Control of environmental variables on species density in fens and meadows: importance of direct effects and effects through community biomass


  • 1

    Present addresses: IHE, Wetland Ecosystems, PO Box 3015, 2601 DA Delft, the Netherlands and 2Alterra, Ecotoxicology, PO Box 47, 6700 AA Wageningen, the Netherlands.

H. Olde Venterink (fax + 31–15–2122921; e-mail


  • 1We tested whether patterns of species density are controlled not only by variations in community biomass but also by variations in environmental conditions, which may be related or unrelated to community biomass. Environmental variables (soil characteristics, acidity, water regime, nutrient availabilities) were measured in 46 sites in wet meadows and fens, and related to above-ground biomass and to densities of all vascular plants and of threatened species.
  • 2Both meadows and fens showed a hump-backed species density–biomass relationship, although there was much variability and the study did not include very highly productive sites. In fens a significant quadratic relationship was observed (R2 = 0.42).
  • 3Environmental factors could explain 57% (in meadows) and 40% (in fens) of variation in species density. The majority of the variance explained was attributable to factors that were not related to community biomass (increasing pH in meadows) or the organic soil–mineral soil gradient in fens.
  • 4Density of threatened species vs. biomass was also consistent with a hump-back curve covering a narrow biomass range. Density of threatened species increased with decreasing P availability, regardless of whether P availability was related to biomass (as in meadows) or not (fens).


Explaining species density is of widespread interest among ecologists, especially because of concern about the world-wide loss of biodiversity. The relationship between species density and community biomass has been studied extensively (see Grace 1999; Waide et al. 1999; for recent reviews) with the conclusion that a hump-back relationship between species density and biomass (cf. Grime 1979) may be general, with this curve encompassing all possible combinations of species density and biomass. This means that biomass alone could explain (1) maximum species density and (2) absence of high species density at very low and very high biomass values. However, several studies revealed that the hump-back curve was only observed when a broad range of vegetation types was compared, and biomass was a poor predictor of species density within a more limited range (Moore & Keddy 1989; Wheeler & Shaw 1991; Guo & Berry 1998).

A conceptual model has been developed from Grime's proposals (cf. Gough et al. 1994; Grace 1999) in which species density is controlled by environmental conditions indirectly via community biomass which subsequently affects species density, and directly by determining the number of species physiologically capable of living at a site. Application of this model in coastal wetlands showed that environmental conditions per se, exerting control on the species pool, were equally important or better predictors of species density than community biomass. In particular, salinity and flooding could explain a large part of variation in species density (Gough et al. 1994; Grace & Pugesek 1997; Grace & Jutila 1999).

We applied the conceptual model of Gough and Grace to inland freshwater wetlands (wet meadows and fens), where there is less flooding stress, to determine whether direct environmental influence on species density is also important in these systems and, if so, which environmental conditions are in control. We focus on the role of nutrient availability, as this may influence species density directly as well as via community biomass.

A second aim was to investigate whether the occurrence of threatened species was controlled by the same mechanisms as overall species density. Although the loss of species in European wetlands is often associated with eutrophication, i.e. increased nutrient availability and consequent biomass increase (e.g. Stanners & Bourdeau 1995), different aspects may vary in their response to other factors.

We measured soil characteristics, above-ground biomass and densities of species and threatened species at 46 sites in wet meadows and fens in the Netherlands and Belgium to test our hypothesis that variation in species density is controlled both directly and indirectly by environmental factors, whereas variation in density of threatened species is controlled primarily via effects on biomass. We therefore predicted that regressions would show that variation in species density is at least partly explained by environmental factors that do not contribute to variation in community biomass.


Study sites

The 46 sites (4 m2) in meadows and fens were located in seven nature reserves along the river Dommel (The Netherlands) and in three reserves along the Zwarte Beek (Belgium) (5°15′–5°40′E and 51°05′–51°45′N). Five of the Dutch nature reserves were surrounded by heavily fertilized agricultural land and the other reserves by forest or extensively used meadows. All sites are subject to annual hay-making but are not fertilized.

Species composition was recorded at all sites in June 1995 by means of the Braun Blanquet method, i.e. one 4 m2 recording per site. Bryophytes, which were not recorded at all sites, are not included in species density. Only 0–3 bryophyte species occurred at our sites compared with 8–43 vascular plant species, and species density vs. biomass or environmental variables relationships have been shown to be very similar for total species density and density of vascular plants (Pollock et al. 1998), suggesting that this omission is not critical.

To enable comparison with other studies we grouped the sites into four types of wet meadows and three types of fens on the basis of vegetation composition (Table 1), clustered by means of Flexclus (van Tongeren 1986). The meadows were classified as Junco-Molinion, Calthion palustris, a degraded Calthion dominated by Holcus lanatus (Schaminée et al. 1996), or Valeriano-Filipenduletum dominated by Glyceria maxima (Stortelder et al. 1999). Fens were classified as Caricion nigrae, Carici curtae-Agrostietum caninae or Caricion gracilis. (Schaminée et al. 1995). Meadows and fens also differed in their hydrology, with fens being were constantly wet, whereas meadows were generally dryer and subject to a more dynamic water regime (Table 1) (cf. Mitsch et al. 1994).

Table 1.  Environmental, biomass and species density characteristics of the study sites, as averaged (± SD) for the major types of meadows and fens. Soil variables were measured in the top 10 cm soil layer. Two meadow sites could not be classified into one of the vegetation types
Number of sites:Junco-Molinion 8Calthion palustris 7Holcus community 6Valeriano-Filipenduletum 9Caricion nigrae 6Carici curtae Agrostietum 3Caricion gracilis 5
Environmental variables
Bulk density (g cm−3)0.48 ± 0.320.20 ± 0.020.58 ± 0.140.52 ± 0.110.32 ± 0.060.28 ± 0.120.30 ± 0.12
Organic matter (%)    35 ± 25    63 ± 4    23 ± 4    29 ± 7    38 ± 7    43 ± 17    40 ± 13
Clay content (%)     6 ± 3     9 ± 4     6 ± 1    14 ± 3    11 ± 8    12 ± 11    6 ± 2
pH-H2O 5.6 ± 0.8 5.6 ± 0.2 5.7 ± 0.2 5.1 ± 0.4 5.9 ± 0.2 5.5 ± 0.5 5.7 ± 0.2
Extractable Al (g m−2)    21 ± 17    19 ± 3    35 ± 7    34 ± 6    21 ± 13    26 ± 15    25 ± 16
Extractable Fe (g m−2)    35 ± 10    55 ± 11   118 ± 39    94 ± 15   251 ± 190    71 ± 13   139 ± 124
Extractable Ca (g m−2)    79 ± 61   107 ± 24    87 ± 25    70 ± 16    70 ± 28    72 ± 50    72 ± 9
Extractable K (g m−2) 7.2 ± 0.6      8.2 ± 0.5 7.2 ± 2.1 8.9 ± 2.0 5.3 ± 1.7 5.7 ± 0.8 7.7 ± 2.0
Extractable P (g m−2) 1.3 ± 0.5 2.0 ± 0.6 2.7 ± 0.4 4.1 ± 1.7 2.9 ± 1.7 2.4 ± 1.3 2.8 ± 1.3
N mineralization (g m−2 y−1) 2.9 ± 1.8 4.5 ± 3.710.8 ± 5.813.3 ± 4.4 1.9 ± 2.2 4.3 ± 2.3 5.8 ± 2.2
Summer moisture content (%)    49 ± 20    66 ± 6    41 ± 18    37 ± 10    69 ± 6    69 ± 10    69 ± 11
Soil moisture dynamics (%)    15 ± 8    15 ± 5    15 ± 10    25 ± 6    6 ± 3    5 ± 3    7 ± 5
Spring water level (cm-surf.)    40 ± 15    23 ± 9    35 ± 26    29 ± 7    7 ± 4    11 ± 1    5 ± 5
Water level dynamics (cm)    73 ± 21    73 ± 9    68 ± 21    88 ±     47 ± 16    59 ± 6    60 ± 16
Number of flooded sites   0   7   1   9   1   1   2
Flood/inundation depth (cm)   0     5 ± 2     2 ± 4    16 ± 6    2 ± 4    2 ± 3    4 ± 5
Above-ground biomass (g m−2)
Vascular plants   352 ± 67   479 ± 112   581 ± 173   915 ± 168   339 ± 81   584 ± 63   902 ± 132
Graminoids   274 ± 66   376 ± 66   541 ± 165   802 ± 138   279 ± 92   512 ± 73   801 ± 206
Herbs    78 ± 81   103 ± 62    40 ± 38   113 ± 85    60 ± 31    73 ± 54   101 ± 141
Bryophytes   131 ± 102    24 ± 20     5 ± 4    3 ± 3    53 ± 53    10 ± 11    16 ± 16
Total standing crop   483 ± 115   503 ± 111   586 ± 173   918 ± 168   392 ± 81   594 ± 73   918 ± 131
Species density (no. 4 m−2)
Vascular plants    25 ± 11    25 ± 5    23 ± 2    17 ± 4    23 ± 5    26 ± 6    22 ± 7
Threatened species 1.4 ± 1.8 0.6 ± 0.5   0   00.5 ± 0.6   0   0

The number of threatened species was determined according to the ‘red list’ of threatened species for the Netherlands (van der Meijden et al. 1991), including only those that had disappeared in at least 25% of map units (25 km2) compared to a historical reference.


Above-ground biomass of the vegetation (standing living and standing dead) was harvested at the peak of the growing season (17–21 July 1995). Within every 4 m2 site, vegetation in three quadrats (50 × 50 cm) vegetation was cut near the soil surface, and bryophytes were also collected. Samples from the three quadrats were combined and sorted into graminoids, herbs and bryophytes. The samples were dried for 48 h at 70 °C and weighed. Soil litter was not collected, but was not a large fraction because of the annual mowing and removal of hay.

Environmental conditions

Net N mineralization was measured at every site by in situ soil incubation (without within site replication) between 1 May 1995 and 1 May 1996. We followed the incubation technique as described in detail by Olff et al. (1994). Soil cores were incubated during five periods of 8 weeks and one winter period of 12 weeks. At the start of every period, paired soil cores (10-cm depth, 4.8-cm diameter) were taken at 1–5 cm from each other, with PVC tubes. One of the paired tubes was incubated and the other transported to the laboratory and stored overnight at 2 °C before soil extraction. Soil extractable N was determined using 1 m KCl (Olff et al. 1994) and after centrifugation, NO3 and NH4+ concentrations were measured colorimetrically with a continuous flow analyser. Net N mineralization was calculated by subtracting soil extractable NO3 and NH4+ in the initial soil cores from that in the incubated cores. Values per unit area were calculated from values per unit dry soil and the average bulk density of the top 10 cm soil layer.

Although soil extractable pool indices for P and K provide no information on flow rates, they reflect P and K availability for plants better than release rates measured by means of soil incubation techniques (cf. Jungk & Claassen 1986; Walbridge 1991; Olde Venterink 2000). Soil extractable K was determined in all initial soil cores of the mineralization experiment using 1 m HCl (Jungk & Claassen 1986; Scheffer & Schachtschabel 1989) and extractable P using ammonium acetic-acid lactic-acid (Scheffer & Schachtschabel 1989; Koerselman et al. 1993). Concentrations of K and P (and also Ca, Fe, and Al) were determined by inductively coupled plasma atomic emission spectrometry (ICP-AES), after centrifuging the extracts. Seven measurements over time were made for each element at every site: no temporal trends were observed and average values were calculated.

Most soil characteristics were also determined in the initial soil cores for the mineralization experiment. Soil moisture content was determined by measuring loss of weight of the top 10 cm soil after 48 h in a drying oven at 105 °C. Summer moisture content was measured at the end of August 1995. ‘Moisture content dynamics’ was represented by the difference between the highest and lowest of the seven values over the year. Bulk density was determined from the fresh weight of the cores and the moisture content. Organic matter content was measured by loss on ignition at 550 °C. Average bulk density and organic matter values were calculated from all initial mineralization cores of a site. Clay contents of the top 10-cm soil layer (fraction < 2 µm) were determined by sieving dried and mixed subsamples (n = 5), taken at every site in November 1995.

Acidity was measured in soil moisture derived from the top 10-cm soil, by means of Rhizon soil moisture samplers (Eijkelkamp Agrisearch Equipment, Giesbeek, the Netherlands). The average acidity values of eight soil moisture samples, taken throughout the year, were used.

Groundwater levels were measured at regular intervals (14 times) between August 1995 and September 1996, in piezometres at a depth of 0.8–1.2 m below the surface. Spring water level was taken to be the average value for the months February–April 1996 and water level dynamics to be the difference between the highest and lowest values. Some sites were flooded in spring 1995 for about a week and the flood depth was then measured.


The statistical procedure was based on Verhoeven et al. (1996a) and Bridgham et al. (1998). A PCA (with varimax rotation) was performed to reduce the number of autocorrelated environmental variables, all variables were standardized to a Z-distribution and multiplied by the factor coefficients to produce factor scores for each sample and non-normal distributed variables were log-transformed. Stepwise multiple regressions were then carried out for biomass and species richness variables using factor scores of the PCA as explanatory variables. Using regressions, instead of correlations, means that we explicitly assumed that variance in the y direction (species density or biomass) was dependent of variance in PCA score, and not vice versa.


Values of all environmental, biomass and species density variables, as averaged for the major vegetation types (groups of sites with a comparable vegetation), are shown in Table 1.

Species density vs. biomass

Species density in the meadows (or in the whole set) was neither significantly related to biomass of vascular plants, nor to total standing crop (Fig. 1a,b). For the fens, a significant quadratic relationship was observed between species density and total standing crop (P = 0.049, R2 = 0.42). Threatened species only occurred at sites with a relatively low biomass of vascular plants (Fig. 1c,d) and there was a significant negative inverse relationship for meadows (and whole set) between their number and biomass of vascular plants (P = 0.015), but not with total standing crop (P = 0.22). The explained variance was, however, weak (R2 = 0.15, 0.18, 0.28 for whole set, meadows and fens, respectively).

Figure 1.

Density of vascular plant species and threatened species vs. above-ground biomass of vascular plants (standing living and dead) or total standing crop (including bryophytes) in meadows (○) and fens (●). Threatened species are from the ‘red list’ of plant species in The Netherlands (van der Meijden et al. 1991).

Species density and productivity vs. environmental factors

A principal component analysis of the environmental variables of Table 1 resulted in four factors for meadows and three factors for fens (Table 2), with total explained variance 77% and 74% for meadows and fens, respectively.

Table 2.  Principal component analysis of environmental variables, including nutrient availability, for meadows (n = 32) and fens (n = 14). Only correlations < −0.5 and > 0.5 are shown
 Factor 1Factor 2Factor 3Factor 4Factor 1Factor 2Factor 3
Bulk density   0.97     0.96  
Organic matter content−0.95   −0.93  
Soil moisture in summer−0.88   −0.80  
Extractable Al   0.54    0.55    0.51−0.59 
N mineralization   0.570.52       0.76
Flood depth 0.81     
Extractable Fe 0.79     0.74  
Extractable P pool 0.78     0.54−0.78 
Clay content 0.69      0.80 
Extractable Ca  −0.87     0.89 
pH  −0.90   −0.70
Extractable K    0.74      0.70
Soil moisture dynamics   0.72     0.51
Water level dynamics   0.71     0.80
Water level in spring       
Variance explained (%)35191310352415

For meadows, the first factor was positively related to bulk density, extractable Al, and N mineralization, and negatively to organic matter and soil moisture (Table 2). This factor can be interpreted as representing the mineral soil–peat gradient, with higher N mineralization rates at the mineral soils. The second factor, which was positively related to flood depth, soil clay content, N mineralization and extractable P and Fe, represents differences in flooding, which is associated with higher N and P availabilities for plants at the flooded sites. The third factor, negatively related to acidity and extractable Ca, and positively related to extractable K and Al, represents the acidity gradient that was associated with a higher K availability for plants at the relatively acid sites. The fourth factor represents water regime dynamics, as it is positively related to annual dynamics of both water level and soil moisture.

For fens, as for meadows, the first factor was positively related to bulk density and negatively related to organic matter and soil moisture, but was also positively related to extractable Fe, Al and P (Table 2). The second factor, positively related to extractable Ca and clay content, and negatively to extractable P and Al, represents P availability and associated chemical complexes. The third factor represented the gradient from hydrological stable fens with low N and K availabilities on the one hand to hydrological dynamic fens with relatively high N and K availabilities on the other. The stable fens also had a slightly lower pH (cf. Table 1 for differences in pH).

Slightly more than half of the variance in biomass of vascular plants was explained by PCA factors, in both meadows and fens (Table 3). In meadows, factors 2 and 1, which represented N and P availabilities as well as flooding and bulk density, explained 54% of biomass of vascular plants. In fens, factor 3, which represented both hydrological and acidity variation and differences in N and K availabilities, explained 51% of vascular plant biomass. In both meadows and fens, the same factors explained the biomass of vascular plants, graminoids and (in the reverse direction) bryophytes. The total above-ground biomass of plants was generally related to the same PCA factors but the variance explained was less, due to the opposite influence on graminoids and bryophytes. Little of the variation in herb biomass could be explained by the PCA factors.

Table 3.  Stepwise multiple regression of above-ground biomass and species density variables vs. PCA factors of Table 2, for meadows (n = 32) and fens (n = 14)
 Entered variablesR2FPSCEntered variablesR2FPSC
  1. R2 are cumulative values per step, F and P show significant change per step, standardized coefficients (SC) are shown for the final step.

Above-ground biomass
Vascular plants
Step 1Factor 20.4423.50< 0.001 0.66Factor 30.5112.57      0.004 0.72
Step 2Factor 10.54 6.34      0.018 0.32     
Step 1Factor 20.5232.79< 0.001 0.72Factor 30.43 9.20      0.010 0.66
SteP 2Factor 10.61 6.85      0.014 0.30     
Step 1Factor 40.12 4.20      0.049 0.35   > 0.1 
Step 1Factor 20.5942.36< 0.001−0.77Factor 30.41 8.20      0.014−0.64
Step 2Factor 10.66 6.27      0.018−0.27     
Total standing crop
Step 1Factor 20.23 9.16      0.005 0.48Factor 30.42 8.61      0.013 0.65
Step 2Factor 40.36 5.51      0.026 0.35     
Step 3Factor 10.46 5.36      0.028 0.32     
Species density
Vascular plants
Step 1Factor 30.4019.75< 0.001−0.63Factor 10.39 7.81      0.016 0.63
Step 2Factor 20.5711.88      0.002−0.42     
Threatened species
Step 1Factor 20.2710.83      0.003−0.52Factor 20.41 8.40      0.013 0.64

Variation in vascular plant species density was explained by other PCA factors than those related to their biomass (Table 3). In meadows, 40% of the variance in species density was explained by factor 3 (acidity and K availability) and only an additional 17% by the ‘biomass-related’ factor 2 (flooding, N and P availability). In fens, 39% of variance in species density was explained by factor 1 (bulk density, P availability), a factor not related to biomass.

Density of threatened species, in both meadows and fens, was related to factors other than those related to species density in general. Density of threatened species decreased with the ‘biomass-related’ factor 2 in meadows and with fen factor 2 (P availability), which was neither related to biomass nor to species density in general (Table 2).

All statistical analyses were repeated for fens and meadows combined, but this did not provide new insights. The PCA for the entire data set was similar to that for meadows, although hydrological variables became more differentiating. The multiple regressions also yielded the same explanatory variables as for meadows only, but less variance was explained. Meadows were the dominant subset because of the greater number of sites and the wider range seen in most variables (Table 1).


Our results were consistent with a hump-back relationship between species and biomass (Grime 1979), given that species density is zero at zero biomass, and we had no very high biomass sites (cf. Vasander 1982; Moore & Keddy 1989; Wheeler & Shaw 1991). Most of our meadows and fens fell in the biomass range where species density is potentially high. The hump-back curve derives from the effects of nutrient stress and competitive exclusion on species density but the species pool, via factors such as niche differentiation, may affect the height of the peak and variation under the curve (cf. Grime 1979; Grace 1999). Near the peak (as for most of our sites) species density should therefore be sensitive to direct effects of environmental variables. Much variation (40%) in species density could indeed be explained by direct environmental effects rather than by environmental factors that influence biomass.

Biomass, particularly of graminoids, increased with increasing N mineralization in both meadows and fens, and with extractable P or extractable K in meadows and fens, respectively (Tables 2 and 3; bivariate regressions, P < 0.05, R2 values 0.32–0.43). This result is consistent with N, P and K as the major growth-limiting nutrients in herbaceous wetlands (Verhoeven et al. 1996b; Olde Venterink et al. 2001). Biomass related factors might explain 17% of variance in species density in the meadows (Table 3), although even this might simply reflect increased N and P availabilities (as wells as biomass) with increasing flooding depth (Table 2). Flooding was an important explanatory factor for species density in coastal wetlands, through its effect on the size of the potential species pool (Gough et al. 1994; Grace & Pugesek 1997; Grace & Jutila 1999). Although the flooding in coastal wetlands imposes much more stress than on our sites where it is restricted to a few weeks in winter or spring, it may be important for species density in freshwater wetlands (Day et al. 1988; Pollock et al. 1998).

The PCA factor that explained 40% of species density in meadows was primarily related to soil acidity. Bivariate regressions showed that, of the variables related to this factor, pH (+, R2 = 0.50) and extractable Al (−, R2 = 0.45) were the best predictors of species density. Grime (1979) found a similar strong increase of species density with increasing pH in English grasslands and Gough et al. (2000) in Arctic tundra, suggesting an effect of acidity on the potential species pool. When European plant species of wet meadow or comparable habitats are sorted according to their habitat preferences for acidity (cf. Ellenberg et al. 1991), it is clear that a much larger proportion is adapted to slightly acid to base-rich soils (R7-R8) than to acid soils (R < 4) (Fig. 2). This suggests that here too, acidity effects on density may be acting via species pool (cf. Pärtel et al. 1996; Zobel 1997). The negative correlation with extractable Al suggests that aluminium toxicity at a low pH contributes to the relatively low number of species adapted to a low pH.

Figure 2.

Distribution of Ellenberg indication values for acidity (R: 1, extremely acid to 9, alkaline; × indifferent) for all European plant species from fen, meadow or comparable habitats (i.e. species of non-shaded (L > 5), non-saline (S < 3), and wet to moist terrestrial ecosystems (F5-10), which are not restricted to alpine or warm areas (T4-6); cf. Ellenberg et al. 1991).

Although in fens 42% of variation in species density was explained via biomass, a further 39% could be explained by environmental factors not related to biomass (Table 3). The lack of an acidity effect in this habitat was probably due to the narrow pH range (Table 1), with species density increasing instead from relatively wet organic soils to dryer, relatively mineral-rich soils (Table 3). Grace & Pugesek (1997) also observed a negative relationship between species density and soil organic content in American coastal wetlands, and ascribed this relationship to a species pool effect: fewer species may be adapted to waterlogged peat-forming conditions than to aerated soils (cf. Silvertown et al. 1999).

We did not measure potential species pool sizes in our geographically separated sites but it is likely that these mediate the direct influence of environmental factors on species density in our sites, as in other grasslands and wetlands (e.g. Grime 1979; Gough et al. 1994; Grace & Pugesek 1997; Grace & Jutila 1999). However, other explanations, such as position within the landscape or historical effects (Grace & Guntenspergen 1999), which could lead to correlations between environmental factors and species density, cannot be ruled out.

About half of the variance in overall species density remained unexplained by PCA factors. Factors such as disturbance, seed bank size and seed dispersal, sampling errors, light, plant density, soil ecosystem composition and small-scale heterogeneity (Grace 1999) may be involved at our sites. Heterogeneity was also likely to have been a major factor for the unexplained biomass variance, given the lack of within-site replication in measurements of environmental variables as mineralization rates.

A second aim of this study was to find out whether the occurrence of threatened species was controlled by the same mechanism as overall species density. Moore et al. (1989) and Wheeler & Shaw (1991) showed that the hump-back curves for species density in Canadian and British wetlands, were narrower when only rare species were considered. The narrow envelope at low biomass seen for our threatened species was more marked for vascular biomass than for total standing crop (Fig. 1). This suggests that competition from more productive vascular plants (graminoids), which can benefit from increased anthropogenic nutrient sources (e.g. atmospheric N deposition), is crucial.

In meadows, density of threatened species was negatively related to the biomass-related factor 2 (Table 3) and, in particular, to P availability (bivariate regressions, P = 0.001, R2 = 0.31). A similar effect of P availability was seen in fens, although it was no longer related to biomass (Table 3). Many threatened European wetland species occur in P-limited conditions (Olde Venterink 2000), but there is large-scale disappearance of P-limited low productive wetlands in Western Europe (e.g. seepage fens; Boyer & Wheeler 1989; Wassen et al. 1996). We therefore conclude that although maintaining relatively low biomass of vascular plants (500–600 g mminus;2) is essential for conservation of threatened species (Fig. 1c), the importance of P availability for threatened species and of environmental variables, such as acidity, for species density in general means that practices such as haymaking do not, by themselves, guarantee high species richness. Management to maintain threatened species might also need to consider reducing P availability, for example, by liming or sod-cutting.


We thank N. Pieterse, A. de Hamer and W. Peeters for their help in the field and laboratory, as well as P. de Ruiter, N. Pieterse, P. Denny, J. Grace, A. Davy, L. Haddon and an anonymous referee for their helpful comments to improve the manuscript. AMINAL Ekologisch Impulsgebied Zwarte Beek, Belgian Ministry of Defence, Staatsbosbeheer, Natuurmonumenten, and the city of Eindhoven are acknowledged for permission to carry out measurements and experiments in their nature reserves. We thank the local staff of these organizations for logistical support.

Received 25 May 2000 revision accepted 2 March 2001