascular plants and Sphagnum mosses: for vascular plant traits, the interspecific variability dominated, indicating low resistance of the community structure to the changes in peat thickness and pH (Albert et al., 2011;Lepš et al., 2011;Violle et al., 2012). Supporting this, Bubier et al. (2011) and Lin et al. (2020)

| Phylogenetic signal of peatland plant traits
We found a phylogenetic signal for most of the measured vascular plant and moss traits indicating phylogenetic conservatism, that is, the tendency for closely related species to share similar ecological and biological attributes (Blomberg et al., 2003). The signal was particularly strong for the late successional species groups such as Ericoid shrubs and Sphagna belonging to subgroup Acutifolia (e.g. S. fuscum, S. capillifolium). Based on this, the harsh environmental con-

| Consequences of climate change on peatlands
Climate change that increases evaporative demand (Helbig et al., 2020) and likely reduces spring water excess in the northern peatland-dominated landscapes (Teutschbein et al., 2015) is suggested to accelerate peatland hydrosere succession, such that fens are more rapidly overgrown by bog vegetation (e.g. Väliranta et al., 2017). Both wet and warm periods, which induce greater peat growth rates (e.g. Hughes, 2000), and drier conditions with water-table drawdown, which lead to surface isolation and support Sphagnum growth (e.g. Tahvanainen, 2011), steer towards ombrotrophication. Our study shows that such shift from a fen to a bog community is connected to a trait trade-off from acquisitive to conservative strategy and changes in community resistance, which may strongly influence peatland C and nutrient cycling, with potential feedbacks to future climate via plant-soil microbe feedbacks (Bragazza et al., 2013;Jassey et al., 2013). Through the Holocene, fen to bog transitions induced by warmer and dryer climate have been connected to a sharp increase in rates of C accumulation (Loisel & Yu, 2013;Loisel et al., 2014). Consequently, accelerated succession towards bogs, which are more stable in their trait and species composition as found here, and in their C sink function (Leppälä et al., 2008;Wu & Roulet, 2014) could have a major effect on the climate-related impact of peatlands, potentially creating a cooling feedback. The persistence of Sphagnum mosses as the elemental part of the plant community seems, however, crucial for this positive development. Low functional diversity and low intraspecific variability in the bog vascular vegetation found in our study indicate that alternative development, that is, climate change-induced large-scale shrubification (e.g. Sturm et al., 2001;Tape et al., 2006) at the expense of mosses, could lead to reduced resistance of peatlands. Such ecosystem change has been shown to lead to a decreased peat accumulation (Turetsky et al., 2012).

ACK N OWLED G EM ENTS
The authors thank Prof. Steve Frolking for his contributions to the review process, Kimmo Vuokare and Antti Pöllänen for assistance in field and Tuija Maliniemi for constructing the map of locations.
Funding from Kone Foundation and Academy of Finland (project code 287039) made the study possible.

PEER R E V I E W
The peer review history for this article is available at https://publo ns.

DATA AVA I L A B I L I T Y S TAT E M E N T
Primary data are stored at the Pangaea Data Library https://doi. org/10.1594/PANGA EA.926827 (Laine et al., 2021).