Introduction

Soil resource regime has a paramount influence on plant community productivity and structure (Grime 1977; Chapin 1980; Tilman 1982; Grime 2001). Plant communities may be assumed to be largely adapted to the current nutrient regime and therefore not to be nutrient limited per se (Körner 2003a).

However, the productivity and abundance of some species or functional groups may increase after nutrient fertilization so those parameters may be considered as nutrient limited. Production of plants in cold regions, i.e. in alpine and arctic communities, is often limited by nitrogen availability, as has been shown in a range of in situ fertilisation studies (e.g. Haag 1974; Shaver et al. 2001; Körner 2003b; van Wijk et al. 2003; Soudzilovskaia et al. 2005; LeBauer and Treseder 2008). Contrary to the well established role of nitrogen limitation, the roles of other nutrients and water limitation for productivity of arctic and alpine communities are not very clear, but can not be dismissed. There is some evidence that phosphorus can limit productivity of alpine communities (Seastedt and Vaccaro 2001), arctic tundra (Chapin 1981; Chapin and Shaver 1989), wetland communities (Gough and Hobbie 2003; Olde Venterink et al. 2003; Gusewell 2004), dry-land communities (Lambers et al. 2008), mountain grasslands on carbonate soils (Sebastia 2007), tropical rainforests (Raaimakers and Lambers 1996; Lambers et al. 2008), but was also shown to be a secondary limiting factor in a Caucasian alpine community (Soudzilovskaia et al. 2005). In some ecosystems raising pH via calcium addition lead to increasing plant biomass production due to a temporary increase of mineralization (De Graaf et al. 1998; Hobbie and Gough 2004). Soil moisture can also be an important factor limiting the production of alpine plant communities (Billings 1974; Walker et al. 1994). Nevertheless, there are few experimental investigations (e.g. Bowman et al. 1995; Soudzilovskaia et al. 2005) testing responses of alpine tundra to water availability manipulations.

Returning to the likely major role of nutrient limitation of production in terrestric ecosystems, two main types of community response to nitrogen or nitrogen-plus-phosphorus fertilizations have been reported repeatedly (Bret-Harte et al. 2008). The first response type features strong increases of graminoid biomass (e.g. Dähler 1992; Jonasson 1992; Press et al. 1998; Gerdol et al. 2000; Graglia et al. 2001; Gough et al. 2002; van Wijk et al. 2003; Bret-Harte et al. 2004; Fremstad et al. 2005; Klanderud 2008) at the apparent expense of dwarf shrubs, mosses, lichens as well as other plants typical for poor soils (e.g. Aerts and Chapin 2000; Cornelissen et al. 2001; van Wijk et al. 2003). The second response type is characterized by herbaceous species (and cryptogams) being replaced by taller woody plants (Tilman 1988; Shaver et al. 2001; van Wijk et al. 2003). However, it is not clear yet (a) which factors drive different communities to the first or second response type, and/or (b) whether alternative response types occur. Here we compare four Caucasian alpine plant communities contrasting in mesorelief position, productivity and dominance structure (Onipchenko 2004) for biomass responses to experimental fertilizations of different mineral nutrients and water. Our primary objectives were to determine: 1) which soil resources limited production of four alpine communities with different mesorelief position, productivity and dominance structure, 2) which plant functional groups react positively or negatively on different nutrient or irrigation; 3) does the response of alpine plant communities to nutrient fertilization or irrigation depend on initial dominance structure, i.e. by the relative abundance of dominant species or functional groups in control plots? To our knowledge, this is the first comprehensive, experimental study on resource limitation of productivity along a toposequence (catena) of alpine communities developing on the same geological substratum.

Methods

Study sites and communities

This study was conducted at the Teberda Biosphere Reserve (Northwestern Caucasus, Russia). The experimental site was located on the south and south–east slope of Mt. Malaya Khatipara (43°27′N, 41°41′E at 2700–2800 m a.s.l.). Mean annual temperature in the area is 1.2°C and mean July temperature is 7.9°C (Grishina et al. 1986). Annual precipitation is about 1400 mm. We investigated four alpine communities: alpine lichen heath (hereafter, simply lichen heath), Festuca varia grassland (Festuca grassland), Geranium-Hedysarum meadow (Geranium meadow) and snow bed community (snowbed). This sequence of community types reflects their location along the snow accumulation gradient from snow-free to snow-bed communities. Experimental plots of different communities were situated within a few hundred meters of one another.

Lichen heath occupies windward crests and slopes with little (up to 20–30 cm) or no snow accumulation in the winter. Deep freezing is typical for the soils there. The growing season lasts about 5 months, from May through September. Fruticose lichens are the main dominants (mostly Cetraria islandica (L.) Ach. ). There is no absolute dominant among vascular plants, but the following species contribute more than 5% to aboveground vascular plant biomass: Carex sempervirens, C. umbrosa, Anemone speciosa, Festuca ovina, Antennaria dioica, Trifolium polyphyllum, and Vaccinium vitis-idaea. Total annual primary production is about 150 g m−2 (Onipchenko 2004).

Festuca grasslands are firm-bunch grass communities, which occupy slopes with little snow accumulation (about 0.5–1 m). Snow cover stays until the second half of May or the first half of June, followed by a growing season of about 4 months. Festuca varia and Nardus stricta are the main dominants, together often contributing more than 75% of aboveground biomass. Total annual primary production is about 400 g m−2 (Onipchenko 2004).

Geranium meadow occupies the lower parts of slopes and small depressions with typical snow depths of 2–3 m. It thaws out by the end of June or by early July, followed by a growing season of only 2.5–3.5 months. Geranium gymnocaulon is the main dominant and Nardus stricta, Phleum alpinum, and Hedysarum caucasicum create more than 5% of aboveground biomass. The population density of voles [Pitymys (Microtus) majori Thomas] can reach 940 animals per hectare during a “peak-year” in these communities (Fomin et al. 1989), causing severe disturbances due to their burrowing activity. Voles prefer this community due to taller close plant cover (protecting against birds of prey) and high food plant abundance. The community is, at about 550 g m−2 year−1, the most productive within this alpine toposequence (Onipchenko 2004).

Snowbeds occupy depressions and bottoms of nival and glacial cirques with heavy winter snow accumulation (4 m or more). They have the shortest vegetative season of about 2–2.5 months from mid July until September. Short rosette and dwarf trailing forbs (Sibbaldia procumbens, Minuartia aizoides, Gnaphalium supinum, and Taraxacum stevenii) prevail here. Due to the short growth season the total production is only about 200 g m−2 year−1 (Onipchenko 2004).

Soils (Umbric Leptosols) of the studied communities have silty loam texture. The proportion of sand decreases slightly from lichen heath to snowbed. Alpine soils are relatively poor in available nitrogen and phosphorus, while they are rich in potassium due to the chemical composition of the bedrock (granite and biotitic schist) (Vertelina et al. 1996, Table 1). Soil acidity gradually increases from lichen heath to snowbed (pHH20 ranges from 5.6 in lichen heath soil to 4.7 in snowbed soil – Onipchenko 1994, see also Table 1).

Table 1 Soil acidity and available inorganic N and P (mg kg−1) after 10 years of fertilizer application (2008), n = 8, standard deviation in parentheses Different letters indicate significant difference among treatments within a community (p < 0.05, post-hoc Tukey test following one-way ANOVA)
Full size table

Field methods

The study was conducted during five growing seasons from 1999 through 2003. The experiment included controls and five treatments: Ca addition (in order to raise soil pH), P, N and NP fertilizations and irrigation. Our previous results (Onipchenko 1994) showed that potassium had no influence on plant aboveground production in potassium-rich biotite-derivated soils. Therefore, we did not use K fertilisers in this experiment.

A visually homogeneous area of 19 × 6.5 m was selected for each community and divided into 24 plots. Each plot was 1.5 m × 1.5 m with 1 m buffer zones between plots. The plots were randomly assigned to the treatments, which were replicated four times.

In 1999 Ca, N, P and NP fertilization treatments were started. N, P and NP plots were fertilized annually at the beginning of the growing season, in dry form on the soil surface. Nitrogen was added as urea (9 g N m−2 year−1), phosphorus as double super phosphate (2.5 g P m−2 year−1).

Calcium was added twice: in 1999 as lime and in 2002 as chalk (equivalent amount). The amount of added lime ranged from 52 g m−2 for lichen heath, 84 g m−2 for Festuca grassland, 119 g m−2 for Geranium meadow, to 183 g m−2 for snowbed. Different doses were used because soil acidity differs significantly among the studied communities increasing from the upper (lichen heath) to the lower (snowbed) positions along the catena. The dose of lime or chalk was to neutralize half of the potential acidity of the upper soil horizons, and had been calculated based on published soil properties of the studied communities (Grishina et al. 1993). Irrigation (H2O treatment) was conducted in 1999–2003 during the vegetation period (July–August). The mean daily value of evapotranspiration in the area is about 3 mm (Grishina et al. 1986). Every day the precipitation was measured. If the precipitation over a 3-day period did not compensate for the water loss due to evapotranspiration, the plots were irrigated with 9 mm of water. The total amount of added water varied according to natural precipitation from 0 (2002—wet season) to 45–63 mm (2000—dry season).

In each treatment and each community plot two soil cores (diameter 5.6 cm, depth 10 cm) were sampled for chemical analyses in 2008 (total 8 replications). Soil pH was measured in H2O suspension by a glass electrode (soil to H2O ratio of 1:2.5). Available phosphorus was extracted from soils with 0.5 M CH3COOH at soil to solution ratio of 1:25 and shaking time 1 h. NH +4 –N and NO −3 –N were extracted with 0.05 M K2SO4 at soil to solution ratio of 1:5 and shaking time 1 h. PO 3−4 –P, NH +4 –N and NO −3 –N were determined on a spectrophotometer Genesys-10-UV by colorimetric reactions by the molybdenum-blue, salicylate-nitroprusid and cadmium reduction methods, respectively.

Total above-ground biomass (including both lichens and vascular plants) was sampled in all plots of all treatments in summer 2002 and 2003 during the first half of August, at the peak of the growing season of this relatively late and cold summer. In each plot two 0.25 m × 0.25 m subplots were cut close to the soil surface each year. Thus, there were 16 (8 in 2002 and other 8 in 2003) subplots for each treatment within each community. Vascular plants were sorted by species. Dead leaves of the current season were added to the biomass of the corresponding species; over-wintered dead plant material of all species was considered as litter. Lichens and mosses were not sorted by species. Plant material was air-dried, then oven-dried (90°C, 12 h) and weighed. The three morphologically similar Carex species in lichen heath (Carex umbrosa, C. sempervirens and C. caryophyllea) were pooled as Carex spp.

Seven functional groups were considered: dwarfs shrubs, forbs, hemiparasites, legumes (N-fixers), grasses, sedges and other monocots. Not all of them were represented in all studied communities, so for our analysis we used only groups which had considerable biomass and/or abundance in a community (details in Table 2). We considered grasses and sedges as separate groups instead of general graminoid group due to their different biomass response to fertilization in lichen heath (Soudzilovskaia et al. 2005, 2006).

Table 2 Functional groups of vascular plants
Full size table

Data analysis

We analysed the data of total vascular plant aboveground biomass between treatments at final harvest. Data for the subsets of 4-year and 5-year harvests deviated only marginally and non-significantly (data not shown here), so they were pooled for the one way ANOVA. Assumption of normality was tested prior to all statistical tests and logarithmic transformation [log10(x + 1)] applied if necessary. We subsequently applied Tukey post hoc tests. To detect shifts in aboveground biomass composition, functional group fractions (% of aboveground vascular plant biomass) were analysed individually per functional group with one way ANOVA. In order to improve normality the data were transformed by natural logarithm [ln(x + 1)]. We did not run two-way ANOVA for simultaneous analysis of treatment and group effects, because such analysis conducted on percentage data would violate the assumption of independence among functional groups.

To compare high and low productive communities, we considered lichen heath and snowbed as the low-productive and Festuca grassland and Geranium meadow as the highly productive communities. We plotted the mean absolute aboveground biomass increase of each functional group in response to NP treatment (BNP-BContr) against the fraction of the biomass of each functional group relative to total aboveground biomass in the control treatment. Here we used absolute rather than relative biomass increase because only the former would highlight any functional group that would monopolize extra nutrients to outcompete others based on its initial strong contribution to the community. In low-productive communities without clear initial dominance by any functional group we did not expect any functional group to outcompete the others when heavily fertilised. To compare total biomass response between low and high productive communities we used relative biomass increase (BNP-BContr)*100/BContr. All calculations were made with the Statistica 6.0 software package.

Results

Soil nutrient and total biomass response to resource additions

In response to fertilizer application, soil available phosphorus increased consistently in P and NP fertilizations of all communities (Table 1). Availability of inorganic nitrogen increased less strongly, and the difference with the control was not always significant because of high variability in N and NP fertilizations. It is interesting to note that soils of the most productive community (Geranium meadow) in contrast to the other ones did not show a significant increase in N after N or NP fertilization (Table 1).

Since measurements on shoot abundances in the control plots (V.G. Onipchenko et al., unpublished data) indicated that the species abundance (and presumably biomass) structure did not change substantially between the start and end of this study in any of the four communities, we take any differences between different treatments in 2002/2003 to also represent changes relative to initial composition and biomass structure.

Total aboveground biomass significantly increased after NP fertilization in all communities, but there were some differences in response (Fig. 1): it was 156% higher in NP plots relative to control plots for lichen heath, 72% higher for Festuca grassland, 32% higher for Geranium meadow and 110% higher for snowbed. Vascular plants of lichen heath responded positively to N but more to NP treatment, while there were no differences in aboveground biomass between N and NP treatment for Festuca grassland. On the other hand, vascular plant biomass response in the most productive Geranium meadow was significant after NP fertilization only. Interestingly, snowbed showed equal biomass increases in response to NP and Ca treatments. Neither irrigation nor P fertilisation-alone affected aboveground biomass in any of the communities.

Fig. 1
figure 1

Aboveground vascular plant biomass (g m−2) response to experimental treatments for 4-5 growing seasons in four alpine communities: a alpine lichen heath (ALH), b Festuca varia grassland (FVG), c Geranium-Hedysarum meadow (GHM), d snowbed community (SBC). Different letters indicate significant difference among treatments (p < 0.05, post-hoc Turkey test following one-way ANOVA)

Full size image

The absolute increase in total aboveground biomass upon NP treatment (as compared to the control) ranged from 99.4 g m−2 in Geranium meadow to 118 in snowbed, 165 in Festuca grassland and 197 g m−2 in lichen heath. Thus on average the absolute biomass response of the two low-productive communities (157 g m−2) was not lower, but apparently even somewhat higher, than that of the two highly productive communities (132 g m−2). The productivity of snowbed may be more limited by the short vegetation season than by soil nutrients, which may have influenced our results.

Functional groups and response of dominants

Significant changes in aboveground biomass among treatments were obtained for forbs (increase in NP treatment in comparison with the control), grasses (increase in P and NP), hemiparasites (increase in irrigation and NP fertilization) and sedges (increase in N) in lichen heath, for grasses (increase in N and NP) in Festuca grassland, for forbs (increase in N and NP) and sedges (increase in N) in Geranium meadow and for forbs and grasses (increase in NP) in snowbed (Table 3). Legumes showed marginal response (P = 0.07, increase in P) in Geranium meadow. There were significant changes in functional group fractions in lichen heath (for grasses – increase in P, decrease in N, hemiparasites – increase in fertilization and NP, and sedges – increase in N) and Geranium meadow (for sedges only – decrease in Ca and P). Marginal responses (0.10 > p > 0.05) were noted in Geranium meadow (for legumes, grasses – decrease in NP, and forbs – increase in NP) and in snowbed (for grasses only – increase in NP). Other monocots (OM group) had negligible biomass in all communities and treatments.

Table 3 Aboveground biomass of plant functional groups (FG) after different treatments (g m−2, mean and standard error, n = 4) for 4 alpine communities for 4–5 years (see Table 1 for plant functional group abbreviations). Different letters indicate significant difference among treatments (p < 0.05, post-hoc Tukey test following one-way ANOVA)
Full size table

Dwarf shrubs and subshrubs (Vaccinium vitis-idaea in lichen heath, Sibbaldia procumbens in Geranium meadow and snowbed) did not react significantly to nutrient or irrigation. Forbs were abundant in all studied communities, but their response to fertilizations and irrigation varied. They did not show significant changes in Festuca grassland, increased their biomass after NP fertilization in lichen heath (2.7-fold in comparison to control) and Geranium meadow (2.1-fold). Nitrogen fertilization significantly increased forb biomass in Geranium meadow (1.5-fold), as well as its percentage in biomass. Ca addition apparently doubled forb biomass in snowbed (Table 3, P = 0.08). Hemiparasites were generally uncommon, but they increased their biomass 20-fold after NP fertilization in lichen heath. Legumes were common in lichen heath and Geranium meadow. They responded marginally (positively) to P, and (negatively) to N and NP fertilizations in Geranium meadow (Table 3). The two main graminoid groups, grasses and sedges, responded very differently to fertilizations. Grasses significantly increased their biomass after NP fertilization in all communities but Geranium meadow. They responded positively to N fertilization in Festuca grassland (mostly owing to Festuca varia) and to P fertilization in lichen heath (mostly owing to F. ovina). In lichen heath the proportion of grasses in aboveground biomass significantly increased in P, significantly decreased in N, but did not change in the NP treatment. In contrast, for the same treatment, the role of grasses marginally decreased in Geranium meadow and marginally increased in snowbed. The biomass of sedges increased most remarkably after N fertilization in lichen heath (5.7-fold) and in Geranium meadow (2.3-fold) (Table 3), also in relative terms.

Across the four communities, there was an overall positive relationship between biomass response of functional groups to NP treatment and the fraction of biomass of these functional groups in control treatments (Fig. 2).

Fig. 2
figure 2

Aboveground biomass response of functional groups to NP fertilisation treatment (4–5 year) in four alpine communities, as a function of the aboveground biomass fraction of each functional group in the control treatment for each community. ALH, Alpine lichen heath; FVG, Festuca varia grassland; GHM, Geranium-Hedysarum meadow; SBC, Snowbed community. F, forbs; G, Grasses. Linear regression relates to all functional groups present in all four communities

Full size image

Discussion

Which resources limit alpine biomass production?

Our results demonstrate that our four alpine communities differed in which resources limited productivity. For alpine lichen heath (see also Soudzilovskaia et al. 2005) and Geranium-Hedysarum meadow, in which forbs play an important role (see below), nitrogen was the principal and phosphorus the secondary limiting nutrient. Grassland dominated by Festuca varia responded equally to N alone and N combined with P fertilization, thus indicating N limitation only of F. varia in particular (see below). In the snow-bed community aboveground biomass increased equally in NP and Ca treatments, implying co-limitation of N, P and Ca. Possibly greater Ca availability or higher pH promoted nutrient mineralization and thereby reduced N and P limitation (see below). The overall productivity responses to N, P and NP fertilization did not line up with the toposequence and associated snow depth sequence, as N and P colimitation was seen at both ends of the toposequence, with the N limited Festuca varia community in the middle. Altogether, while P and Ca play important roles in specific communities, our results generally support Körner’s (2003b) hypothesis that vascular plant biomass production in alpine communities is mostly nitrogen limited. The same order of production limitation -first N, then P- was also reported for other alpine communities in different mountain areas (Jeffrey and Pigott 1973; Shatvoryan 1978; Molau and Alatalo 1998; Shaver et al. 2001; Heer and Körner 2002).

There are several ecosystems for which P limitation of biomass production has been reported, in many cases from P-depleted old soils after long-term leaching (alpine – Seastedt and Vaccaro 2001; arctic tundra – Chapin 1981; Chapin and Shaver 1989; tropical and subtropical - Walker et al. 1981; Lambers et al. 2008; arctic and temperate wetlands – Gough and Hobbie 2003; Olde Venterink et al. 2003; grasslands on basic soils – Jeffrey and Pigott 1973; Roem et al. 2002). The fact that P was not a main limiting resource in any of our communities may be explained by the relatively high P contents in the soils (1.4–3.3 mg kg−1, Table 1; see also Makarov et al. 1996, 2004), which are only of mid-Holocene age and have therefore been depleted little by substrate weathering (radiocarbon age up to 5000 years, Grishina et al. 1987).

Ca addition lead to doubling of aboveground biomass in snowbed. This unexpected result has no precedent for other alpine communities in the literature. Reduced soil acidity in response to Ca additions can improve nitrogen and phosphorus availability to vascular plants (Rorison 1980). Ca addition, at least temporally, stimulates nitrogen mineralisation (De Graaf et al. 1998; Hobbie and Gough 2004) as well as organic phosphorus mineralisation (Halstead et al. 1963; Islam and Mandal 1977; Condron and Goh 1989; Trasar-Cepeda and Carballas 1991). Indeed, in our experiment we measured an increase in nitrate and mineral phosphorus in snowbed soils, as well as a pH increase from 4.1 in controls to 5.1 after Ca addition (Table 1). Other communities showed a weaker (Festuca grassland, Geranium meadow) or no obvious pH increase (lichen heath).

Irrigation did not influence aboveground biomass in any of our four communities. Similar results were obtained for several arctic and subarctic species and communities (Wookey et al. 1994; 1995; Press et al. 1998; Robinson et al. 1998; Dormann and Woodin 2002; but see Phoenix et al. 2001) and in general for alpine plants, which appear to rarely experience water stress in mesic areas (Welker et al. 2001; Körner 2003b). On the other hand production of several species in mountain meadows in drier regions (Colorado, Alps) show significant water limitation (Bowman and Fisk 2001; Liancourt et al. 2005; Brancaleoni et al. 2007). In our communities, only hemiparasites increased biomass after irrigation treatment. Euphrasia ossica showed a 10-fold increase of population density in irrigated lichen heath plots (Soudzilovskaia and Onipchenko 2005). It seems that seedling establishment of annual hemiparasites strongly depends on soil humidity and Euphrasia is very sensitive to drought (Grubb 1984). However, such annual plants do not play an important role in total community biomass production.

Functional group response to fertilisation

Since most species occurred only in one of the four communities, we cannot be sure that their response to nutrient amendments in one community is representative for the species, as species by soil interactions for fertilization response are possible. At the level of functional types such interactions are also likely to occur, but some general response patterns can still be detected by putting our findings into the broader context based on available results from fertilisation experiments in cool-temperate, alpine and arctic ecosystems in Appendix 1. Phosphorus fertilizations lead to increase in biomass of grasses (especially Festuca ovina) in several studies (Jeffrey and Pigott 1973; Bowman et al. 1993; Soudzilovskaia et al. 2005) including ours. Festuca ovina has very low leaf P concentration in the study area (0.05%—Voronina et al. 1986; 0.022%—Soudzilovskaia et al. 2005; mean for all plant species 0.16%—Voronina et al. 1986) and compared to the global mean of 0.123% (Kattge et al. 2011) and might be limited by this nutrient. Indeed, in a monoculture experiment this species decreased soil P concentration to a greater extent than other species (Onipchenko et al. 2001). However, when P fertiliser was combined with hay cutting in subalpine meadow, the closely related species Festuca varia decreased its abundance with increasing forb biomass (Bush 1940). It is interesting to note the contrasting responses of the two congeneric grasses: production of Festuca varia was limited by N while that of F. ovina. was limited by P. These species differ greatly in traits and corresponding strategies (F. varia – competitor, F. ovina – stress tolerant – Onipchenko et al. 1998). The latter finding suggests that nutrient status and/or nutrient cycling may differ considerably even among species belonging to the same growth form (see also Bombonato et al. 2010).

Legumes, due to their nitrogen fixing ability, often respond positively to P fertilization in temperate grasslands (Rabotnov 1973) as well as in alpine plant communities (Bowman et al. 1993; Theodose and Bowman 1997; Walker et al. 2001). Our results for the Geranium meadow community support these observations. However, Trifolium polyphyllum in lichen heath did not show a significant positive response to P, likely due to the absence of nodulation and N2-fixing activity in this species (Onipchenko 1994).

Graminoids usually increase their abundance after nitrogen treatment in alpine and arctic communities (Appendix 1; McKendrick et al. 1980; Bowman et al. 1993; Press et al. 1998; Theodose and Bowman 1997; Körner 2003b; Calvo et al. 2005). Grasses and sedges can respond differently to N fertilization. Our results for lichen heath and Geranium meadow as well as studies in Colorado (Bowman et al. 1993; Theodose and Bowman 1997; Walker et al. 2001) and Alps (Bassin et al. 2007) demonstrate that sedges respond to nitrogen fertilisation better than grasses. In our monoculture experiments (Onipchenko et al. 2001) sedges (Carex umbrosa, C. sempervirens) effectively decreased available soil nitrogen concentrations. Higher C/N ratios were observed in soils under Cyperaceae (sedge allies) in alpine area (southwestern Alps) in comparison with other communities (Choler 2005). Positive responses of grasses to P fertilization and of sedges to N fertilization were noted in subalpine grasslands in the Pyrenees (Sebastia 2007). Therefore, our results support the view that sedges have a high nitrogen uptake capacity and this group can be strongly limited by low soil nitrogen availability. Thus we recommend to consider sedges and grasses as separate functional groups rather than pooling them as “graminoids”.

One of our key findings is that, while NP fertilization led to significant biomass increases in all four studied communities, different plant functional groups were responsible for this increase. Both grasses and several forbs responded strongly to NP treatment in low productive lichen heath and snowbed, without any species or group monopolizing the extra nutrients for biomass production over a five-year period. In contrast, in the two more productive alpine meadows we observed the most positive response of the main dominant group, i.e. grasses in Festuca grassland and forbs in Geranium meadow. In Geranium meadow this response resulted predominantly from biomass increase of Geranium gymnocaulon, which perhaps may have outcompeted the otherwise potentially responsive grasses there. Based on our literature survey (Appendix 1), this seems to be one of the first records of such strong forb-only biomass response to NP fertilisation in cold-biome communities. Positive responses of forbs on fertilisation were noted in several studies (Madaminov and Budtueva 1990; Henry et al. 1986; Calvo et al. 2005; Bowman et al. 1993; Jägerbrand et al. 2009 – together with temperature increase only), but we obtained an increase in the relative contribution of forbs to aboveground biomass. Otherwise, such dominance of forbs in response to fertilization can also be observed at the patch scale in the agricultural landscape, for instance patches of stinging nettles (Urtica dioica) in cattle resting areas with high dung concentration. So, in contrast to Bret-Harte et al. (2008) we propose that there are now three main types of cold biome community responses to NP fertilizations (see Appendix 1): (1) graminoid increase (Jeffrey and Pigott 1973; Bowman et al. 1993; Theodose and Bowman 1997; Gerdol et al. 2000; Bowman and Fisk 2001; Graglia et al. 2001; Walker et al. 2001; Gerdol et al. 2002; Gough and Hobbie 2003), (2) deciduous shrub increase (Shaver et al. 2001; van Wijk et al. 2003) and (3) dominant forb increase (e.g. our results for Geranium meadow). The second type was not observed in our communities owing possibly to (1) the absence of relatively tall (and deciduous) shrubs in the local species pool; (2) lack of time for shrubs to migrate from other communities. Thus, our data do not exclude the possibility of shrub expansion in response to fertilisation over longer time scales.

Conclusion

Our 4–5 year long in situ fertilisation and irrigation experiment of four alpine plant communities in the NW Caucasus has yielded three main findings which, we believe, have implications for how we think about nutrient limitation of plant communities in cold biomes: (1) We have shown that, in three out of four alpine communities, N and P fertilizations give greater aboveground biomass increases compared to fertilizations with N or P alone, indicating overall co-limitation of N and P, with N being the principal limiting nutrient. Only Festuca varia grassland responded in a way consistent with N limitation alone, while the snow-bed community also increased biomass substantially in response to Ca addition. (2) Where previous studies have generally reported strong dominance of either graminoids or shrubs upon N or NP fertilisation, we have presented the first firm example of a forb showing a similar response. (3) We have shown that the biomass response (mainly increase) of different functional groups to nutrient fertilisation is to an important degree a function of the initial composition of the community. Our findings will together help to improve our ability to predict community composition and biomass dynamics in cool- and cold-climate ecosystems subject to external nutrient inputs, as related for instance to nitrogen deposition from anthropogenic sources (Bobbink et al. 2010).