Introduction

Productivity of tundra vegetation is often severely limited by soil nitrogen (N) and/or phosphorus (P) availability (e.g. Haag 1974; Shaver and Chapin 1980, 1995). Inherent low availability of these nutrients is likely to make such systems especially sensitive to increased inputs or availability of N and P (Bobbink et al. 1998). Although numerous experiments throughout the Arctic have used high rates of fertiliser application to evaluate nutrient limitation of vegetation, these rates have far exceeded any increases in nutrient availability that may be expected in any reasonable scenario of increased deposition or climate change. To evaluate the effects of smaller, more likely changes in nutrient availability, we used both high and low levels of N and P addition in a 3-year experiment in High Arctic vegetation at Ny-Ålesund, Svalbard.

N deposition and P availability in the Arctic

Nitrogen deposition to non-forested areas in the temperate zone is typically in the range 2–5 g N m−2 a−1 (Bobbink et al. 1998). At higher latitudes, although the rate may reach 1 g N m−2 a−1 in Northern and interior Alaska, the Taimyr Peninsula and Northern Norway, N deposition is usually around an order of magnitude lower (0.1–0.5 g N m−2 a−1; Woodin 1997) than at temperate latitudes. However, as arctic ecosystems are often N-limited, with low rates of N fixation or mineralization (both up to 0.4 g N m−2 a−1; Alexander 1974; Nadelhoffer et al. 1992; Robinson et al. 1995), even small anthropogenic inputs may represent a substantial perturbation to local N cycling processes (Bobbink et al. 1998). Moreover, since global NO x emissions are increasing, NO x removal from combustion gases is technically difficult and NH3 emissions are also rising, it is likely that the Arctic will be subjected to progressively higher N inputs (Woodin 1997).

Global warming is a further component of environmental change, which impacts the Arctic with consequences for soil N availability. Increases in soil temperature in experiments at high-latitudes have been shown to stimulate N availability and mineralization strongly (reviews by Rustad et al. 2001; Robinson 2002).

It is likely that P deposition within the Arctic is extremely low, being c 0.001 g P m−2 during the growing season (Gordon et al. 2001). Phosphorus is generally tightly conserved in arctic ecosystems: soluble and exchangeable soil pools are typically very small and turn over rapidly (Nadelhoffer et al. 1992). A large proportion of soil P is often immobilized within microbial biomass and this biological pool is strongly seasonal in its P release. In dry arctic ecosystems where organic horizons are thin, such as polar semi-desert, P availability also depends on mobilization from mineral horizons (Robinson and Wookey 1997). Increases in temperature in arctic soils have been found to increase available P (+9°C, Van Cleve et al. 1983), to have no effect (+4°C, Jonasson et al. 1993) or to both increase and decrease P release (incubation temperatures of 3, 9 and 15°C, Nadelhoffer et al. 1991).

Previous studies of N deposition and greater P availability on High Arctic vegetation

Over the past three decades there has been a number of studies concerning N and P additions to arctic vegetation, often with N and P applied together, rather than factorially (Bobbink et al. 1998; Gordon et al. 2001) and at high rates or in single annual doses (Baddeley et al. 1994). Most studies have been carried out in closed vegetation communities in the low Arctic, for example in tussock tundra in subarctic Alaska (e.g. Chapin et al. 1995; Gough and Hobbie 2003; Bret-Harte et al. 2004; Hobbie et al. 2005). Fewer nutrient experiments have been carried out in the High Arctic (e.g. Henry et al. 1986; Havström et al. 1993; Robinson et al. 1998).

Polar desert and semi-desert constitute 93% of the High Arctic land area (Bliss and Matveyeva 1992). Their characteristic feature is large areas of bare ground, with disturbed soil microsites, such as frost boils and of frost-shattered rock, presenting the potential for major adjustments in vegetation cover, the establishment of “immigrant” species in uncolonized areas, and thus changes in plant diversity (Klokk and Rønning 1987; Robinson et al. 1998). In High Arctic ecosystems containing a large proportion of bare ground, it is likely that the effects of deposited N and greater P availability, if accompanied by major adjustments in vegetation cover, species richness and abundance, will influence nutrient cycles, whole system carbon budgets (Chapin et al. 2000) and surface energy and water balances (e.g. Foley et al. 1994; Beringer et al. 2001; Chapin et al. 2005). To date, there are no reports of factorial experiments concerning N deposition and greater P availability on vegetation cover and diversity in High Arctic polar desert or semi-desert.

Previous studies concerning the effects of N and P on arctic vegetation cover and diversity have frequently concentrated on the responses of vascular plants, particularly dwarf shrubs and graminoids (e.g. Haag 1974; Shaver and Chapin 1980; Fox 1992; Parsons et al. 1994; McKane et al. 2002). Bobbink et al. (1998) highlighted the need for long-term studies to confirm the sensitivity of arctic–alpine cryptogams to air-borne N pollutants and their impacts on diversity. In the High Arctic, cryptograms, particularly bryophytes, are important indicators of human disturbance (Forbes 1994). Bryophytes are often a dominant component of open tundra vegetation and increase soil moisture and reduce soil temperature because of their water-holding and insulating properties (Beringer et al. 2001; Gordon et al. 2001). Thus, any change in the abundance of bryophytes will strongly influence the abiotic conditions. In addition, efficient scavenging and retention of nutrients and slow decomposition of bryophytes all combine to affect ecosystem nutrient cycling (Longton 1997). Lichens also play a vital role in ecosystem functions, e.g. N2-fixation, nutrient cycling, or biomass contributions to the arctic C sink (Cornelissen et al. 2001). In relation to environmental change, only a few studies concern the overall biodiversity of bryophytes, lichens and vascular plants in arctic and alpine communities (Molau and Alatalo 1998).

In this context, it was useful to study the effects of a relatively realistic deposition rate of N (Bouwman et al. 2002; Tye et al. 2005) in the field, using it alone as a treatment (0.5 g N m−2 a−1) and also factorially with P, since P may limit plant response to N in this ecosystem (Wookey et al. 1995). In addition, the lowest N rate often applied in previous experiments to study impacts of N on arctic ecosystems, 5 g m−2 a−1, was used as a comparison. Effects of these factorial treatments on soil N and P dynamics have been documented in Robinson et al. (2004).

Responses of plant cover and diversity to N and P additions were quantified by extensive species assessments of mosses, lichens and vascular plants. Chlorophyll content of leaves of Polygonum viviparum was measured at the end of the season in the second year, in order to assess whether any of the treatments prolonged greenness into the autumn. The specific hypotheses tested were that the plant community in High Arctic ecosystems with a large proportion of bare ground would respond to simulated increased N deposition and/or greater P availability (1) by major adjustments in vegetation cover, (2) by colonization by ‘immigrant’ species and (3) by increases in species richness and diversity. The objectives of the study were to determine: (1) the separate and combined effects of N and P on vegetation cover and diversity, (2) whether the fertilised plots approach more closed canopy tundra and (3) whether plant species cover and diversity increased across all functional types.

Materials and methods

Site description and experimental design

An area of polar semi-desert within the Dryas octopetala zone of Svalbard (78°56′N, 11°50′E, Brattbakk 1986) was selected close to Ny-Ålesund in the High Arctic (Bliss and Matveyeva 1992). This corresponds to bioclimatic subzone B and vegetation type P1 according to the Circumpolar Arctic Vegetation Map (CAVM Team 2003). The site was located on an exposed south-facing knoll, which is often clear of snow during winter, with a 2° slope. Summer (mid-June until the end of August) precipitation totals and mean air temperatures were 65, 168 mm and 2.4, 5.1°C for 2000 and 2002, respectively (data courtesy of the Norwegian Meteorological Institute, unavailable for 2001). The ground surface was characterised by frost shattered limestone, fine till material (pH 5.6–6.1) and patches of vegetation dominated by the dwarf shrub D. octopetala. Included in on-line supplementary information are: a map showing the location of the study site (Fig. S1), a photograph of the site showing its relation to the surrounding landscape (Fig. S2) and a photograph of the plant community (Fig. S3). Organic matter (0–50 mm depth, pH 5.7–6.1) was present mainly under these vegetation mats.

A wooden fence, with widely spaced rails to stop snow build-up inside the fenced area, was erected around the site to prevent differential grazing of the treatments by reindeer (Fig. S2). Thirty 1 × 1 m2 plots were established in a randomised block design in the last week in June 2000 to form five replicates of six treatments, in a factorial combination of N (0, 0.5, 5 g N m−2 a−1) and phosphorus (0, 1 g P m−2 a−1). The treatments are subsequently referred to as control (C; 0 g N, 0 g P m−2 a−1), low nitrogen (LN; 0.5 g N, 0 g P m−2 a−1), high nitrogen (HN; 5 g N, 0 g P m−2 a−1), control with phosphorus (CP; 0 g N, 1 g P m−2 a−1), low nitrogen with phosphorus (LNP; 0.5 g N, 1 g P m−2 a−1) and high nitrogen with phosphorus (HNP; 5 g N, 1 g P m−2 a−1). N and P were applied in solution as NH4NO3 and NaH2PO4. Solutes were added five times during each of the growing seasons, starting in June 2000, to provide the total annual nutrient addition. Since the solutes were added in small water volumes, equivalent amounts of water minus solutes were added to the C plots. The treatments were applied for each of three consecutive field seasons.

Vegetation analysis

Nomenclature follows Rønning (1996) for vascular plants, Corley et al. (1981) and Corley and Crundwell (1991) for bryophytes and Santesson (1993) for lichens. Each summer at peak plant biomass, determined by plant phenology, three methods of vegetation analysis were used to determine vegetation composition and cover as fully as possible. These were: total plot counts of vascular plant species, point frame analysis and estimation of the percentage cover of ‘new’ bryophyte species.

The number or percentage cover of the vascular plants in each plot was quantified using a 1 × 1 m2 quadrat divided into 10 × 10 cm2. Within each square, the clonal plants D. octopetala and Saxifraga oppositifolia were enumerated by percentage living cover. For the clonal plant Salix polaris, the number of leaves was counted, and the numbers of tillers were counted for graminoids (after Van der Wal et al. 2000). Non-clonal plants were enumerated by counting numbers of individuals.

Point frame analysis was performed in a central, permanent non-destructive area (50 × 50 cm2) within each plot, using a point frame with a hundred intersections, 5 cm apart, to determine changes in the plant species’ cover with treatment. Bryophytes, crustose and foliose lichens, bare ground and litter were included in the analysis. Bare ground was sub-classified into stones, mineral soil or humus (fine organic material); the litter layer was comprised mainly of organic matter originating from D. octopetala. In addition, vascular plants and cryptogams were categorised as living or dead. Point frame analysis was carried out in four, rather than five, replicates of each treatment, as not all plots contained sufficient material for a central, non-destructive area to be preserved.

The term ‘new’ bryophytes refers to those species of moss not present at the first sampling but which appeared in the next 2 years. Percentage cover of ‘new’ bryophytes, which colonised the bare mineral soil in each plot was determined, using the same quadrat as employed for enumerating the vascular plants.

Chlorophyll content of Polygonum viviparum leaves

At the end of the growing season in the second year (16 August 2001), measurements of total chlorophyll content were made non-destructively on the least senescent leaf (youngest mature leaf) of P. viviparum plants, using a Minolta SPAD meter. This plant species was chosen because previously, by visual assessment only, a higher proportion of living versus senescent P. viviparum leaves was found where NPK was applied for 4 years compared with the proportion for plants in unfertilised plots (C. H. Robinson and O. B. Borisova, unpublished data, 19 August 1994). Ten plants were used per plot. To calibrate the SPAD for P. viviparum, SPAD readings were taken from the least senescent leaf of each of five plants outside the experimental area, with the leaves subsequently detached and taken back to the field laboratory. Each leaf was ground, extracted in 80% acetone and the total chlorophyll content estimated using the procedure of Lichtenthaler (1987).

Statistical analysis

All data sets were checked for normality using the Anderson–Darling test (Anderson and Darling 1954). Appropriate transformations of arcsine, log and square-root were performed before repeated measures factorial analysis of variance was carried out between treatments and years, followed by F-protected Least Significant Difference (LSD) tests.

Diversity of the combined vascular plant, moss and lichen community in each treatment in each year was estimated by calculating the Shannon index, using the results from the point frame analysis. The Shannon index is H′ = −Σ p i log p i , where p i is the proportion of the total number of ‘hits’ composed of species i. Values usually fall between 1.5 and 3.5, and are rarely greater than 4.5 (Calow 1999).

Results

Total living and dead vegetation cover and bare ground

There was a significant effect of year of treatment (P < 0.001) on total living vegetation cover (included vascular plants, mosses, lichens and Nostoc, Fig. 1a), with a trend of greater live cover in HNP in 2001 and 2002 compared with 2000 (year × N interaction, P < 0.001; year × P interaction, P < 0.001; ANOVA). Similarly, for total dead vegetation (Fig. 1b), there was a significant effect of year (P < 0.01; ANOVA) and a year × P interaction (P < 0.01; ANOVA). For total bare soil (data not shown, includes bare organic plus bare mineral soil) and bare mineral soil (Fig. 1c), there was a significant year × N × P interaction (P < 0.05; ANOVA), resulting in a trend of more bare mineral soil in the control, CP, LN and HN plots compared with the LNP and HNP treatments (Fig. 1c).

Fig. 1
figure 1

Mean (±SE) a total living vegetation cover, significant effects: year of treatment (P < 0.001), year × N interaction (P < 0.01), year × P interaction (P < 0.01); b Total dead vegetation cover, significant effects: year of treatment (P < 0.01), year × P interaction (P < 0.01); c Percentage bare mineral soil, significant effects: year × N × P interaction (P < 0.05). Data obtained using 50 × 50 cm2 point quadrat, n = 4. Statistical results refer to factorial repeated measures ANOVA. The treatments are control (C; 0 g N, 0 g P m−2 a−1), low nitrogen (LN; 0.5 g N, 0 g P m−2 a−1), high nitrogen (HN; 5 g N, 0 g P m−2 a−1), control with phosphorus (CP; 0 g N, 1 g P m−2 a−1), low nitrogen with phosphorus (LNP; 0.5 g N, 1 g P m−2 a−1), and high nitrogen with phosphorus (HNP; 5 g N, 1 g P m−2 a−1)

Full size image

Cover of established vascular plants and individuals of ‘new’ species

There was a significant effect of treatment year on individual plant species (Figs. 2b–d, 3a, b inclusive). Except for percentage living cover of D. octopetala (Fig. 2a), factorial additions of N and P had clear effects on cover of individual plant species from the second year of treatment onwards, often manifested statistically by interactive effects of N and/or P with year. The interaction results were initially investigated further using the LSD statistics in an attempt to explore plant community development. These more detailed results, however, are largely not presented here because they simply confirm the means and standard errors in Figs. 2a–d and 3a, b inclusive.

Fig. 2
figure 2

Mean (±SE) percentage cover of individual vascular plant species a Dryas octopetala, no significant effects; b Saxifraga oppositifolia, significant effects: N (P < 0.05), P (P < 0.05), year of treatment (P < 0.001), year × N interaction (P < 0.001), year × P interaction (P < 0.05), year × N × P interaction (P < 0.05); c number of live leaves of Salix polaris m−2, significant effects: year of treatment (P < 0.05), year × N interaction (P < 0.05), year × P interaction (P < 0.01); d Number of shoots of Polygonum viviparum m−2, significant effects: year of treatment (P < 0.001), year × P interaction (P < 0.01). Data obtained using 1 × 1 m2 quadrat, n = 5. Statistical analysis and key to symbols as for Fig. 1

Full size image

Each species showed individualistic responses to N and P applications. Percentage cover of S. oppositifolia was significantly greater with N additions (P < 0.05; ANOVA; HN > LN, P < 0.05, LSD) and P additions (P < 0.05; ANOVA), although there was no significant N × P interaction (Fig. 2b). There were several significant interactions of N and P with year (Fig. 2b). In 2001 and 2002, there were significantly greater densities of live leaves of S. polaris with N additions (year × N interaction, P < 0.05; LSD) and P alone (year × P interaction, P < 0.01; LSD; Fig. 2c). The densities of P. viviparum shoots were significantly greater in 2001 and 2002 where P was applied (P < 0.01; year × P interaction; LSD) (Fig. 2d). There was no obvious treatment effect on the tiller density of Carex rupestris, C. misandra and C. nardina because the distribution of the three species across the experimental site was so uneven (data not shown).

An extremely rapid response in the number of Draba oxycarpa seedlings was observed (Fig. 3a). Even within the first year of treatment, there were greater densities of shoots in treatments where P was applied (P, P < 0.01; year, P < 0.01; year × P interaction, P < 0.01; ANOVA; Fig. 3a). By the third year of the experiment (2002), however, there was a trend of greater densities of shoots in plots where N and P had been added together, as well as where P was applied alone (Fig. 3a). Taking the seedlings of all ‘new’ species into account, the greatest densities were present in the HNP treatment, e.g. Saxifraga caespitosa, Sagina nivalis and D. oxycarpa (Table 1, Fig. 3a). There was no significant treatment effect within each year on the number of seedlings of S. oppositifolia m−2 (Fig. 3b), although there was a trend of fewer seedlings in the control, CP and HN treatments compared with HNP in each of the 3 years.

Fig. 3
figure 3

Mean (±SE) number of a shoots of Draba oxycarpa m−2, significant effects: P (P < 0.01), year of treatment (P < 0.01), year × P interaction (P < 0.01); b Seedlings of Saxifraga oppositifolia m−2, significant effect: year of treatment (P < 0.001). Data obtained using 1 × 1 m2 quadrat, n = 5. Statistical analysis and key to symbols as for Fig. 1

Full size image
Table 1 Total number per treatment (total area 6 m2) per year of seedlings of ‘new’ angiosperm species
Full size table

Cover of bryophytes and lichens

Within 2 years of treatment, where N and P were applied together, there was a dramatic reduction in cover of the previously dominant moss species, Schistidium apocarpum (Fig. 4a). Concomitantly, there was a marked increase in the proportion of Bryum arcticum, a species unrecorded on the plots prior to 2001, colonising what was previously bare soil. In 2001 and 2002 (second and third years of experiment), there was a significantly higher proportion of B. arcticum: bare ground where N and P had been applied (Fig. 4b), especially together (N × P interaction, N alone, P alone, P < 0.001; ANOVA). Other moss species were recorded: Andreaea obovata, Aulacomnium turgidum, Dicranoweisia crispula, Dicranum sp., Drepanocladus vevcricosus, Hylocomium splendens, Hypnum hamulosum, Isopterygium pulchellum, Oncophorous virens and Tortula ruralis. However, these were occasional (<1% cover) and were unevenly distributed across the field site. There were no significant effects of the treatment on the most abundant crustose lichen on stones, Protoblastenia incrustans (data not shown).

Fig. 4
figure 4

Mean (±SE) cover a Schistidium apocarpum, n = 4, significant effect: year of treatment (P < 0.001); b Bryum arcticum: bare ground ratio, n = 5, significant effects: N (P < 0.001), P (P < 0.001), N × P interaction (P < 0.001). Data obtained using 50 × 50 cm2 point quadrat. Statistical analysis as for Fig. 1

Full size image

Species richness and diversity for vascular plants, bryophytes and lichens combined

There was no significant effect of treatment or year on species richness (Fig. 5a), although there was a trend of an increase in richness where P was applied at 0 and 0.5 g N m−2 a−1, but not at 5 g N m−2 a−1. Species diversity was significantly greater in plots where P had been applied (P < 0.05; ANOVA; Fig. 5b). There was the same trend of no increase in diversity at 5 g N m−2 a−1 with P application as for richness above. No treatment effects on the diversity of crustose lichens on stones were observed (data not shown).

Fig. 5
figure 5

Mean (±SE) vascular plant and cryptogam a species richness, no significant effects; b Diversity, significant effect: P (P < 0.05), n = 4. Statistical analysis and key to symbols as for Fig. 1

Full size image

Chlorophyll content of Polygonum viviparum leaves

At the end of the growing season in the second year of treatment, there were significantly greater total chlorophyll contents in the youngest mature leaves where N had been applied (P < 0.001; ANOVA). The higher rate of N had the most effect, although there was evidence of greater chlorophyll content in the LN treatment than in the control (HN > C, CP, LNP P < 0.001; LSD; Fig. 6).

Fig. 6
figure 6

Mean (±SE) chlorophyll content of the least senescent leaf of Polygonum viviparum, 16 August 2001, n = 5. Key to symbol: ***P < 0.001, ANOVA, significant effect of N

Full size image

Discussion

Previously, in close proximity to the current study site, a five-year experiment was carried out in polar semi-desert where N, P and potassium (K) were added together at 5, 5 and 6 g m−2 a−1, resulting in an increase in moss cover with a concomitant decrease in bare ground in fertilised plots (Robinson et al. 1998). Fertilisation also promoted establishment of seedlings of nitrophilous ‘immigrant’ species, whereas the live cover of dominant vascular plants D. octopetala and S. oppositifolia significantly decreased. After 5 years, the plant community tended more towards closed bird-cliff vegetation. However, the relationships between different N application rates and interactions with P could not be defined in this previous experiment.

The background N deposition rate at the Ny-Ålesund site is ∼0.1 g N m−2 a−1 (Woodin 1997), meaning that the LN treatment in the current experiment was relatively realistic (five times ambient deposition, but within the ranges already found in Greenland and Iceland, and below that found in northern Norway, the Taimyr Peninusula and Alaska; Woodin 1997). The HN treatment can be compared directly with previous experiments studying the impacts of N on Arctic ecosystems (see Gordon et al. 2001 for a review) but it is 50 times the rate of ambient deposition (Woodin 1997) and 25 times that of N mineralization (Robinson et al. 1995) at this High Arctic site. This 25-fold increase in annual plant-available N supply is much greater than the largest reported increase in N mineralization rate in response to experimental soil warming in Arctic ecosystems, and is much closer to the N deposition rate in non-forested areas of Europe (Bobbink et al. 1998). Clearly, this rate of N addition was designed to be supra-optimal in our experiment. At the Ny-Ålesund site over the summer season, inorganic wet N deposition was largely in the form of nitrate (82% of the total) compared with ammonium (18%; N. J. Madan, unpublished data), suggesting that the application of NH4NO3 to simulate N deposition in our experiment was not unreasonable. Even so, dissolved organic N and particulate N may account for at least 2–18% and 6–44%, respectively, of total deposited N at the site (Hodson et al. 2005). Phosphorus was added to ameliorate any limitation of the N response because of soil P deficiency. The P application was double that used at a mixed tundra heath in the inner fjord zone of Ny-Ålesund (Gordon et al. 2001), but five to tenfold less than used in other Arctic studies (e.g. Chapin et al. 1995; Press et al. 1998; Robinson et al. 1998).

Vegetation responses

Although there were no significant effects of the treatments on total living or dead vegetation cover in each of the 3 years, clear responses to N and P additions were observed in individual plant species. As early as 1987, Klokk and Rønning noted that the response of High Arctic plants to fertiliser was species-specific.

Vascular plants

Because of the paucity of research concerning nutrient applications to polar semi-desert ecosystems, it seemed sensible to compare the responses of individual plant species in the current study where N and P were applied factorially (rates of 0, 0.5, or 5 g N m−2 with P at 0 or 1 g P m−2) for 3 years with those at a site in close proximity where NPK were applied together (rates of 5, 5 and 6.3 g m−2 for N, P and K, respectively) for up to 5 years (e.g. Wookey et al. 1994, 1995; Robinson et al. 1998).

Salix polaris and P. viviparum exhibited similar responses in the current study to those in the previous one where N, P and K were added in combined applications. In the current factorial study, the density of S. polaris leaves was significantly greater with N and P additions, compared with a significant increase in cover with NPK additions after 5 years in the previous experiment (Robinson et al. 1998). In the current study, greater shoot densities of P. viviparum were recorded where P had been applied, which may explain the observations of Wookey et al. (1994) that, after 3 years of treatment, combined NPK additions were associated with a 171% increase in whole-plant biomass per unit ground area compared with unfertilised controls.

There were different responses for D. octopetala and S. oppositifolia in the current study compared with those in the previous combined nutrient experiment. After three growing seasons, D. octopetala showed no response to any treatment, although previously there was an increase in plant cover with combined NPK additions. Subsequently, in the latter experiment winter injury induced by the NPK treatment was particularly detrimental to live cover of this species (Robinson et al. 1998). The damaging effects of fertiliser on Dryas species have also previously been reported by Henry et al. (1986) with 25 g m−2 NPK and by Fox (1992) with 10 g N m−2 combined with P and K. In the current study, S. oppositifolia cover increased with N and P separately, even though a significant decrease in cover was previously observed after five growing seasons with combined applications of NPK (Robinson et al. 1998). Reasons for apparent inconsistencies in species’ responses observed in the current and previous experiments may include different rates and frequencies of N and P (and K) application, interaction with climate (e.g. winter injury) and differing duration of field experiments.

The colonisation of the plots by seedlings of previously unrecorded angiosperm species where P and N were applied (Fig. 3a, Table 1) is in agreement with the results of Robinson et al. (1998), although slightly different, less nitrophilous species, were recorded in the current study. For example, Oxyria digyna, Cochlearia officinalis, Saxifraga cernua and Phippsia algida were not recorded as previously, which could reflect 3 years of treatment only, rather than five, or the characteristics of the local seed bank (Cooper et al. 2004), or response to K. Saxifraga caespitosa was a notable ‘new’ species in the HNP plots in the current study (Table 1). This species was found to be present at or below 5% cover on the Lovén islands, but absent on the glacier foreland of the Midtre Lovénbre (Hodkinson et al. 2003), presumably because of higher nutrient availability from bird manuring on the former site. Indeed, S. caespitosa is diagnostic of moist bird-cliff meadows (Eurola and Hakala 1977; Hadač 1989; cited in Elvebakk 1994), and the largest density of seedlings (4,366 m−2) of this species, out of six dry-mesic habitats on Svalbard, was found in bird cliff vegetation on the NW coast (Cooper et al. 2004). The high densities of seedlings of S. oppositifolia (greater than 20 m−2 in the HNP plots) are consistent with very high densities of seedlings of this species at a gravel pad, a raised beach ridge and a mesic heath at Alexandra Fjord (78°53′N, 75°55′W; Freedman et al. 1982). Such seedlings are likely to originate from long-term seed banks (Cooper et al. 2004). The germination of species (e.g. S. cernua, Cerastium alpinum) from propagule banks associated with nutrient-rich substrates at High Arctic sites has also been discussed by Forbes (1996). Even though they vary in number and functional traits to a great extent, these seedlings provide useful information concerning the pool of potential colonizing species in response to disturbance, or environmental change.

Bryophytes and lichens

Here, individual bryophyte species displayed contrasting responses to fertilisation, suggesting that they should not be grouped as a single functional type (Potter et al. 1995; Gordon et al. 2001). Cover of S. apocarpum was markedly decreased by a combination of HN and P applications, in agreement with the results of Gordon et al. (2001) for the same species in mixed tundra heath. The dramatic increase in B. arcticum in the LNP and HNP treatments (Fig. 4b) corresponds well with the increase in cover of bryophytes, mainly as Bryum caespiticium and Tetraplodon mniodes, after 5 years of combined NPK treatments (Robinson et al. 1998). Similarly, close to Ny-Ålesund, Bryum spp. colonised sand-gravel marine terraces fertilised with a combined application of 5 g N m−2 a−1 and 2.2 g P m−2 a−1 (Klokk and Rønning 1987). This increased colonisation by some moss species in open semi-desert sites is not observed in closed tundra vegetation, where increased shrub and graminoid growth in response to fertilization results in most bryophytes being shaded out (Chapin et al. 1995; Press et al. 1998; van der Wal et al. 2005). For example, in a subarctic dwarf shrub heath, combined NPK applications over 5 years at 10, 10 and 12.6 g m−2 a−1 resulted in significantly lower abundance of three of the dominant mosses (H. splendens, Polytrichum juniperum and Dicranum scorparium) compared with control plots (Press et al. 1998).

In northern and central Europe, the moss S. apocarpum generally indicates basic or calcareous conditions according to several authors (e.g. Smith 1978; Jahns 1980). B. arcticum is described as occurring on ‘basic soils’ in Europe (Smith 1978) and on ‘moist calcareous soils in Alaska’ (Steere 1978) and the Canadian High Arctic (Schuster et al. 1959). The clear reduction of the previously dominant S. apocarpum and the concomitant increase in B. arcticum raises the question of whether there has been a shift in the substrate nutrition or chemical status, particularly in the HNP treatment. There was no apparent soil acidification, however, in the second year as a result of the treatments (Robinson et al. 2004), although any decrease in pH may have occurred only in the very uppermost millimetres of the soil. In the third year of treatment, addition of 0.5 g N m−2 a−1 appeared to exceed N, particularly nitrate-N, demand in the soil even when P limitation was relieved (e.g. HNP plots where cover of B. arcticum was dramatically increased). This suggests further accumulation of nitrate-N and soil acidification is likely with further feedbacks to changes in vegetation growth.

There was no effect of any treatment on cover of P. incrustans here, in agreement with results from our previous study (Robinson et al. 1998). However, in a High Arctic mixed tundra heath, 5 g N m−2 a−1 reduced lichen abundance by almost half, and 1 or 5 g N m−2 a−1 with P exacerbated this negative response, reducing cover by 60% when compared with control plots (Gordon et al. 2001).

Species richness and diversity for vascular plants, bryophytes and lichens combined

In open tundra, increases in vascular plant diversity with NPK application, comparable to the changes observed here (Figs. 5a, b), have been reported (Fox 1992). Greater plant species richness has also been recorded in plots receiving 1 g N m−2 a−1 and 0.5 g P m−2 a−1 compared with unfertilised ones (Gordon et al. 2001). In contrast, in closed tundra in the Low Arctic, fewer species are usually found in plots receiving combined NPK applications compared with plots that remain unfertilised (e.g. Chapin et al. 1995; Press et al. 1998). This reflects increased dominance of a few rapidly growing species (e.g. Betula nana, Calamagrostis lapponica) and competitive elimination of understorey taxa.

Chlorophyll content of Polygonum viviparum leaves

There was an indication, even at the lower rate, that N application could prolong the greenness of leaves of P. viviparum into the autumn (Fig. 6). A similar phenomenon has been observed previously for this species by visual assessment, when a higher proportion of living versus senescent P. viviparum leaves was found where NPK was applied for 4 years compared with the proportion for plants in unfertilised plots (C. H. Robinson and O. B. Borisova, unpublished data, 19 August 1994). Such extensions of the growing season as a result of increased N deposition are likely to have severe effects on the survival of some plant species. For example in our previous experiment, D. octopetala in fertilized plots remained green into the autumn of 1993 (after 3 years of treatment), and winter injury was observed in these plants in spring 1994 after an autumn of exceptionally wet and mild conditions, suggesting that winter injury resulted from delayed hardening (Robinson et al. 1998).

N and P availability

In the current study, the greatest effects on vascular plant and bryophyte cover and on germination of ‘new’ seedlings occurred when N was applied with P. Thus, the effect of increased N deposition (or increased N mineralization in response to soil warming) is likely to depend on soil P availability (and its response to warming) at the micro-scale. Indeed, increased P availability alone resulted in greater species diversity and richness and density of P. viviparum shoots.

Conclusions

This study has shown that if soil N and P availabilities increase sufficiently at the soil surface in polar semi-desert sites, areas of bare ground could be colonised by B. arcticum, by increased cover of S. oppositifolia and S. polaris and by ‘immigrant’ angiosperm seedlings. The experiment was relatively short in duration and it would be expected that much greater changes would be observed in the longer term, especially using the low-level additions. After 3 years, it is clear that the changes in vegetation resulting from the high-rate could be seen beginning at the low rate. Although the greatest responses were observed here at ‘unrealistic’ rates of N and P addition, even so, there is the possibility that the vegetation-growing season could be extended at tenfold lower, more realistic, rates of N deposition. Further work should elucidate the effect of changes in vegetation cover on ecosystem function (e.g. on nutrient cycles, water budgets) and on feedbacks to climate change.