In general, our results are in agreement with the conceptual model describing the influence of N deposition on Sphagnum dominated vegetation proposed by Lamers et al. (2000) and Berendse et al. (2001). As was hypothesised, the concentration of inorganic N in the interstitial water decreased and increased in accordance with the changes in N deposition (Fig. 1a,b). Adding N did encourage vascular plants to grow, although an increase in the deposition above field background values, benefited only few vascular plant species (Table 2). Rhynchospora and Vaccinium increased their biomass when N deposition increased from 40 to 80 kg N ha−1 yr−1. Other species accumulated the excess N in their tissue (Table 3). This observation, coupled with the high N : P ratio in most vascular plant species and Sphagnum (Tables 3 and 4), suggests that those species that did not expand their biomass after adding N were limited by P (Verhoeven & Schmitz, 1991; Koerselman & Meuleman, 1996).
Both Lamers et al. (2000) and Berendse et al. (2001) suggest that in intact vegetation, the depression of Sphagnum by elevated N deposition is a result of increased shading by vascular plants. In our study, the observed growth reduction of Sphagnum in the 80 kg N treatment (Fig. 4) cannot be wholly explained by increased shading (Fig. 5): only less than half the data points indicating a combination of heavy shading and short height increment coincide with the 80 kg N treatment (Fig. 5). An alternative explanation, which may have confounded the effect of shading, is a direct toxic effect of N on Sphagnum at a high N supply (Limpens & Berendse, 2003; Press et al., 1986; Gunnarsson & Rydin, 2000). The hump-shaped relationship we found between shading and height increment (Fig. 5) does point to an additional effect of shading, however. It seems that the observed response to shading below 53% was due to elongation of Sphagnum as a result of diminished light availability. Above 53% shading however, length increment decreased albeit the etiolation process taking place. This result suggests a decreased production, as has been observed by Hayward & Clymo, 1983).
There are some further discrepancies between our study and the conceptual model proposed by Lamers et al. (2000). In our study (Table 4), as well as in other similar studies (Ferguson et al., 1984; Press et al., 1986; Aerts et al., 1992; Pitcairn et al., 1995; Williams et al., 1999; Nordin & Gunnarsson, 2000; Berendse et al., 2001; Heijmans et al., 2001), tissue N concentration of Sphagnum showed a linear increase with N deposition rather than a logarithmic one, and subsequently reached higher values than the proposed maximum N tissue concentration of 12–13 mg Ng−1. For S. magellanicum the maximum value recorded was 24.3 (± 0.06) mg Ng−1 in the 0–3 cm fraction (Heijmans et al., 2001). This value was reached in a fertilisation treatment in which the total deposition load was c. 100 kg N ha−1 yr−1. When these concentrations are taken into account, a maximum organic N concentration of c. 20 mg N g−1 for S. magellanicum, as proposed by Berendse et al. (2001), seems more likely. Of course, one can argue that the cited studies in which nutrient concentration exceeded 12–13 mg N g−1 refer to relatively short-term fertilisation experiments, which lasted 1–3 yr. Experimentally fertilised systems undergo forced rapid change, whereas peat bogs have been subject to at least 10 yr of rather constant, or slowly changing, deposition loads. Lamers et al. (2000) only used values for Sphagnum derived from these latter ‘natural’ conditions. Nevertheless, there is no reason to assume that Sphagnum would behave so differently under experimentally fertilised conditions.
Another anomaly is that, according to the theory as proposed by Lamers et al. (2000), N concentration in the rhizosphere should only begin to increase strongly when the N tissue concentration in Sphagnum has peaked. Our results indicate that the process must be more gradual. While still accumulating N in its tissue, S. magellanicum is not capable of retaining all atmospherically derived N, as indicated by the increasing inorganic N concentration in the interstitial water (Fig. 1a,b) and the increased concentration of N in vascular plants (Table 3). It seems more likely that, as tissue N concentration of Sphagnum increases, its capability to absorb N declines, as shown by Woodin & Lee (1987). The rate, at which Sphagnum and subsequently the rhizosphere become loaded with N, is likely to be determined by Sphagnum production. In turn, the latter may be influenced by factors such as P availability, degree of shading, water table 1evel and the occurrence of extreme climatic conditions (Hayward & Clymo, 1983; Malmer, 1988; Aerts et al., 1992, 2001; Takagi et al., 1999; Heijmans et al., 2001). When modelling the impact of N deposition on Sphagnum dominated peatlands, we must take this interaction between the abiotic and biotic environment into account.
The above implies that we cannot separate phases two and three of the proposed mechanism on account of the maximum N concentration in S. magellanicum: the Sphagnum filter fails before Sphagnum reaches its maximum inorganic N content. As the transition between phase two and three is not well defined, phase two only seems to serve a theoretical purpose, maybe indicating a time lag between exceeding phase one (N no longer limits Sphagnum growth) and the ensuing N effects on the vegetation due to positive feed-back through litter quality (Berendse et al., 1989, van Breemen, 1995). On account of this, we expect N-induced changes in the species composition of Sphagnum dominated peatlands at the end of phase one, shortly after the N concentration of Sphagnum in the upper 3 cm surpasses 8–9 mg N g−1 (Lamers et al., 2000).
We anticipated that growth of Betula and Molinia would improve with an increase in N deposition and decline when deposition approached zero. Although the growth of both species was much less than expected, their response to N was in accordance with our hypothesis (Table 1, Figs 2 and 3). Nevertheless, it seems that the expansion of these species in bogs cannot be solely explained by high atmospheric N deposition. The level of the water table in combination with the availability of other nutrients than N, probably codetermines the success of these species in undrained bogs.
For Betula the circumstances were obviously too severe for establishment; the combination of a high water table and substantial Sphagnum growth successfully prevented survival of Betula seedlings. A less vigorous Sphagnum growth seems a prerequisite to a successful establishment.
Molinia also suffered indirectly from the high water table by becoming vulnerable to competition with Rhynchospora . In the first year Molinia profited from adding N ( Fig. 2 ) and took up a significant part of the N available in the interstitial water ( Fig. 1b ). At the beginning of the following year, the amount of N taken up in the previous year was used to form new leaves ( Thornton & Millard, 1993 ). Between May and June 2000, the presence of Molinia no longer suppressed N availability in the interstitial water ( Fig. 1b ). This period coincided with the rapid expansion of Rhynchospora in the mesocosms treated with extra N. At the end of the growing season, no effect of N treatment was found on biomass of Molinia , but adding N did increase Rhynchospora biomass ( Table 2 ). The extremely high N : P ratio of Molinia ( Table 3 ) indicates limitation by P ( Verhoeven & Schmitz, 1991 ; Koerselman & Meuleman, 1996 ). On this account it seems likely that during the second growing season, insufficient P was available to sustain further growth of both Molinia and Rhynchospora at the same time. Although Rhynchospora is not characterised as a successful competitor ( Ohlson & Malmer, 1990 ), under the conditions in our experiment it was apparently able to outcompete Molinia . That a low P availability, at least in natural wet ecosystems, can curb expansion of Molinia , has also been shown by Aerts & Berendse (1988 ). In their fertilisation experiment, conducted in wet heath, they only found a significant increase in cover for Molinia when P had been added.
From the above, we infer that extensive expansion of Molinia in intact bog vegetation is likely to occur only when P availability is also enhanced.
Raised bog vegetation seems able to recover its nutrient poor state after the source of enrichment is taken away. The latter is illustrated by the decrease in inorganic N concentration of interstitial water (Fig. 1a,b) and the N tissue concentrations of both vascular plants and Sphagnum (Tables 3 and 4) in the 0 kg N treatment. In addition, the poor performance of Betula and Molinia in this treatment (Table 1, Fig. 2) shows that when N availability decreases, survival of invasive species becomes challenged. These findings, combined with the vigorous Sphagnum growth (6 cm in 2 yr, Fig. 4), suggest that in time vascular plant establishment and growth would slowly decrease and again be in line with the re-created extreme nutrient-poor environment. These observations are supported by the work of Maksimova & Yudina (1999), who mention a similar reversibility of fertilisation effects. They observed that the vegetation had almost completely recovered 18 yr after a 3-yr application of high doses of mineral fertiliser (30 kg N and 60 kg P ha−1 yr−1). These findings imply that prospects for the conservation of bogs are good, if critical loads are met in the near future.