Limited decadal growth of mountain birch saplings has minor impact on surrounding tundra vegetation

Abstract Temperatures over the Arctic region are increasing at three times the rate of the global average. Consequently, Arctic vegetation is changing and trees are encroaching into the tundra. In this study, we examine the establishment and growth of mountain birch (Betula pubescens ssp. tortuosa), which forms the treeline in subarctic Europe, and its impact on community composition across the treeline ecotone nearby Abisko, Sweden. Birch advancement along elevational gradients was studied by comparing data collected in 2016 with data collected 10 and 15 years previously. Species identity, cover, and phylogenetic relatedness were used to assess the impact of birch encroachment on community composition. Our results show that birch occurrence above the treeline did not affect plant community composition, probably owing to the observed lack of significant growth due to herbivore browsing, nitrogen limitation, or a reduction in snow cover. Independent of birch performance, the tundra community structure shifted toward a novel community dissimilar from the forest plant community found below the treeline. Taken together, our findings are explained by species‐specific responses to climate change, rather than by a linear forest advance. Future treeline advancements are likely more restricted than previously expected.

warming include an increase in primary productivity (Arctic greening) and in the cover of shrubs and trees (e.g., Bjorkman et al., 2018;Elmendorf et al., 2012;Myers-Smith et al., 2020). The taller vegetation as a result of the increase in productivity remains more exposed above the snow cover in winter and hence reduces surface albedo (Bjorkman et al., 2018;Sturm et al., 2001). Models have shown a reduction in winter albedo of almost 2% between 1980 and 2006 due to longer growing seasons (Bonfils et al., 2012). A decline in snow cover, coupled with a decrease in albedo due to forest expansion, can accentuate further warming to such an extent that it overrides the negative climate feedback resulting from increased carbon uptake by trees (Betts, 2000;de Wit et al., 2014). Thus, the expansion of trees and shrubs into the Arctic is of urgent global concern (Larsen et al., 2015).
Shifts in treeline position uphill or poleward into tundra ecosystems under climate change have been observed in many locations around the world (e.g., Bryn & Potthoff, 2018;Harsch et al., 2009;Rees et al., 2020). Globally, 52% of all treelines studied have advanced, with only 1% reporting treeline recession (Harsch et al., 2009). Particular treelines in areas experiencing strong winter warming were more likely to advance (Harsch et al., 2009). In addition, the altitudinal position of the treeline has been found to positively correlate with a ground temperature of 6-7°C in the mean growing season (Körner & Paulsen, 2004), which suggests that the treeline positions are sensitive to a changing climate. Climate scenarios of future treeline advance in the Abisko region of northern Sweden predict the treeline to advance between 45 (RCP 2.6) and 195 (RCP 8.5) elevational meters by the year 2100 (Gustafson et al., 2021).
Despite strong global evidence for climate-driven treeline advance, scale-dependent factors such as varying thermal conditions, soil type, and nutrient availability can have a strong influence on local patterns in treeline advance (Holtmeier & Broll, 2017). In addition, mammal browsing and more local patterns of herbivory (e.g., by invertebrates) can restrict tree advancement into the tundra (Kaarlejärvi et al., 2013;Van Bogaert et al., 2011). Due to large regional variation in these variables (Hofgaard et al., 2012), studies have identified slow advance (Mathisen et al., 2014;Rees et al., 2020), retreat (Dalen & Hofgaard, 2005), stability in distribution (Hofgaard et al., 2009), and increased forest density (Lescop-Sinclair & Payette, 1995) in forests within the same region. Tree species in the Swedish Scandes have advanced uphill in recent decades (Kullman, 2010), where, for instance, in the year 2000, mountain birch saplings were identified almost 400 m higher than the uppermost limit for this species recorded in 1955 (Kullman, 2002).
Though the establishment of these birch saplings is restricted to local enclaves with a more favorable microclimate, they do not necessarily indicate treeline advance (Løkken et al., 2020;Sundqvist et al., 2008). The plant community composition in the direct vicinity of birch individuals that have been established on the tundra was also found to shift by becoming more similar to the composition in the nearby birch forest (Sundqvist et al., 2008). This may indicate that mountain birch has a positive effect on the growth of some subalpine species, whereby birch saplings ameliorate the habitat and buffer some neighboring plants from open tundra-related stress (Bertness & Callaway, 1994;Brooker et al., 2008). For example, mature mountain birches have been shown to facilitate birch seedling performance and survival at high-stress sites with strong winds and extreme temperatures (Eränen & Kozlov, 2008). They have furthermore been found to facilitate the establishment of other trees, such as Scots pine (Mikola et al., 2018). In addition, both warming and mountain birch leaf litter can increase moss-associated nitrogen fixation in subarctic tundra environments, promoting shrub expansion (Rousk & Michelsen, 2017).
Here, we study the growth of mountain birches that have been established in the Swedish tundra for over 15 years (Sundqvist et al., 2008;Truong et al., 2007), and explore how alpha (α) diversity and beta (β) diversity have varied over time in tundra communities with and without birch saplings. Further, we compare diversity patterns in local tundra communities with and without mountain birch with those in a nearby mountain birch forest. Based on previous observations (Sundqvist et al., 2008), we predict that birch-infiltrated tundra communities have become more similar to birch forest communities in terms of their α-and β-diversity over time. Finally, we assess whether these patterns are consistent over multiple aspects of diversity by comparing species richness, evenness, and phylogenetic diversity.

| Study sites
The study area is located in the area around Latnjajaure Field Station (68°21′ N, 18°30′ E) at the shoreline of lake Latnjajaure (990 m a.s.l.).The closest forest in this area is located in the lower valley of Kårsavagge south of Latnjajaure (610-810 m a.s.l.), northern Swedish Lapland ( Figure 1). The mean annual temperature at Latnjajaure Field Station ranges from −2.9°C to −0.5°C  and has increased at a rate of 0.4°C [CI = 0.2-0.7°C] per decade from 1993 to 2016 (Scharn, Brachmann, et al., 2021). The mean annual precipitation is 847 mm, with a range of 605 and 1091 mm (1990Scharn, Brachmann, et al., 2021). There is no in-site climate data for the forest, but at the nearby Abisko meteorological station (385 m a.s.l.), the mean annual temperature is +0. 3°C (1991-2011), with an increase of 2.6°C from 1913 to 2006 (Callaghan et al., 2013).
The mean annual precipitation is 332 mm (2002Callaghan et al., 2013). The Latnjavagge valley is an Oroarctic tundra (Virtanen et al., 2016) covered by snow from October to mid-June most years.
Vegetation composition in the valley varies depending on altitude, soil moisture, and the time of final thaw, as well as soil nutrients and pH (Lindblad, 2006). The treeline reaches about 810 m a.s.l. (where the uppermost 2-m tall mountain birch tree has been found; Sundqvist et al., 2008).
The upper range of the closed birch forest in the area is at around 700 m a.s.l., but at Latnjajaure, 100 birch saplings have been recorded above the treeline within an altitudinal range of 810-1170 m a.s.l. (Sundqvist et al., 2008;Truong et al., 2007). The growth (height, number of leaves, and number of branches) of these birch saplings above the treeline has been found to be positively correlated with abiotic factors such as southern exposure resulting in higher air temperatures and longer growing season due to earlier snowmelt (Truong et al., 2007). Further, at approximately 1060 m a.s.l., around 30 saplings have been established within a 10-m radius of each other in an area, which, due to local topography, has a warmer microclimate than the valley floor near the Latnjajaure Field Station.
Most of the mountain birch saplings in Latnjajaure are situated on a southwest-facing slope, where soil pH and nutrient content are high, and where Garnet mica schists with some inclusions of marble and dolomite dominate the bedrock (Kulling, 1964).

| Data collection
In July 2016, we revisited birch individuals previously observed by Sundqvist et al. (2008) in Latnjavagge in 2006. Seven of 18 individuals from the previous dataset were found and included, supplemented by nine individuals that were identified in this study.
Hence, in total 16 individuals were measured above the treeline at elevations of 960-1180 m a.s.l. A 50 × 50 cm plot was placed around each birch individual (named "Betula" plots), and Control plots were assigned approximately 10 m away from each "Betula" plot following the methodology in Sundqvist et al. (2008).

| Phylogeny
We used the time-calibrated phylogenetic tree of vascular plants in Latnjajaure generated by Scharn, Little, et al. (2021). We pruned the tree to only contain species observed in this study using the package ape (v5.3; Paradis et al., 2004) in the R programming language (version 4.0.0; Arbor Day, R Core Team, 2018).

| Diversity metrics
For each plot, we calculated species richness (SR) as the total number of each species in each plot, and Simpson's index of diversity (D), where an increase in this index represents an increase in diversity (Magurran, 2003). Based on the maximum-likelihood tree obtained from the phylogenetic inference, α-and β-diversity indices were calculated using the interMPD (interspecific abundanceweighted mean pairwise distance; Miller et al., 2017)  F I G U R E 1 Elevation map over Latnjajaure and Kårsavagge derived from ground elevation model at 1-m grid cell size (Lantmäteriet, 2019a). Locations and heights of birch saplings included in the study are marked by dots. Solid lines and green shading mark the treeline and closed forest area, respectively. Both were derived from a synthesis of RGB orthophoto (Lantmäteriet, 2019b), the Swedish National Land Cover Database (Naturvårdsverket, 2020), and field data Scharn, Little, et al., 2021). In order to measure β-diversity based on pure presence/absence, as well as abundance-weighted input, we used Jaccard and Bray-Curtis dissimilarity indices, respectively.

| Statistical analyses
All analyses described below were done in R (version 4.0.0; Arbor Day, R Core Team, 2018), using the following packages: ggplot2 year as a fixed effect, and altitude (factor) as a random effect (see below). All Bayesian models utilized Markov chain Monte Carlo with two chains each consisting of 30,000 iterations of which the first 20,000 were discarded as burn-in. We define a "significant" effect where the difference between categories did not overlap with 0 in its 95% credible interval (CI). We used noninformative priors for all coefficients and checked for convergence of the chains for all parameters using the Gelman-Rubin convergence statistic (Gelman & Rubin, 1992).
where μ y = mean value for each community year, α y = intercept value for a certain community and year, α a = altitude-level (factor) random effect, σ = the variance parameter of the normal distribution, and = the intercept mean.
To estimate the variation of each diversity metric between community types and years, we used a similar model to the birch height model described above. The model included diversity metrics as a dependent variable and community, year, and their interactions as fixed effects. The example below describes the model used to model interMPD. The models used to describe SR and D responses follow the Poisson and beta error distributions, as they represent count and 0-1 bounded data, respectively, but are otherwise identical.
where D mpd = the observed interMPD per plot, μ c,y = mean value for each community and year, α c,y = intercept value for a certain community and year, α a = altitude-level (factor) random effect, σ = variance parameter of normal distribution, and = the intercept mean. The models also included altitude as a random effect to account for the nonindependence of plants measured within the same altitude and the same plot measured in both years.
In addition to the intercept-only model, we tested whether community differences varied with altitude. Similar to the intercept-only model, the linear model included diversity metrics as a dependent variable and community, year, and their interactions as fixed effects using altitude as a linear predictor rather than as a random effect.
The example below describes the model used to model inter-MPD. The models used to describe species richness and Simpson's responses follow the Poisson and beta error distributions, as they represent count and 0-1 bounded data, respectively, but are otherwise identical.
where D mpd = the observed mean pairwise distance (interMPD) per plot, μ c,y = mean value for each community and year, α c,y = intercept value for a certain community and year, c,y = slope value for a certain community and year over altitude A a , σ = variance parameter of normal distribution, and and represent the intercept and slope means, respectively.
We used the betadisper function in vegan (Anderson, 2006) to calculate and compare within-community β-dispersion using Jaccard, Bray-Curtis, and netMPD indices among both years of measurement. We then compared taxonomic community composition through time using permutational multivariate analysis of variance (PerMANOVA) implemented with the Adonis function in vegan, constraining the permutations to each unique altitude to account for repeated sampling through time (Anderson, 2001). We used nonmetric multidimensional scaling (NMDS) to identify patterns between communities and over time. Confidence interval ellipses and community centroids were extracted using the ordiellipse function in the R package vegan (Kruskal, 1964;Minchin, 1987).

| Species turnover between years
In 2006, we recorded an average of 11.69 species in Control, 11.89 species in Betula, and 11.26 species in Reference plots. In 2016, we recorded 11.51, 11.40, and 11.14 species for Control, Betula,  we did not observe any difference in diversity over altitude between communities nor between years for any of the metrics (Figure 3).

| Birch effect on plant diversity
Results of the β-dispersion analysis did not reveal any significant difference in dispersion between communities for any of the tree metrics (Jaccard, Bray-Curtis, and netMPD). When comparing community composition, however, all metrics showed significant differentiation between the Reference and both the Betula and Control communities (Figure 4). Differences between the Jaccard and net-MPD metrics were only observed between the 2006 and 2016 Control plots, a trend that was also observed for the Bray-Curtis metric (Table 1).

| DISCUSS ION
We found that change over time rather than the presence of birches drove shifts in most α-and β-diversity measures in tundra communities at Latnjajaure. Furthermore, both species richness and evenness aspects of α-diversity decreased over time in the tundra sites, resulting in these communities becoming more similar to the Reference birch forest. However, in contrast to our prediction this decrease in diversity was not driven by the presence of birches. Thus, even though the birch saplings in Latnjajaure have been established for more than 15 years (Truong et al., 2007), our result points out that these mountain birches do not ameliorate their habitat enough to buffer surrounding neighbor plants from stress that would lead to a vegetation shift (Bertness & Callaway, 1994;Brooker et al., 2008). To date, studies on ameliorating effect of the presence of birches have focused on mature trees (Eränen & Kozlov, 2008;Mikola et al., 2018). Thus, the unde- with improved growth (Kullman, 2010). Although it is unlikely that the birch saplings currently are strong drivers themselves, birch treeline advance over the past century has generally been through the rapid growth of established individuals (Kullman, 2010). Thus, if future warming increases growth, they might still be an important component in driving community change.
The limited birch sapling growth we observe was restricted to growth in stem width but not height. Air temperatures have increased in the Abisko area over the last century (Callaghan et al., 2013), but in Latnjajaure, the increase has not been as strong during the last decade as during the 1990s and early 2000s (Björk et al., 2007;Scharn, Brachmann, et al., 2021). Furthermore, the wind is causing local variation in physiognomy seen in the world's treelines (Richardson & Friedland, 2009). Thus, the windier conditions prevailing in tundra compared within the birch forest may explain the increase in stem width but not in height, and the increase in mean air temperature may yet not be enough to overcome the wind effect at this altitude. In addition, the length of the growing season has been shown to play a larger role in birch height than mean air temperature (Kullman, 2010). Hence, the prolongation of the growing season in the Abisko area (Kohler et al., 2006) may not yet have reached the threshold needed for the saplings to grow into trees. This is in line with findings that age distribution and sapling height above the treeline do not corroborate predictions of rapid forest advance in northern areas such as Abisko (Hofgaard et al., 2009). In this region, the winter is the season that is warming the most (Scharn, Brachmann, et al., 2021), which most likely does not affect birch growth itself, but may influence snow conditions.
The thickness, continuity and duration, of the winter snowpack are common limiting factors for arctic vegetation as the snow provides shelter from the harsh winter climate (Wipf & Rixen, 2010), as well as protection from large herbivores such as mountain hare, ptarmigan, and reindeer (Christie et al., 2014;Nordengren et al., 2003;Rödel et al., 2004;Vowles et al., 2016). Browsing may have played a significant role in controlling birch growth (Kaarlejärvi et al., 2013;Løkken et al., 2019;Speed et al., 2011). Unlike the North American Betula glandulosa and Betula neoalaskana, Betula pubescens ssp.  tortuosa is resin-free and only produces condensed tannin as an anti-browsing defense (Bryant et al., 2014). For this reason, it has been suggested that browsing has considerably more impact when considering B. pubescens, compared with that in areas such as North America, where caribou have less impact on the expansion of birches in the tundra (Bryant et al., 2014). In addition, warming has shown to not significantly affect birch saplings in the Norwegian forest-tundra ecotone, whereas browsing had a much larger impact on their growth (Hofgaard et al., 2010;Løkken et al., 2019).
One alternative explanation is nitrogen limitation, which is commonly found in high latitude ecosystems (Tamm, 1991). A recent modeling study in the Abisko area suggested that nitrogen could become a limiting factor once climate conditions are favorable for growth (Gustafson et al., 2021). Nitrogen fertilization experiments have also shown that mountain birch at the treeline displays additional growth after nitrogen additions (Sveinbjornsson et al., 1992), as well as improved birch seedling survival at the treeline (Grau et al., 2012). shifted toward a graminoid-dominated tundra have been reported to be driven by herbivory (Wal, 2006;Zimov et al., 1995). However, the long-term impact of warming climate may override the herbivore effect through an increase in shrub abundance shifting the vegetation state (Vowles et al., 2017). Climate change has been predicted to cause idiosyncratic responses and therefore generates entirely new communities, rather than a movement of preexisting ecosystems (Molau, 1997;Wang et al., 2021). Shifts to novel communities after drastic climatic change are not uncommon throughout the history of the Arctic flora, as exemplified by the flora shifting from a dry steppe, forb-dominated system before, to a shrub-and graminoiddominated system after the last glacial maximum (Willerslev et al., 2014). Thus, our results, though restricted to one valley, are broadly in line with the projected climate-driven altitudinal vegetation changes in the forest-tundra ecotone. However, the changes are small and, despite some indicative results, together with a 10- year time frame could be too short to result in any major vegetation transitions. Furthermore, the small number of birches available for sampling may not have had the precision required to observe small shifts in diversity if they did occur.

| CON CLUS ION
The expansion of birch forest is known to facilitate vegetation change through buffering stress on neighboring plants (Eränen & Kozlov, 2008;Mikola et al., 2018). The birch saplings found above the treeline at Latnjajaure, however, are growth-limited and do not affect the surrounding plant diversity. The lack of growth in spite of regional climate warming is likely explained by herbivory, nitrogen limitation, or a reduction in snow cover, though their importance needs to be experimentally verified. Independent of the presence of birches, the surrounding tundra community structure is shifting. Its direction is unclear, though it does not show any indication of shifting toward the birch forest communities. Thus, more evidence from both observations (Hofgaard et al., 2009;Kullman, 2010;Løkken et al., 2020;Rees et al., 2020), experimental (Løkken et al., 2019;Speed et al., 2011) and modeling (Gustafson et al., 2021), suggests that future treeline advancements are more restricted than commonly expected, implying that the vegetation change may be slower than hitherto assumed.