Introduction

Many alpine treeline ecotones have shown upward elevational shifts as a result of climate warming (see reviews by Holtmeier and Broll 2005; Harsch et al. 2009). Further treeline advances are anticipated because, in the absence of large-scale natural disturbances or human land use, low temperature appears to be the primary cause of high altitude treeline formation globally (Körner and Paulsen 2004). In the Swiss Alps, where temperature has increased more rapidly than the global average (Beniston et al. 1997), the treeline might advance at higher rates than expected based on the global average. It has been suggested that many treelines may currently be below their potential maximum elevation even in the absence of anthropogenic impacts, and their reported advances may lag behind current climate warming rates (Holtmeier 1994; Harsch et al. 2009). During these lag phases, factors other than temperature may be important for treeline dynamics at regional and local scales (Holtmeier and Broll 2005; Danby and Hik 2007).

The establishment of seedlings (≤10-year-old trees, sensu Smith et al. 2003) and their performance during early years are key processes determining shifts in treeline positions (Germino et al. 2002; Smith et al. 2009), and it is essential to analyze seedling establishment, survival, and growth in combination with various environmental factors to be able to understand the underlying mechanisms of treeline dynamics. Seedlings, moreover, respond differently to environmental drivers compared to upright trees in later stages of development because they are less coupled to atmospheric conditions than upright trees (Wilson et al. 1987) and often have different sensitivities to environmental factors (Hofgaard et al. 2009; Bansal and Germino 2010). Numerous studies have investigated how seedlings of treeline species respond to changes in a variety of environmental factors, such as soil temperature (e.g., Domisch et al. 2002; Peng and Dang 2003; Juntunen and Neuvonen 2006), exposure to light (e.g., Johnson and Smith 2005), nitrogen availability (e.g., Sveinbjörnsson et al. 1992; Kabeya 2010), drought (e.g., Hättenschwiler and Körner 1995), competition and facilitation (e.g., Moir et al. 1999; Germino et al. 2002; Batllori et al. 2009; Venn et al. 2009), and herbivory (e.g., Myster 2009). While interactions between several of these factors have been studied before (e.g., Domisch et al. 2002; Myster 2009; Munier et al. 2010), there are only few studies on the relative importance of multiple factors (e.g., Kabeya 2010; Barbeito et al. 2012). In the present study, we measured proxies for four of these important factors—soil temperature, light conditions, water use efficiency, and nitrogen availability—and evaluated their relative importance for seedling performance of seven tree species at and above the current treeline.

Analyses of stable carbon and nitrogen isotopes (δ 13C, δ 15N) can provide an integrated measure of water use efficiency and plant nitrogen availability (Farquhar and Richards 1984; Craine et al. 2009; Resco et al. 2011). However, foliar δ 13C is also influenced by temperature, atmospheric pressure, and higher carboxylation efficiency at higher elevations, and is, therefore, expected to increase with elevation (Körner et al. 1991; Körner 2003), making causal interpretations difficult. Foliar δ 15N has been shown to be closely linked to soil N availability (Garten and Van Miegroet 1994; Craine et al. 2009) where δ 15N values decrease with tighter N cycles, characterized by stronger microbial N immobilization, a smaller and more available soil N pool, and higher plant N-uptake rates via mycorrhizal associations (Hobbie et al. 2000; Hobbie and Colpaert 2003).

Our analysis of early species-specific seedling performance, over several years and along an elevational gradient, aims at a more detailed understanding of the key processes determining the position of the upper treeline and of potential future species-specific elevational shifts. Seeds of seven regionally important tree species were sown at three sites along an elevational transect, and seedlings were monitored over four growing seasons. We hypothesized that (1) measures of seedling performance—proportions of living seedlings, and seedling growth and biomass—should be higher at sites at or below the current treeline than above, and that such performance differences among sites should reflect (2) the species’ current elevational ranges, and (3) local environmental factors such as nitrogen or light availability, in addition to soil temperature.

Methods

Study area and tree species

We conducted the study across an elevational gradient (1,990–2,400 m a.s.l.) on a NE-exposed slope of the Dischma Valley near Davos (Switzerland). The transect crosses the Stillberg treeline research site, where the treeline is currently at 2,100 m a.s.l. (Barbeito et al. 2012). At the meteorological station Stillberg, at 2,090 m a.s.l., the annual mean air temperature is 2.0 °C and the annual precipitation is 1,150 mm (years 1975–2005). The average snow cover lasts from mid October to the end of May, with a difference of 1–2 weeks in snowmelt date between the highest and lowest site of the transect (Walder 1983). The growing season lasts (on average) from end of May to end of September (Rixen et al. 2012). The local treeline ecotone is characterized by a Larch-Swiss stone pine forest (Larici-Pinetum cembrae) with dwarf shrubs on siliceous (Paragneis) bedrock (Schönenberger 1975).

Seeds of the tree species Alnus viridis (Chaix) DC., Larix decidua L., Picea abies (L.) Karst, Pinus cembra L., Pinus mugo ssp. uncinata Ramond, Pinus sylvestris L., and Sorbus aucuparia L. were used; all of which occur naturally in close vicinity of the study area, but have different elevational ranges and thus are thought to represent a range of temperature sensitivities: P. cembra, L. decidua, and P. mugo are treeline species, whereas P. abies, P. sylvestris, S. aucuparia, and A. viridis have lower elevational ranges (Table 1; Brändli 1998). The seeds were obtained from natural populations (one population per species) from the Swiss Central Alps in the upper elevational range of each species (Table 1), and the numbers of seeds sown per plot were adjusted to the expected species-specific viability rates based on greenhouse germination tests (A. Burkart, pers. comm.). Due to the low initial germination success of P. cembra and S. aucuparia in the first year of the experiment, pre-germinated seedlings of these two species were added in the second year.

Table 1 Current elevational ranges of seven species in Switzerland, with ranges spanning 90 % of all observations in Switzerland (Brändli 1998); and elevation of seed sources (A. Burkart, personal communication)
Full size table

Experimental design

The elevational transect was set up in May 2007, with one site each at 1,990 m (ca. 100 m below the current treeline), 2,220 m (ca. 100 m above the current treeline), and 2,400 m a.s.l. (ca. 300 m above the current treeline). All sites were in areas with similar microtopography in treeless alpine vegetation. The three sites were outside forested areas and without direct shading from trees, but with varying densities and heights of surrounding ground vegetation, which consisted mostly of graminoid and herbaceous species, with scattered deciduous and evergreen dwarf shrubs (e.g., Dawes et al. 2011). Potential N-fixing species (e.g., Lotus alpinus (DC) Ramond) occurred at similar very low densities at all sites. Each elevation level (“site”) contained five replications (“groups”) of seven species-specific circular plots (“plots”) of ca. 15 cm diameter. Within each site, each of the five groups spanned a diameter of 2–3 m, with distances of 5–6 m between group centers. Within each group, the seven species-specific plots were located at distances of 0.5–2 m between plots. At the beginning of the experiment, above-ground vegetation was removed inside and within a 10-cm radius of each plot. A 2-cm layer of sieved soil taken from the top 5-cm layer near the intermediate elevation site was added to each plot before sowing. This treatment ensured similar starting conditions for all plots. Reinvasion of ground vegetation into plots was minor, and any reinvading plants were removed twice yearly. Plots were covered over the course of the experiment with a 2-mm metal mesh attached to a plastic ring around the seeds, in order to exclude herbivores while only slightly changing light availability.

Measurements

For each plot, photosynthetically active radiation (PAR) was measured once in July 2010, when the surrounding ground vegetation (outside the weeded 17.5 cm radius from the center of the plots) had reached its maximum height. Measurements were taken with a light sensor (LI-190SA Quantum Sensor, LI-COR Environmental, USA) held horizontally directly above the seedling canopy at c. 10 cm height, and well above any grass or shrub layers near the seedling canopies at 60 cm height. In order to avoid direct radiation that could negatively influence the measurements of relative differences in radiation, the measurements were taken during mostly cloudy conditions. The percentage PAR reaching the seedling crowns, relative to the PAR at 60 cm (radC), provides an indirect estimate for light competition in the immediate surroundings of the plots.

Soil temperatures at 5 cm depth were measured using one data logger (Thermochron iButtons, Maxim Integrated Products, USA) per site at 0.5 h intervals during July–September 2010. Due to the lack of continuous data over the whole experiment, soil temperatures were not used as predictor variable in seedling performance analysis, but used for post hoc interpretation of site differences instead.

For each plot, the seedlings were counted and measured twice per year during 2007–2010 (only once in 2008), over a total of 39 months. Yearly census dates were early and late in the vegetation period (June and October). Since it was impossible to follow the many seedlings individually, germination and survival rates could not be explicitly calculated, and the proportion of living seedlings relative to the expected number of viable seeds per plot was calculated instead for each time point at M months after sowing (%aM).

Individual shoot length was measured to the nearest mm from the soil surface to the apical meristem for three random seedlings per plot, and averaged for each plot. Relative shoot growth rate was calculated as RGRsh = [ln(L 2010) − ln(L 2007)] × Δt −1, where L 2007 and L 2010 were shoot lengths at the end of the growing seasons of 2007 and 2010, respectively, and Δt was 3 years (growing seasons 2008–2010). For the two species with delayed germination (P. cembra and S. aucuparia), the last two growing seasons were used (2009–2010). To assess accumulated biomass at the end of the experiment, three seedlings per plot were harvested in September 2010, separated carefully with tweezers, cleaned, and separated into roots, stems, and foliage. The seedlings were oven-dried for 2 days at 60 °C and weighed to ±0.1 mg accuracy for all compartments separately, but pooled over the three individuals per plot. These biomass (bm) data were also used for the calculation of root to shoot ratio, R:S = bmroot/(bmstem + bmfoliage), and the foliar mass ratio, FMR = bmfoliage/bmtotal. Foliar C and N concentrations (%C, %N) and stable C and N isotope ratios (δ 13C, δ 15N) were measured from the dried foliage samples (see Online Resource 1A).

Data analysis

To analyze potential differences in nine variables (%a39, RGRsh, bm, R:S, FMR, and foliar %N, N allocation, δ 15N and δ 13C) among sites for each species, we used 50–50-Manova (Langsrud et al. 2007; see Online Resource 1B).

In a second analysis, we examined the relationship of seedling performance with different environmental variables, in a mixed effects regression model on the data pooled over all species (spc) and sites. The response variables used as measures for seedling performance were %a4, %a16, %a28, %a39, i.e., the proportion of living seedlings at the end of each vegetation period, and RGRsh, bm, R:S, and FMR. For each response variable (y), site, spc, radC, and δ 15N were used as fixed predictors, and group identity (1–15) as random factor (Eq. 1). The spc:radC interaction was included because L. decidua, P. sylvestris, P. mugo were expected to require more light than the other species (Landolt 2010), and the spc:δ 15N interaction was included after an initial correlation test showed that δ 15N was positively correlated with %N and negatively correlated with R:S in the same three species (Table 2).

Table 2 Species-specific Pearson’s correlation coefficients (r) between foliar N concentration (%N) and foliar δ 15N natural abundance heavy isotope, and between foliar δ 15N and root to shoot ratio (R:S)
Full size table

We did not include δ 13C as a predictor variable because it was not significantly different from the expected altitudinal lapse rate (see Results section). To better meet the assumptions of the mixed effects regression model, response variables were transformed where necessary (see Online Resource 1B).

$$ y_{\text{transf}} \sim {\text{site}} + {\text{spc}} { } + {\text{radC}} + \delta^{15} {\text{N}} + {\text{spc}} :{\text{radC}} + {\text{spc}} :\delta^{15} {\text{N}} + (1|{\text{group}}) $$
(1)

The significance of the predictor variables was assessed with likelihood ratio tests (Bolker et al. 2009), and a Bonferroni correction was used for the four time-series response variables (see Online Resource 1B). Overall model fit for each response variable was calculated as linear regression of the predicted versus the observed values (Byrnes et al. 2006). The parameter estimation methods and results can be found in the Online Resource 1B.

Results

Site characteristics

The relative light availability at the seedling crown level (radC) increased with site elevation, with means (±SD) of 79.3 % (±5.9 %) at the lowest site, 86.6 % (±1.7 %) at the intermediate site, and 95.0 % (±0.5 %) at the highest site.

Soil temperature at 5 cm depth averaged over the 0.5 h-measurements in July–September 2010 decreased with site elevation: mean (±SD) temperatures were 10.9 (±3.2) °C at the lowest site, 9.3 (±4.1) °C at the intermediate site, and 9.0 (±4.8) °C at the highest site. Site differences were more pronounced during cloudy weather (4–5 °C), while in clear weather, temperature differences were smaller (<1 °C). Overall, 0.5 h-measurements in July–September 2010 ranged from 4.4 to 22.8, 2.2 to 22.5, and 1.2 to 24.8 °C at the lowest, intermediate, and highest site, respectively. No temperatures below 0 °C were recorded. Temperatures below 2 °C, at which the surface soil may have been frozen, were only recorded at the highest site, during a total of c. 93 h in September. Soil temperatures below 5 °C, the approximate minimum temperature required for tissue formation (Körner 2003), occurred during 4, 15, and 22 % of the temperature measurement period at the lowest, intermediate, and highest site, respectively.

Seedling characteristics

Foliar carbon and nitrogen concentration and isotopes

No significant differences were found among sites for foliar %C (data not shown). The elevational differences in δ 13C were not significantly different from the expected mean lapse rate of 1.2 ‰ per 1,000 m (Körner et al. 1991; Körner 2003). Foliar N concentration (%N) and N allocation significantly decreased from the lowest to highest site in L. decidua and P. cembra (by 0.48 and 0.46 % for %N, and by 0.84 and 1.76 mg for N allocation, respectively), but were lowest at the intermediate site in P. abies and P. sylvestris (Fig. 1). δ 15N significantly decreased with elevation in three species (L. decidua, P. mugo, P. sylvestris; by 3.3, 3.0, and 2.6 ‰, respectively, from the lowest to highest site), but increased by 1.1 ‰ from the lowest to intermediate site in P. abies.

Fig. 1
figure 1

Seedling characteristics per species (panels) and for three sites (elevations m a.s.l. on x-axis). Each row of panels shows one variable: Foliar nitrogen concentration relative to dry weight (%N), foliar N allocation, and foliar heavy isotope ratios δ 15N and δ 13C. Different letters (a, b, c) indicate significant among-site differences (50–50-MANOVA, Tukey Honest Significant Difference tests, p < 0.05; see Online Resource 1B for details). Labels were omitted where no significant differences between sites were found. Due to the low initial germination success of P. cembra and S. aucuparia in the first year of the experiment, pre-germinated seedlings of these two species were added in the second year

Full size image

Proportions of living seedlings

The proportion of seedlings alive at the end of the experiment (%a39) relative to the number of viable seeds sown (Table 3) was not significantly different among sites except in P. abies, where the mean %a39 decreased by 28 % between the lowest and highest site and by 38 % between the lowest and intermediate site (Fig. 2). Changes in proportions of living seedlings over time (Fig. 3) differed markedly among species. P. cembra and S. aucuparia had low initial germination success, whereas in A. viridis, despite high initial germination success at the lowest site, almost all individuals died before the end of the experiment (Table 3; Fig. 3), and the species was excluded from further analysis. All species except S. aucuparia and P. cembra tended to show larger differences among sites early in the experiment with converging proportions of living seedlings among sites toward the end of the experiment (Fig. 3).

Table 3 Species-specific expected number of viable seeds (based on seed quality assessed in a greenhouse germination experiment; A. Burkhart, personal communication), mean (±SD) number of surviving seedlings at the end of the experiment (39 months), and mean (±SD) seedling height of three seedlings in cm
Full size table
Fig. 2
figure 2

Seedling characteristics per species (panels) and for three sites (elevations m a.s.l. on x-axis). Each row of panels shows one variable: Percentage of living seedlings after four vegetation seasons (39 months) relative to the number of viable seeds sown, relative shoot growth rate (RGRsh), total dry weight biomass per seedling, root to shoot biomass ratio (R:S), and foliar mass ratio (FMR). The significant differences between sites are labeled as in Fig. 1. Due to the low initial germination success of P. cembra and S. aucuparia in the first year of the experiment, pre-germinated seedlings of these two species were added in the second year

Full size image
Fig. 3
figure 3

Percentage of living seedlings relative to viable seeds sown in May 2007, over four vegetation seasons at three sites (1,990, 2,220, and 2,400 m a.s.l.) along an elevational transect across and above the treeline in the Swiss Alps. Mean ± SD (5 replications per site and species) are shown for all censuses at 4–39 months after sowing. Growing seasons (May–September) are shaded in gray. Due to the low initial germination success of P. cembra and S. aucuparia in the first year of the experiment, pre-germinated seedlings of these two species were added in the second year (dashed lines)

Full size image

The regression analyses for the yearly %aM indicated that site location and species identity were significant predictors, whereas no relationship with δ 15N was found. A significant and species-specific relationship of the proportion of living seedlings with light availability was found 28 months after sowing but not at any other census (Table 4; see also Table S1 and Fig. S1a in Online Resource 1B).

Table 4 Significance (p values) of predictor variables (rows) site, species (spc), light availability (radC), and N isotope ratio (δ 15N) for the eight variables of interest (columns) percentage of living seedlings (%a4, %a16, %a28, %a39), relative shoot growth (RGRsh), dry weight biomass per seedling (bm), root to shoot ratio (R:S), and foliar mass ratio (FMR)
Full size table

Seedling growth

The relative shoot growth rate (RGRsh) decreased from the lowest to highest site in L. decidua, P. abies, and P. mugo, by 0.41, 0.21, and 0.16, respectively (Fig. 2). However, in P. mugo the decrease in mean RGRsh from the lowest to intermediate site (0.21) was largest, and in P. sylvestris the mean RGRsh at the intermediate site was significantly lower than at both other sites. In all species except S. aucuparia, the mean RGRsh was ≥0.2 at all three sites. Absolute seedling height varied among species and among sites, but showed relatively low variability among plots (Table 3).

Mean seedling biomass decreased from the lowest to highest site in L. decidua, P. cembra, and P. mugo, by 88, 37, and 51 %, respectively (Fig. 2). In P. abies and P. sylvestris, differences between the lowest and intermediate site (−69 and −81 %, respectively) were largest, while those between the lowest and highest site were not significant (Fig. 2).

Biomass allocation to roots (R:S) increased from the lowest to highest site in L. decidua, P. abies, P. mugo, and P. sylvestris, by 0.44, 0.23, 0.12, and 0.22, respectively (Fig. 2). This shift in R:S mostly reflected reduced stem biomass relative to root biomass, as the proportion of biomass allocated to foliage (FMR) was not significantly different among sites for most species (L. decidua, P. cembra, P. mugo, S. aucuparia), or showed the lowest mean at the intermediate site (P. abies). Only P. sylvestris had the highest FMR at the lowest site (Fig. 2).

Apart from site and spc, δ 15N was a significant predictor of bm, R:S and FMR, with a positive relationship with bm and R:S, and a negative relationship with FMR. Moreover, the responses to δ 15N were species-specific for bm and FMR (Table 4; see also Table 1 and Fig. S1b–d in Online Resource 1B). Light availability (radC) showed no relationship with growth or biomass.

Discussion

Proportions of living seedlings

Despite some site differences in seedling dynamics (Fig. 3), our first hypothesis that seedling performance should decrease with increasing elevation was not supported by the proportions of living seedlings. Their means were similar among sites, and higher than expected based on past findings in the European Alps (Hättenschwiler and Körner 1995; Körner 2003), the boreal treeline in Finland (Juntunen and Neuvonen 2006), the Pyrenees (Batllori et al. 2009) and the Snowy Range, Wyoming (Germino et al. 2002). The differences between our results and these previous studies may have resulted from factors that were not studied in our experiment, such as seed dispersal, herbivory, facilitation, or competition (e.g., Moir et al. 1999; Juntunen and Neuvonen 2006; Myster 2009). Apart from soil temperature, spring or summer frosts have a strong impact on seedling survival (Venn et al. 2009), although frost events were mostly absent during our experiment (Rixen et al. 2012). In the absence of such events, the observed high proportions of living seedlings may suggest that the seedlings may only become limited at a later age or size (e.g., Barbeito et al. 2012), or that the local treeline is not in equilibrium with climate.

Likewise, our second hypothesis, that species-specific performance should broadly match their current elevational ranges (Table 1), was also not confirmed by the proportions of living seedlings. The three species currently found at or near the treeline, L. decidua, P. mugo, and P. cembra, showed no significant decrease in their proportions of living seedlings with site elevation. P. sylvestris and P. abies—common at elevations 400–500 m below the regional treeline—were more successful than expected at the highest site 300 m above the current treeline, with >20 % seedlings alive 39 months after sowing in both species. The results for P. sylvestris agree with a previous study, which suggested that in the Swiss Alps, this species is more limited by species interactions than by temperature (Hättenschwiler and Körner 1995).

While P. cembra and S. aucuparia germinated poorly, their performance in terms of proportions of living seedlings and growth was high after we added pre-germinated seedlings. Based on their current elevational ranges (Table 1), P. cembra would be expected to germinate well at least at the two lower sites, while the occurrence of S. aucuparia at high elevations may be determined more by vegetative reproduction. Similarly, even though A. viridis is common in Swiss alpine treeline ecotones, our experiment suggests rather poor sexual reproduction success at high elevations. Species-specific seedling dynamics over time (Fig. 3) indicated that different species were likely limited at different stages of their development. Even though the seedling performance data were not significantly correlated with seedling densities, the high seedling densities in our plots may have led to interactions among the seedlings (Moir et al. 1999; Batllori et al. 2009; Venn et al. 2009), and we could not entirely exclude seedling interactions as potential co-drivers of seedling performance.

Our third hypothesis, that proportions of living seedlings should be related to nitrogen or light availability, was only partly confirmed. The proportions of living seedlings were not significantly related to δ 15N. Light availability (radC) was a significant and species-specific predictor of the proportion of living seedlings at one point in time (%a28; Table 4, see also Table S1 and Fig. S1a in Online Resource 1B). However, the differences in %a28 did not correspond to the species-specific light requirements, i.e., the relationship of %a28 with radC was not more positive in the more light-demanding species L. decidua, P. sylvestris, or P. mugo, compared to the other species. While light availability may have an influence on early seedling survival (Johnson and Smith 2005), our results suggest that other factors not studied here may have been more important for the species analyzed here during the course of our experiment.

Seedling growth

According to our first hypothesis, seedling RGRsh and biomass decreased with increasing elevation in most species. These results are in agreement with a previous study of four subalpine tree species (L. decidua, P. cembra, P. mugo, and P. abies) in the same geographic area (Turner et al. 1975), with laboratory results of L. decidua and P. mugo (Hoch and Körner 2009), and partly with results of boreal tree species (Peng and Dang 2003).

Contrary to our second hypothesis, the differences in RGRsh among sites provided no evidence for distinct growth potential according to current species-specific elevational ranges, at least in the very early life of germinating seedlings. The high elevations of the seed sources (Table 1) may have partly contributed to higher than expected growth rates, based on provenance effects (e.g., Castanha et al. 2012). Additionally, species-specific differences in RGRsh may become stronger in later stages of development. Two of the species that occur at comparatively lower elevations, P. abies and P. sylvestris, did not show significant differences in biomass between the lowest and highest site, while the three species L. decidua, P. cembra, and P. mugo that occur at treeline in our study area experienced strong reductions in total biomass accumulation with increasing elevation. R:S ratios increased significantly with site elevations for all species except P. cembra and S. aucuparia (Fig. 2), indicating more biomass allocation to the root system while maintaining similar biomass investments in photosynthetic tissues as shown by the rather stable FMR among sites.

Our third hypothesis, that seedling growth and biomass should be related to light or N availability, was partly confirmed. While light availability (radC) was not related to differences in RGRsh, biomass, or biomass allocation, δ 15N was a significant predictor of biomass and biomass allocation. The significant spc:δ 15N interaction in the prediction of biomass and FMR suggested a species-specific pattern of N nutrition, especially in combination with the low overall %N and foliar N allocation in some species (Fig. 1), and the low soil N availability at our research sites (Blaser 1980; Dawes et al. 2011). Species-specific N-uptake traits (Miller and Bowman 2002) could, therefore, lead to the apparent differences in N availability and biomass allocation (Domisch et al. 2002). The elevational decline in foliar δ 15N could arise either from a decreasing soil N availability or declining N uptake, both likely influenced by temperature (Weih and Karlsson 2001; Smith et al. 2009), and N availability may have been additionally influenced by competition with other seedlings or surrounding ground vegetation (Moir et al. 1999).

At relatively low N availability, mycorrhizae-mediated N uptake becomes more likely, which would also be reflected in lower foliar δ 15N (Hobbie et al. 2000; Hobbie and Colpaert 2003). Particularly low δ 15N were found at the highest site in L. decidua, P. sylvestris and P. mugo, which in conjunction with the positive correlation of δ 15N with %N and negative correlations of δ 15N with R:S in these species (Table 3) could indicate that these species are taking up more N via mycorrhizae than the other species. Previous reports have shown the importance of mycorrhization and N uptake in tree seedlings in culture and field conditions (Hobbie and Colpaert 2003; Hasselquist et al. 2005; Gruffman et al. 2012). Furthermore, the species-specific decrease in R:S and the increase in biomass with increasing foliar δ 15N in our study also support a possible link between belowground C allocation and N availability. In line with this interpretation, fertilization experiments at treeline in the Scandes showed a nitrogen-induced growth enhancement of Betula pubescens (Sveinbjörnsson et al. 1992).

Conclusions

We acknowledge the limitations of our case study, along an elevational gradient of one treeline site, for broad generalizations. However, the responses observed here contribute to a better mechanistic understanding of control factors on early performance of germinating seedlings of a wide range of treeline species of the Swiss Alps. Our results showed that tree seedlings overall performed better than expected at and above treeline, and that differences in performance among species could not be predicted based on the species’ current elevational range. These results suggest that the early establishment of seedlings of a variety of tree species is not limited by current climatic conditions even beyond their actual upper distribution limit. We further showed that δ 15N as an indicator for N availability and uptake showed a significant and species-specific relationship with seedling biomass and biomass allocation. Nitrogen dynamics may, therefore, be an important co-driver of treeline seedling dynamics in situations where the climatic limit is not reached, and where seed input and herbivory are not limiting.