Determinants of tree seedling establishment in alpine tundra

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. © 2020 The Authors. Journal of Vegetation Science published by John Wiley & Sons Ltd on behalf of International Association for Vegetation Science 1Plant Ecology and Nature Conservation, Wageningen University & Research, Wageningen, The Netherlands 2Department of Biology, Norwegian University of Science and Technology, Trondheim, Norway 3Faculty of Biosciences and Aquaculture, Nord University, Steinkjer, Norway 4Centre for Biodiversity Dynamics, Department of Biology, Norwegian University of Science and Technology, Trondheim, Norway 5Research Centre for Ecological Change, Faculty of Biological and Environmental Sciences, University of Helsinki, Helsinki, Finland 6Forest & Nature Lab, Ghent University, Gontrode-Melle, Belgium


| INTRODUC TI ON
Northern high-latitude ecosystems are strongly affected by climate change due to fast and intense warming (Elmendorf et al., 2015) and because their biota are limited by low temperatures (Michelsen et al., 2011;Vanneste et al., 2017). Ongoing changes in climate and herbivore pressure are affecting the species composition of these systems in several ways. For example, treelines, the lower boundaries of tundra ecosystems, are expected to advance in elevation and latitude, but the observed trends vary (Millar et al., 2004;Dalen and Hofgaard, 2005). In addition, shrubs (e.g. birch, willow and alder) are expanding in tundra communities worldwide (Myers-Smith et al., 2011;Frost and Epstein, 2014;Vanneste et al., 2017;Bjorkman et al., 2018), yet there are again exceptions (García Criado et al., 2020).
Interestingly, shrubs often facilitate tree establishment in tundra (Castro et al., 2004;Akhalkatsi et al., 2006;Chen et al., 2020). Thus, woody species have the potential to expand in these ecosystems (Myers-Smith et al., 2011), resulting in vegetation shifts from open herbaceous or dwarf-shrub-dominated to closed shrub-dominated communities that potentially are beneficial for tree establishment as well. However, it is still poorly understood in which situations and to what extent these shifts will occur.
Previous work has shown that woody encroachment occurs at unequal rates across heterogeneous alpine landscapes (Wookey et al., 2009;García Criado et al., 2020). Variation is thought to depend on the invasibility (i.e., susceptibility to the establishment of new species) of current communities, a characteristic determined by the interplay between biotic and abiotic factors (Graae et al., 2011;Milbau et al., 2013). Invasibility is often assumed to be driven by resistance from the receiving community (Bruno, Stachowicz, and Bertness, 2003;Bulleri, Bruno, and Benedetti-Cecchi, 2008). However, facilitation (i.e., positive biotic interactions) is suggested to be common across ecosystems with effects at least as strong as other factors shaping plant communities (Maestre et al., 2009;McIntire and Fajardo, 2014). For example, tree recruitment in tundra is often facilitated by shrub, tree or krummholz canopies (Castro et al., 2004;Akhalkatsi et al., 2006;Chen et al., 2020), though varying with the canopy-forming and recruiting tree species (Körner, 2012;Liang et al., 2016). This facilitation seems predominantly important in early recruitment phases (Cranston and Hermanutz, 2013;Brodersen et al., 2019). Amelioration of abiotic growing conditions (e.g. protection against temperature extremes, high irradiance and wind) during vulnerable recruitment stages is an important mechanism behind this facilitation (Akhalkatsi et al., 2006;Holmgren et al., 2015;Chen et al., 2020). Conversely, dense ground covers of herbaceous plants mainly suppress tree seedling recruitment by shading (Loranger, Zotz, and Bader, 2017). These plant-plant interactions are expected to shift from competitive to facilitative with increasing abiotic stress level (Choler, Michalet, and Callaway, 2001;Callaway et al., 2002;Blonder et al., 2018), suggesting that environmentally benign communities are less invasible than are more stressful communities.
Natural and anthropogenic disturbances, such as bare soil patches resulting from trampling, human recreation or transport, landslides or rock falls, are another strong driver of tree seedling recruitment in alpine habitats (Hättenschwiler and Körner, 1995;Munier et al., 2010;Tremblay and Boudreau, 2011). Small-scale disturbances, by removing or reducing the abundance of competitors, generate new microhabitats suitable for seedling emergence and establishment (Milbau et al., 2013;Nystuen et al., 2014;Lembrechts et al., 2016). However, severe disturbances can counteract the benefits of reduced competition, for instance when it leads to too hot and too dry soils that can be detrimental for tree seedlings (Kambo and Danby, 2018;Nystuen et al., 2019).
Browsing, grazing and trampling by herbivores such as ungulates and small rodents also create disturbances, and have therefore been suggested to indirectly facilitate plant recruitment in tundra (Ims Vistnes and Nellemann, 2008;Tremblay and Boudreau, 2011;Milbau et al., 2013). However, other studies have detected the opposite effect: experimental herbivore exclusion increased seedling establishment (Olofsson et al., 2009;Munier et al., 2010;Ravolainen et al., 2014). Herbivory can thus either increase or reduce the invasibility of a plant community, thereby either stimulating or reducing tree seedling establishment.
Alpine tundra vegetation comprises a patchwork of distinct community types that differ in abiotic conditions created by strong gradients of environmental stress. In low-alpine areas of Fennoscandia, typical topographical gradients occur from harsh wind-exposed and dry heaths to more benign sheltered shrublands, meadows and snowbeds (Graae et al., 2011). Consequently, new species trying to establish in these communities will not only be subjected to differences in abiotic stress but also to differences in biotic interactions with co-occurring plant species, soil biota and herbivores.
Here, we explored the mechanisms underlying tree invasibility in alpine tundra in a full-factorial pine seed sowing experiment manipulating canopy cover, herbivore exclusion and shrub introduction in three alpine plant communities differing in abiotic stress.
Treatment effects on emergence, survival and performance of Scots pine (Pinus sylvestris) and microclimatic conditions were monitored for five years. To explore the relative importance of biotic and abiotic drivers, we assessed the effects of community type, vegetation removal, herbivore exclusion, shrub introduction and microclimate on the establishment, growth and survival of pine seedlings, and whether these effects vary among distinct community types, over a period of five years. We compared treatment effects on the invasibility of a heath, a meadow and a Salix shrubland, three representative plant community types at an alpine tundra site in Central K E Y W O R D S above-ground competition, alpine tundra, exclosure, herbivory, invasibility, microclimate, Pinus sylvestris, shrub encroachment 2 | MATERIAL S AND ME THODS

| Study site and plant community description
The study was conducted in the low-alpine zone near Hjerkinn (62.22° N, 9.56°E) at Dovrefjell, Central Norway, a part of the Scandes mountains ( Figure 1a). Here, Scots pine (Pinus sylvestris L., Pinaceae, hereafter referred to as pine) exist as scattered trees up to the tree line, which is dominated by birch (Betula pubescens ssp. czerepanovii (N.I.Orlova) Hämet-Ahti) (See The Norwegian Biodiversity Information Centre (NBIC) (https://www.biodi versi ty.no/, accessed 5 October 2020) for unified nomenclature of plant species.), and smaller pine individuals occur sporadically above the treeline. The field sites are all located on podzolic soils around 1,100 m above sea level just above the local treeline. In the period between January 2013 and December 2018, the mean February and July temperatures were −6.4°C and 11.5°C, respectively, and the annual mean precipitation was 531 mm at the closest weather station (Hjerkinn II,1,012 m a.s.l.,62.22° N,9.54° E;Norwegian Meteorological Institute,eklima.met.no). Study sites were selected within three common alpine plant community types in the alpine tundra ecosystem: (1) evergreen dwarf-shrub heath dominated by Empetrum; (2) meadow with mixed herbaceous vegetation of grasses, forbs and cryptogams; and (3) deciduous shrubland dominated by Salix sp.
with a heterogenous ground layer rich in bryophytes and lichens (see Appendix S1, Table S1, for community characteristics and species composition). The three plant communities were situated on different mountain slopes with similar aspect and elevation and located within 5 km of each other (Figure 2), thus sharing roughly the same macroclimate. All sites are subjected to low-intensity summer grazing by Norwegian white sheep (Ovis aries) (Norwegian Institute of Bioeconomy Research: http://kilden.skogo gland skap.no/), and animal husbandry has probably been present in the area for about 400 years BC (Risbøl, Stene, and Saetren, 2011

| Study design
The experiment was established in 2013 as a randomized block design within each of the three plant communities, with eight replicates (blocks) per treatment . The eight blocks were randomly located in each plant community. Within each block, four plots (25 cm × 25 cm) were randomly assigned to a full-factorial F I G U R E 1 (a) Location of the study area in the low-alpine zone near Hjerkinn (62.22° N, 9.56° E), Dovrefjell, Central Norway. (b) Schematic overview of the treatments within one block, replicated eight times per plant community. Each block consisted of four plots, with a factorial combination of the treatments herbivore exclosure (yes, no) and willow transplants (yes, no). Each plot was subdivided into four subplots assigned to a factorial combination of a Pinus sylvestris seeding treatment (yes, no) and a canopy removal treatment (yes, no  and the two species were therefore randomly distributed among the plots (the two species commonly occur in mixed stands in the study area). To exclude herbivores, 80 cm × 80 cm × 50 cm cages were placed permanently over half of the plots. The cages were constructed from galvanized iron with a mesh size of 1.27 cm × 1.27 cm, and buried 5-10 cm into the soil .
In all seeded subplots, 10 pine seeds were sown in late autumn 2013. Seeds were supplied by the Norwegian Forest Seed Center, and originated from a natural forest near Oppdal (600-650 m above sea level), 50 km north of the study sites. While seeding, a cardboard box was placed around the subplot to protect against wind and to make sure that the subplot received exactly 10 seeds. The unseeded subplots provide an experimental control for spontaneous emergence at the study sites and were not used directly in the analysis.
In half of the subplots, all above-ground biomass of all co-occurring plants was removed to ground level to reduce above-ground interactions of surrounding species with the pine seedlings. Vegetation removal was done with a knife, leaving soil and roots intact.
In summary, the experiment comprised three community types × eight blocks × two vegetation-removal treatments × two herbivore treatments × two willow transplant treatments × two sowing treatments = 384 subplots.

| Seedling emergence, survival, and performance
Pine seedling emergence was monitored yearly in all subplots during summer or early autumn from 2014 to 2016. The emerged seedlings were assigned a unique ID, marked with a toothpick F I G U R E 2 (a) Spatial configuration of the three vegetation communities, and spatial configuration of plots within communities (b) Salix shrubland, (c) Meadow and (d) Heath and marked on a seedling map, so that every seedling could be followed individually. Litter was removed to facilitate seedling counts, and subsequently replaced. In the first growing season In 2018 only, the performance of the pine seedlings was quantified in terms of their growth and condition. Pine seedling performance was measured in three ways: stem height, new stem growth and the fraction of healthy needles per seedlings. Stem height was measured as the length of the stem from the soil to the highest point of the stem, pressing the measuring stick firmly into the ground to minimize the deviation due to the moss layer. New stem growth was defined as the length of the green part of the main stem, which indicates the yearly seedling growth (Holmgren et al., 2015). The fraction of healthy needles per seedling was based on the colour of the needles.
Colour change in needles is a good indicator of stress and nutrient deficiency in Pinus sylvestris (Hytönen and Wall, 2006). All needles were counted and scored as either 'healthy' (when the needle was fresh and green) or 'unhealthy' (when the needle had turned yellow or brown) and the fraction of healthy needles was monitored per seedling. Seedlings that were missing, or had only brown needles, were scored as dead.

| Microclimate
To quantify microclimatic conditions, we measured soil temperature, soil moisture and light availability for every subplot.

| Data analysis
The invasibility of the study sites to pine seedlings were tested with linear mixed models (LMMs) with Gaussian error distributions or were treated as random factors (see Appendix S2). Because our focal community types were concentrated in one site, we focused our hypothesis testing on the treatment effects, and their possible variation among communities. A significant interaction between community and treatment provided evidence that the treatment effects differed among communities. To further explore these differences, we fitted models for each community separately following: To test for effects of plant community type and treatment on microclimate (soil temperature, moisture and light availability), we fitted LMMs with Gaussian error distributions (see Appendix S2).
Some of the soil temperature variables exhibited multicollinearity.

MARSMAN et Al.
Therefore, we analyzed only maximum summer temperature (which correlated with mean summer temperature, r = 0.85) and minimum winter temperature (which correlated with mean winter temperature, r = 0.98), because temperature extremes are most likely to limit establishment. We also analyzed soil moisture and light availability.
Block and plot (nested within block) were treated as random factors.
As above, we fitted models for each community when treatment effects differed among communities. Statistical analyses were performed in R version 3.4.4, using the functions lmer and glmer from the package lme4 (Bates et al., 2014) for model fitting, the function drop1 from the base package for backward selection, and the function dunn.test from package dunn.

| RE SULTS
A total of 578 pine seedlings (30% of sown seeds) emerged during the first three years of the experiment and 159 (almost 30%) of the emerged seedlings survived until the fifth year. Mean emergence and survival rates tended to be similar across the three communities ( Figure 3).

| Differences in treatment effects on pine establishment among communities
The effects of the treatments on seedling emergence and performance differed among communities, while effects on seedling survival were consistent (Appendix 3, Table S3.1, Figure 3). In the following analyses, the three-way interactions among vegetation removal, herbivore exclusion and willow introduction were never statistically significant, and we focus only on direct and two-way interaction effects.
Pine seedling survival increased when protected from herbivores (χ 2 = 15.76, p < 0.001) for all communities. The effect of herbivore exclusion on stem height was inconsistent among sites (Appendix S3; Table S3.1, Figure 3). /in the shrubland and heath, the effect depended on complex interactions with vegetation removal and willow introduction, respectively, while we detected no effect at the meadow. When herbivores were excluded, pine seedlings had greater fractions of healthy needles on the heath and Salix shrubland F I G U R E 3 Effects of experimental treatments on pine emergence, survival and performance in seeded subplots in the three different plant communities after fiveyear (means ± standard error). (a-c) Pine emergence per seeded subplot, as fraction of seeds sown (10 seeds per subplot); (d-f) pine survival per seeded subplot as fraction of emerged pines; (g-i) mean stem height per pine seedling; (j-l) mean new stem growth per pine seedling; and (m-o) mean fraction healthy needles per pine seedling. Different letters represent statistically significant differences between treatments (lower case) or communities (upper case) (Tukey HSD or Dunn's test, p < 0.05). For survival probability, hypothesis testing was performed jointly across the three communities | 7  (χ 2 = 4.17, p = 0.041, χ 2 = 6.63, p = 0.010, respectively; Table 2, Figure 3), but not in the meadow.

| Willow transplant effect on pine establishment
The

| Biotic treatment effects on microclimate
Compared to the Salix shrubland, the heath was warmer in summer and colder in winter and the meadow community was generally intermediate. The effects of the treatments on the temperature variables were consistent among communities (Appendix 3, Table   S3.2). Vegetation removal increased maximum summer temperatures (F = 22.78, p < 0.001, Appendix 3, Table S3.3, Figure 4) and decreased minimum winter temperatures (F = 10.37, p = 0.002).
Maximum summer temperatures were generally lower inside the exclosures (F = 17.14, p < 0.001, Appendix 3, Table S3.3, Figure 4), and minimum winter temperatures were higher in exclosed plots compared to open plots (F = 26.69, p < 0.001). Treatment effects on soil moisture were highly variable among communities (Appendix 3, Table S3.2, Figure 4). Vegetation removal increased light availability at all communities and in the exclosed plots light availability was also lower at the heath and meadow (Appendix 3, Table S3.4, Figure 4).

| Relationship between pine establishment and microclimate
None of the microclimatic variables detectably affected pine emergence. Seedling survival tended to increase with warmer maximum summer temperature, minimum winter temperature and light availability, although the estimated effects were weak (Table 3, Figure 5).
Seedlings in plots characterized by high maximum summer temperatures, moister soils and higher light availability had greater fractions of healthy needles. Furthermore, seedlings grew taller in plots where less light was available. All of the estimated effects were weak.

| D ISCUSS I ON
We detected strong effects of biotic factors (

| Pine seedling establishment in alpine plant communities and vegetation interaction effect
In intact vegetation, pine seedlings emerged and survived about equally well in all three alpine tundra community types we considered, but establishment rates were generally low. This low invasibility of intact tundra vegetation is in line with previous studies reporting predominantly negative effects of tundra vegetation on tree seedling recruitment (Hättenschwiler and Körner, 1995;Loranger et al., 2017;Lett and Dorrepaal, 2018). As expected, experimental reduction of competition through vegetation removal strongly increased invasibility. Vegetation removal had positive effects on emergence in the Salix shrubland and in the heath and meadow communities when combined with willow introduction.
Seedling survival was considerably better at all three sites when canopies were removed when this effect was combined with willow introduction. This illustrates biotic resistance of alpine plant communities also for later life stages of the pine seedlings. The negative effects of tundra vegetation on tree seedling recruitment probably act through competition for light, nutrients, water and space, but could also relate to allelopathy or higher susceptibility to pathogen infections in dense vegetation (Sedia and Ehrenfeld, 2003;Loranger et al., 2017;Lett and Dorrepaal, 2018).

TA B L E 1
Parameter estimates from the selected minimal linear model describing direct and interaction effects of vegetation removal (R), herbivore exclosure (E), and willow transplants (T) on survival probability Furthermore, pine seedlings in intact vegetation had a lower fraction of healthy needles than in vegetation removal subplots in heath and in Salix shrubland when combined with the introduction of willow transplants, suggesting a negative impact of the standing vegetation on tree seedling performance. On the other hand, seedlings in the heath were taller in intact than in vegetation removal subplots, perhaps due to facilitation through protection from, for instance, strong abrasive winds during periods with shallow snow cover (Batllori et al., 2009;McIntire, Piper, and Fajardo, 2016;Piper et al., 2016). Also in the Salix shrubland, when protected from herbivores, seedlings grew taller in intact subplots than vegetation-removal subplots. In addition to the presence of a shrub canopy, the Salix shrubland is characterized by relatively thick understorey ground layers of lichens or bryophytes (mean thickness ± SD in mm: 73.7 ± 30.6). Therefore, the greater height of the seedlings in undisturbed Salix shrubland could be attributed to the need of outgrowing this ground layer to reach high-light conditions. Consistent with this hypothesis, we found that the Salix shrub community was associated with the lowest light availability of all three communities. Decreased tree growth is often the cost of recruiting below shrubs (Castro et al., 2004;Kambo and Danby, 2018), but this was not evident for the performance measures in our study. Since we found a strong interaction between vegetation removal and willow introduction, Salix shrublands might provide favourable regeneration sites, provided that gaps and enough light are available in the vegetation.

| Herbivory reduces invasibility
We detected strong effects of experimental herbivore exclusion, suggesting that sheep, rodents and other herbivores affect pine seedling emergence, survival and performance. Rodents may also have eaten some of the experimental seeds (Nilson and Hjältén, 2003;Nystuen et al., 2014). Interestingly, the study area experienced a rodent population build-up during the year of seed sowing (2013), resulting in a TA B L E 2 Parameter estimates from the selected minimal (generalised) linear models describing direct and interaction effects of vegetation removal (R), herbivore exclosure (E), and willow transplants (T) on pine emergence, stem height and fraction of healthy needles in the heath, meadow and Salix shrubland   low rodent peak in 2014 when most pine seeds germinated (Framstad, 2017). Previous studies have shown that small rodents have a stronger effect on alpine plant communities than do large herbivores (Olofsson et al., 2004;Bougnounou et al., 2018). These results add to previous studies (Moen, Lundberg, and Oksanen, 1993;Boulant et al., 2008;Post and Pedersen, 2008;Munier et al., 2010;Bougnounou et al., 2018) in suggesting that herbivory may be important in limiting tree regeneration in alpine tundra ecosystems.

| Willow introduction increases invasibility when above-ground vegetation has been removed
Willow introduction resulted in higher pine seedling emergence in the heath and meadow communities and higher pine survival in all three communities, but only when combined with vegetation removal, which could indicate facilitation from the willows on pine seedling recruitment in these short-statured vegetation types (Akhalkatsi et al., 2006;Holmgren et al., 2015;Chen et al., 2020). The introduction of willows had no detectable effect on the microclimate (Figure 4), and transplants were relatively small. Therefore, the effect observed here on invasibility may not relate to amelioration of microclimatic conditions as previously suggested (Sturm et al., 2005;Holmgren et al., 2015). Instead, this apparent facilitative effect may relate, for example, to changes in nutrient content (Chen et al., 2020), soil biota, ectomycorrhizal activity (Nara, 2006) or increased CO 2 levels due to higher soil respiration . Disentangling the mechanisms behind this apparent facilitative effect will require further studies as the introduced willows grow larger.

| Weak effects of microclimate on invasibility
Contrary to expectations, pine seedling survival and fraction of healthy needles increased with higher maximum summer temperatures. Thus, high summer temperatures do not seem to limit seedling establishment and performance. In addition, seedling survival increased with higher minimum winter temperatures. Although temperature extremes are more ecologically meaningful for explaining seedling establishment than are temperature means, we could not distinguish with certainty between the variables, due to their correlation. Magnitudes of all microclimate effects were very small (Table 3) and statistical support was weak ( Figure 5), indicating only subtle effects of abiotic factors on pine invasibility in this system.
These results suggest that abiotic factors are of limited importance for pine establishment compared to the biotic drivers discussed above. It is possible that abiotic stress is an important factor driving differences in woody encroachment at larger scales, but on the scale we operated, abiotic factors seem to play limited roles compared to biotic factors.

| CON CLUS IONS
Our results demonstrate that, when experimentally introduced into treeless alpine tundra, Pinus sylvestris seedlings have the potential to emerge and establish in all three plant community types considered.
Despite successful initial establishment, the pines remained small, depending on their microsite and community characteristics (Körner, 2012;Brodersen et al., 2019). Furthermore, we provide field evidence that biotic factors are the key drivers of pine seedling establishment into the alpine tundra ecosystem. Above-ground vegetation biomass and herbivory inhibited both pine seedling establishment and performance, and this inhibition was relieved the most when both factors were removed. In contrast, seedling responses to variation in microclimate were subtle. This suggests that effects of climate on vegetation dynamics in alpine ecosystems are mediated through disturbances and herbivory. Studies aiming to predict future vegetation changes should therefore incorporate local biotic interactions in addition to abiotic factors even in alpine communities.

ACK N OWLED G EM ENTS
We thank all ECOSHRUB field team members (Andreas Baele,  F I G U R E 5 Regression plots showing the statistically significant (p < 0.05, Table 3) relationships between: (a) survival probability and maximum summer temperature; (b) fraction healthy needles and maximum summer temperature; (c) survival probability and minimum winter temperature; (d) fraction healthy needles and moisture; (e) survival probability and light availability;(f) fraction healthy needles and light availability; and (g) stem height and light availability. Other explanatory variables in the models are kept constant at their median values. Non-significant relationships are not shown. Shaded areas indicate 95% confidence intervals