Vegetation composition promotes carbon and nitrogen storage in model grassland communities of contrasting soil fertility

Authors

  • Gerlinde B. De Deyn,

    Corresponding author
    1. Soil and Ecosystem Ecology Laboratory, Lancaster Environment Centre, Lancaster University, Lancaster LA1 4YQ, UK
    Search for more papers by this author
    • Present address: Department of Multitrophic Interactions, Centre for Terrestrial Ecology, Netherlands Institute of Ecology, PO Box 40, 6666 ZG Heteren, The Netherlands.

  • Helen Quirk,

    1. Soil and Ecosystem Ecology Laboratory, Lancaster Environment Centre, Lancaster University, Lancaster LA1 4YQ, UK
    Search for more papers by this author
  • Zou Yi,

    1. Soil and Ecosystem Ecology Laboratory, Lancaster Environment Centre, Lancaster University, Lancaster LA1 4YQ, UK
    Search for more papers by this author
  • Simon Oakley,

    1. Centre for Ecology and Hydrology, Lancaster Environment Centre, Lancaster LA1 4AP, UK
    Search for more papers by this author
  • Nick J. Ostle,

    1. Centre for Ecology and Hydrology, Lancaster Environment Centre, Lancaster LA1 4AP, UK
    Search for more papers by this author
  • Richard D. Bardgett

    1. Soil and Ecosystem Ecology Laboratory, Lancaster Environment Centre, Lancaster University, Lancaster LA1 4YQ, UK
    Search for more papers by this author

*Correspondence author. E-mail: gerlindede@gmail.com; g.dedeyn@nioo.knaw.nl

Summary

1. The benefits of plant functional group and plant species diversity for sustaining primary productivity have been extensively studied. However, few studies have simultaneously explored potential benefits of plant species and functional group richness and composition for the delivery of other ecosystem services and their dependency on resource availability.

2. Here, we investigated in soils of different fertility the effects of plant species and functional group richness and composition on carbon (C) and nitrogen (N) stocks in vegetation, soil and soil microbes and on CO2 exchange and the loss of C and N from soil through leaching. We established plant communities from a pool of six mesotrophic grassland species belonging to one of three functional groups (C3 grasses, forbs and legumes) in two soils of contrasting fertility. We varied species richness using one, two, three or six species and one, two or three functional groups.

3. After 2 years, vegetation C and N and soil microbial biomass were greater in the more fertile soil and increased significantly with greater numbers of plant species and functional group richness. The positive effect of plant diversity on vegetation C and N coincided with reduced loss of water and N through leaching, which was especially governed by forbs, and increased rates of net ecosystem CO2 exchange.

4. Soil C and N pools were not affected by the number of plant species or functional group richness per se after 2 years, but were enhanced by the presence and biomass of the legumes Lotus corniculatus and Trifolium repens.

5.Synthesis. Collectively, our findings indicate that changes in plant species and functional group richness influence the storage and loss of both C and N in model grassland communities but that these responses are related to the presence and biomass of certain plant species, notably N fixers and forbs. Our results therefore suggest that the co-occurrence of species from specific functional groups is crucial for the maintenance of multifunctionality with respect to C and N storage in grasslands.

Introduction

Understanding the relationship between ecosystem biodiversity and function is not only of academic interest but is also crucial in order to ensure sustainable management for the support of valued ecosystem services (Carpenter et al. 2009). To date, most studies on diversity function relationships have focussed on primary productivity, but it is clear that ecosystems provide many other services (Hooper et al. 2005; Balvanera et al. 2006; Diaz et al. 2007). Plant diversity and its interaction with soil has been shown to influence primary productivity, soil C and N sequestration (Catovsky, Bradford & Hector 2002; Oelmann et al. 2007; Fornara & Tilman 2008; Steinbeiss et al. 2008) and the retention of nutrients and water in soil and vegetation (Hooper & Vitousek 1998; Scherer-Lorenzen et al. 2003; Oelmann et al. 2007; Phoenix et al. 2008). The function of ecosystem C sequestration, in which soils play a central role (Lal 2004; Bardgett, Freeman & Ostle 2008; Smith et al. 2008), is of special importance because it offers a means to reduce atmospheric CO2 concentrations. The maintenance of ecosystem functions of water and nutrient retention are crucial in order to not only sustain ecosystem productivity but also to reduce the risk of eutrophication and the loss of diversity that this can cause (Stevens et al. 2004; Schlesinger 2009).

Several studies have addressed diversity–function relationships and found positive effects of species richness on productivity and soil processes (Hooper et al. 2005). However, these effects were often driven by plant community composition rather than by plant species or functional group richness per se (Grime 1998; Hooper & Vitousek 1998; Hooper et al. 2005). For instance, diversity effects on primary productivity in grasslands appeared to strongly depend on the presence of N-fixing legumes (Spehn et al. 2002; Fornara & Tilman 2008), while in grasslands without legumes increased primary productivity was attributed to increased nutrient use efficiency and complementarity in nutrient acquisition (van Ruijven & Berendse 2005). Reduction of soil N loss also appears to be strongly determined by plant community composition. For instance, several grassland diversity studies have found suppression of N leaching by perennial grasses across the diversity gradient (Scherer-Lorenzen et al. 2003; Oelmann et al. 2007; Phoenix et al. 2008), whereas the presence of legumes strongly enhances the risk of N loss through leaching (Scherer-Lorenzen et al. 2003; Oelmann et al. 2007).

Net storage of C and N in soil can be promoted by plant species diversity and/or composition via two main ways: first, by increasing the amount of C and N entering the soil, and second by reducing their loss from soil through respiration, volatilization and leaching (Catovsky, Bradford & Hector 2002; De Deyn, Cornelissen & Bardgett 2008; Phoenix et al. 2008). While several studies have examined how variations in plant species richness and composition influence the amount of C and N entering the soil (Oelmann et al. 2007; Fornara & Tilman 2008; Steinbeiss et al. 2008), very little is known about the residence times of this plant-derived C and N in soil, which may in part be lost quickly through leaching, or the overall drivers of the balance between the input and output of C and N (Bardgett, Freeman & Ostle 2008; De Deyn, Cornelissen & Bardgett 2008). For example, in a N-limited prairie grassland on sandy soil, plant diversity was found to enhance soil C and N accumulation by stimulating primary productivity (Fornara & Tilman 2008). This effect was mostly attributed to the co-occurrence of two functional groups, namely, legumes and C4 grasses which facilitated each others’ growth above and below ground through enhanced N input (legumes) and N retention (C4 grasses). Also, in temperate grassland communities on alluvial soil and in the absence of C4 grasses, soil C pools were found to increase with higher plant species richness (Steinbeiss et al. 2008). In this system, tall herbs acted as a key functional group through their suppression of initial C loss during the establishment of the experimental grassland communities on arable soil. Hence, different mechanisms and different plant functional groups may explain changes in soil C and N content.

Soil nutrient status is a key factor determining ecosystem properties such as primary productivity and plant diversity (Grime 1979; Tilman 1987; De Deyn, Raaijmakers & van der Putten 2004; Hooper et al. 2005), rates of decomposition (Wardle 2002) and soil C and N cycling (Wardle 2002; Dijkstra et al. 2005, 2007; Bradford, Fierer & Reynolds 2008). Also, it is well established that effects of individual plant species on soil microbial communities can vary with soil fertility (Bardgett et al. 1999; Innes, Hobbs & Bardgett 2004; van der Heijden, Bardgett & van Straalen 2008) and that soil N enrichment can modify the outcome of plant–soil feedback relationships (Manning et al. 2008). Soil nutrient status could therefore be a very important mediator of the relationship between plant species or functional group richness and ecosystem functioning. Interactive effects of soil fertility have indeed been demonstrated for the relationship between plant diversity and primary productivity, with often stronger responses being detected in more productive soil (Reich et al. 2001; Fridley 2002, 2003; Wacker et al. 2009). However, as highlighted by Hooper et al. (2005), very few studies have investigated interactive effects of soil fertility and plant diversity on ecosystem functions and, as far as we are aware, none have explicitly considered their interactive effects on ecosystem services such as soil C and N sequestration.

In this study we investigated the potential benefits of plant species richness and composition for the delivery of multiple ecosystem services related to C and N cycling and the dependency of such benefits on soil fertility. In outdoor mesocosms we investigated the effects of plant species and plant functional group richness and identity on C and N storage in soil and soil microbes and in above-ground and below-ground vegetation as well as net CO2 exchange rates and the loss of C and N from soil through leaching. Specifically, we test the hypotheses that: (i) plant species and functional group richness promote C and N pools in vegetation as well as in soil by increasing primary production and that these effects are stronger in more fertile soil; (ii) plant species and functional group richness increase C losses but reduce N losses in leachate, especially at higher soil fertility; and (iii) pools of C and N are determined by biomass of specific plant species rather than by changes in plant species richness per se.

Materials and methods

Experimental design

The model plant communities were established in September 2006 in outdoor mesocosms consisting of high-density polypropylene pots (38 × 38 cm, 30 cm deep), filled with 20 cm of soil (which was reduced to on average 14.4 cm after settling) on top of a 10-cm layer of limestone chippings (Bardgett et al. 2006). To test for effects of soil fertility on the relationship between plant community composition and diversity on C and N storage, we used two soils of contrasting fertility taken from The University of Newcastle-upon-Tyne Farm at Nafferton, Northumberland, UK (54o1′ N, 0o23′ W). These soils were collected from two adjacent grasslands with a similar free-draining alluvial sandy loam soil as determined using a particle size analyzer (Mastersizer 2000, Malvern Instruments Ltd, Malvern, UK), but of contrasting long-term fertilizer management. One grassland was intensively managed, with a history of nitrogenous fertilizer application, which has enhanced soil fertility, while the other was a low-fertility, unimproved, extensively managed grassland that had never received artificial fertilizer (Table 1). Hereafter we refer to these soils as high and low fertility, respectively. All soils were collected in April 2006 using a mechanical digger to carefully strip the surface vegetation and then collect soil from the top 20 cm. The collected soil was transported back to Lancaster University field station where the experiment was set up. Each soil was homogenized by mixing, and visible roots and stones were removed by hand prior to filling the mesocosms.

Table 1.   Initial soil characteristics of the two soils with low and high fertility
Soil fertilityTotal N–KCl (mg kg−1 soil dry wt.)%C%N C (kg m–2) N (kg m−2)%Loss-on-ignitionBulk density (g cm−3)%Clay%Silt%Sand pH
Low12.9 ± 0.4b1.59 ± 0.06a0.129 ± 0.004b2.37 ± 0.05a0.192 ± 0.004b3.1 ± 0.1b1.03 ± 0.02a2 ± 127 ± 171 ± 16.3 ± 0.1
High16.5 ± 0.4a1.75 ± 0.13a0.175 ± 0.007a2.42 ±  0.03a0.242 ±  0.003a4.3 ± 0.1a0.96 ± 0.01b3 ± 142 ± 155 ± 15.8 ± 0.1

The experiment was set up in a complete random block design containing plant communities of varying species richness, which were planted according to a substitutive design in order to keep initial plant density constant (as suggested by Balvanera et al. 2006). We used a species pool of six common grassland species belonging to one of three functional groups: two grasses, Lolium perenne L. (Lp) and Anthoxanthum odoratum L. (Ao); two forbs, Plantago lanceolata L. (Pl) and Achillea millefolium L. (Am); and two legume species Trifolium repens L. (Tr) and Lotus corniculatus L. (Lc). Among these Lp, Tr and Am are common in British fertile grassland, while Ao, Lc and Pl are common in less productive grassland (Grime & Hunt 1975; Rodwell 1992). In total, there were 15 plant treatments: no pants (1) and monocultures of Lp (2), Ao (3), Tr (4), Lc (5), Am (6) and Pl (7); two species mixtures with only one functional group, namely, grasses (G) (8), legumes (L) (9) and forbs (F) (10); two species mixtures with two functional groups, namely, GL (11), GF (12) and LF (13); mixtures with three functional groups GFL (14) with three species or with all six species (15). All communities were planted in the two soils of contrasting fertility and replicated in four blocks. For treatments with two species from two functional groups, and of three species from three functional groups, different species combinations were used per replicate such that all plant species were present in an equal number of mesocosms (van Ruijven, De Deyn & Berendse 2003) (Table 2).

Table 2.   Plant species composition and the levels of species and functional group richness and their number of replications in the model plant communities. The species used were: two grasses (G) Lolium perenne (Lp) and Anthoxanthum odoratum (Ao), two forbs (F) Plantago lanceolata (Pl) and Achillea millefolium (Am) and two legumes (L) Trifolium repens (Tr) and Lotus corniculatus (Lc)
Functional richness (composition)GroupSpecies richnessTotal/soil fertility (four blocks)
01236
0 14
1G (Lp, Ao)236
L (Tr, Lc)2
F (Pl, Am)2
G + G (Lp + Ao)1
L + L (Tr + Lc)1
F + F (Pl + Am)1
2G + F (e.g. Lp + Am)112
G + L (e.g. Ao + Lc)1
F + L (e.g. Pl + Tr)1
3G + F + L (e.g. Lp + Pl + Tr, Ao + Am + Lc)2 12
G + F + L (Lp + Ao + Pl + Am + Tr + Lc)1
Total/soil fertility (four blocks) 424248464

Plants were grown from seed: seeds were surface-sterilized in diluted sodium hypochlorite and germinated at room temperature in Petri dishes with filter paper soaked in demineralized water. Germinated seeds were transplanted in plug trays filled with autoclaved sterilized soil (1:1 mixture of low:high-fertility soil) and maintained in the glasshouse. After 8 weeks, plug trays with seedlings were put outside to acclimatize and planted a week later in the mesocosms in August 2006. Each mesocosm was placed on a saucer, above-ground vegetation was trimmed in August 2007 by clipping to 2 cm above the soil surface, and plant communities were left to regrow. Throughout the experiment mesocosms were weeded in order to maintain the original species composition.

C and N stocks in vegetation, soil and soil biota

In 2008 C and N content in the vegetation was determined from the biomass and C and N concentration in plant tissue. Above-ground vegetation was collected from each mesocosm at the end of August 2008 by clipping all shoot material above the soil surface and below-ground parameters were determined using four intact soil cores (34 mm diameter to full soil depth) collected immediately after clipping. We recorded the precise depth of all the cores and their weights and used these to calculate soil bulk density. Total mass of soil and roots was obtained by separating bulk soil and roots by sieving (2.6 mm mesh), handpicking and washing. Soil bulk density was determined using the dry weight of soil contained in the volume of the four intact cores. Vegetation and soil C and N pools were expressed on a per m2 basis by extrapolating from the sampled area (38 × 38 cm) and for the soil volume using an average mesocosm soil depth of 14.4 cm. Net changes in soil C and N pools were calculated as the difference between the amounts in 2008 and those at the start of the experiment in 2006. Vegetation was dried at 70 °C and bulk soil at 105 °C before weighing and grinding in a ball mixer. Ground material was oven-dried at 105 °C and a subsample of 5 mg analysed for %C and %N on an Elementar Vario EL elemental analyser (Hanau, Germany). The percentage of inorganic and organic C in soil was verified by analyzing the %C in oven-dry soil before and after burning the soil at 550 °C for 6 h. Soil C was predominantly (>95%) organic C and total soil %C correlated strongly positively with the % organic soil C (R2 = 0.93) and not with % inorganic soil C (R2 = 0.007); the results for total soil C hence reflected changes in organic soil C. Organic matter content in soil was determined as loss-on-ignition (% LOI) by determining weight loss after burning the soil at 550 °C for 6 h.

Soil microbial biomass C and N were determined using the fumigation extraction technique of Vance, Brookes & Jenkinson (1987). Briefly, soil samples (15 g fresh weight) were fumigated with CHCl3 for 24 h at 25 °C. After removal of CHCl3, soluble C was extracted from fumigated and unfumigated samples with 75 mL 0.5 M K2SO4 for 30 min on an orbital shaker. Total organic C in filtered extracts (Whatman No. 1, Whatman GmbH, Germany) was determined using a Shimadzu 5000A TOC analyser (Asia Pacific, Kyoto, Japan). Microbial C flush (difference between extractable C from fumigated and un-fumigated samples) was converted to microbial biomass C using a kEC factor of 0.35 (Sparling et al. 1990). Extractable N in the above extracts was determined by oxidation with K2S2O8 using the methodology of Ross (1992) and measurements of the resultant NO3-N and NH4+-N were carried out using autoanalyzer procedures. The microbial N flush was converted to microbial biomass N using a kEN factor of 0.54 (Brookes et al. 1985). Dissolved inorganic N (DIN, i.e. nitrate and ammonium) and organic N (DON) in leachates were measured by autoanalyser procedures, as was total N after oxidation of a subsample with potassium persulphate (K2S2O8). DIN was calculated by subtracting DIN from the total N measured. Plant available mineral N was measured by adding 50 mL 1.0 M KCl to 10 g soil samples, which were shaken on an orbital shaker for 1 h prior to being filtered through Whatman No. 1 paper, and nitrate and ammonium in the extract were measured by autoanalyzer procedures.

Leachate C and N

Leachate was collected from 21 April 21 until 23 June 23 2008 by capturing the drainage water that ran through each mesocosm. This period represented the time of the year during which plant growth rates were fastest. Leachates were collected in bottles via a funnel attached to the perforated bottom corner of each saucer which was placed on two breeze blocks during the leaching period, while the area between saucers and mesocosms was sealed with plastic to prevent direct rainfall in the saucers. Leachates were kept cool and were analyzed for dissolved organic C (DOC) and DIN and DON within 3 days. Leachates were filtered through Whatman No 1 filter paper; the concentration of DOC in the leachates was determined by a Shimadzu 5000A TOC analyser, and the concentrations of ammonium (NH4+) and nitrate (NO3) and DON were measured as described above. Loss of DOC, DIN and DON was determined from the volume of leachate and the concentrations of C and N for each sampling and summed to calculate total loss.

Field-based C fluxes

Ecosystem respiration rates were measured in unplanted soil, monocultures and six species mixtures at monthly intervals from February to June 2008 between mid-day and 14:00 h in the afternoon using a portable IRGA EGM-4 with SRC-1 soil respiration chamber (PP Systems, Herts, UK), Photosynthetically Active Radiation (PAR) sensor and temperature probe. To determine net ecosystem CO2–C exchange rates the soil respiration chamber was fitted with a 155-mm transparent extension hood (total volume = 3764 cm3), which transmitted 90% of ambient PAR, to enclose the vegetation with minimal disturbance.

Data analysis

The data were analyzed to investigate effects of soil fertility, plant species richness and functional group richness on various response variables. In addition, effects of species identity were analyzed using only the monocultures, while effects of species composition and functional group richness were tested using the two species mixtures. The levels of species and functional group richness were limited because of logistical constraints. Multiple regressions (standard as well as stepwise with forward selection of significant predictors) were used across all mesocosms to test the explanatory value of species or functional group richness, of above-ground biomass of specific species, and of the presence or absence of specific species (i.e. being a component species of the plant community or not) for several response variables.

We used General Linear Models (GLM) with block as a random and treatments as fixed factors to test effects of species richness, soil fertility and their interaction on change in the soil C pool, on C and N in microbes, and on DOC and water loss. In a similar way we tested effects of monoculture species identity, soil fertility and their interaction on C in vegetation and of community composition within two species mixtures on change in the soil N pool. Differences between treatments, when significant, were tested with a post hoc Tukey test.

In order to obtain equal variances values for root C and DOC were square root transformed before analysis. Because of unequal variances (tested with Levene’s test), nonparametric Kruskal–Wallis and Mann–Whitney U-tests were used to investigate effects of species richness, of community composition across two species mixtures, of soil fertility on C and N in vegetation, and on DON and DIN leaching. The same tests were used to investigate effects of species richness and soil fertility on change in the soil N pool, of plant community composition within two species mixtures on changes in the soil C pool, and of monoculture species identity on N in the vegetation, root C:N ratio and change in soil C and N pools. There was one missing value for shoot and root C and N for a monoculture of T. repens in low-fertility soil, two missing values (two different two-species mixtures in high-fertility soil) for soil microbial C and 12 for soil microbial N (six in low fertility soil: two different monocultures plus 2 two-species mixtures and 1 three- and 1 six-species mixture; six in high-fertility soil: 1 monoculture and 1 three-species mixture and 4 different two-species mixtures).

The relationship between species richness, above-ground biomass of specific species, and changes in soil C and N content and leaching losses were investigated by multiple regression analyses using species richness and above-ground biomass of each species as predictor and changes in soil C and N pools and in leached volume and DIN and DON loss as response variables. These predictors together with presence or absence of specific plant species were also used in stepwise multiple regression analysis with forward selection of significant predictors. Relationships between changes in the amount of C or N in soil and total above-ground and below-ground biomass, as well as changes between leaching volume, amount of DOC leached, and species richness were tested with simple linear regressions using Pearson correlations.

Net ecosystem CO2–C exchange rates were analyzed with repeated-measures anova to test effects of month, soil fertility, plant species richness (1 vs. 6) or community composition (six species and the monocultures) and their interactions. Net ecosystem CO2–C exchange rates in May were analyzed using GLM with block as a random factor and soil fertility, species richness or composition and their interaction with soil fertility as fixed factors. To obtain equal variances, exchange rates were square root-transformed prior to testing the effects of community composition. Differences between treatments were determined with Tukey post hoc tests. Data were analyzed with STATISTICA version 8 (StatSoft, Groningen, The Netherlands).

Results

Vegetation C and N pools

Across all plant communities, soil fertility had a strong effect on the mass of C (Mann–Whitney U-test, = −2.7, < 0.01) and N (Mann–Whitney U-test, = −2.5, < 0.05) stored in vegetation; both these measures were greatest in communities grown in the high-fertility soil (Fig. 1). The separate results for above-ground and below-ground vegetation are provided in Fig. S1. Plant species richness also strongly affected the total mass of C (KW-H = 30.1, < 0.0001) and N (KW-H = 23.2, < 0.0001) stored in vegetation (Fig. 1a,c). In both soils, plant communities with three species and monocultures contained on average most and least C and N in vegetation respectively. Mass of C and N in six-species mixtures was comparable with the values in two- and three-species mixtures. Hence, in communities with three functional groups doubling the number of species did not yield any benefits for C and N pools in vegetation, while these pools were consistently largest in communities with three and smallest in communities with one functional group (Fig. 1a,c).

Figure 1.

 Plant species and functional group (FG) richness (a,c) and monoculture species identity (b,d) effect on total (above-ground +  below-ground) carbon (c) (a,b) and nitrogen (N) (c,d) in vegetation. Abbreviations for monocultures see Table 2. Bars represent mean values ± 1 SE. Different letters denote significant differences between groups within soil fertility at < 0.05.

In monocultures, soil fertility (F1,31 = 34.38, < 0.0001) and species identity (F5,31 = 73.04, < 0.0001) had strong effects on the mass of C in vegetation (Fig. 1b) and the interaction between both factors was marginally non-significant (F5,31 = 2.58, = 0.06). In both soils, L. corniculatus monocultures contained most C in vegetation, while also the large mass of C in roots of A. millefolium monocultures is noteworthy (Fig. S2). Differences between species in the mass of N stored in vegetation (KW-H = 38.97, < 0.0001) were comparable with the pattern found for C (Fig. 1d), while soil fertility had no significant effect on mass of N stored in vegetation in monocultures (Mann–Whitney U-test; = 1.12, = 0.27). Overall, across all monocultures and in both soils, L. corniculatus consistently stored most C and N in vegetation. Not only the quantity but also the quality of plant tissue entering the soil in terms of the C:N ratio of roots strongly differed between species (KW-H = 33.0, < 0.001), independent of soil fertility. Across all species both legumes had the lowest root C:N ratios (19.5 ± 0.4 for L. corniculatus and 21.2 ± 3.1 for T. repens), while the forb A. millefolium had the highest root C:N ratio (74.5 ± 6.0).

In order to asses the impact of the most productive plant species in monoculture (L. corniculatus) on the relationship between species richness and C and N pools in vegetation, the effect of plant species richness was tested in plant communities with or without L. corniculatus (Fig. 2). In plant communities with L. corniculatus mass of C in vegetation did not increase with species richness, while in plant communities without L. corniculatus it did (Fig. 2). Overall, the mass of C stored in vegetation was higher in plant communities with L. corniculatus than in those without. Similar responses were found for the mass of N in vegetation (results not shown). The separate results for above-ground and below-ground vegetation are provided in Fig. S3.

Figure 2.

 Effect of dominant plant species (Lc) on carbon (C) in vegetation. Bars represent mean ± 1SE. Different letters denote significant differences between groups within soil fertility and within all communities or only those with (+Lc) or without (−Lc) Lotus corniculatus as component species at < 0.05.

The importance of plant functional group richness and composition for C and N stored in vegetation was tested at a constant species richness level of two species. In both soils, functional group composition affected the mass of C (KW-H = 29.96, < 0.0001) and N (KW-H = 36.77, < 0.0001) stored in vegetation. Plant communities with two legume species or with one legume plus one grass or one forb species contained most C in vegetation (Fig. 3a). A similar pattern was found for the mass of N in vegetation (Fig. 3b). The separate results for above-ground and below-ground vegetation are provided in Fig. S4. Overall, for two-species mixtures the presence of at least one legume species had an overriding effect on C and N stored in vegetation rather than functional group richness per se.

Figure 3.

 Effect of functional group (FG) richness and composition (G = grass, F = forb, L = legume) within two-species mixtures on carbon (C) (a) and nitrogen (N) (b) in vegetation. Bars represent mean values ±1SE. Different letters denote significant differences between groups within soil fertility at < 0.05.

Across all planted mesocosms, stepwise multiple regressions showed that total C stored in vegetation was positively related to the biomass of all plant species and most strongly to that of L. corniculatus (β = 0.93, B = 0.72, < 0.0001), while species or functional group richness per se or the presence or absence of specific plant species showed no significant relation to total C in vegetation. Total N stored in vegetation related positively to above-ground biomass of all plant species except that of P. lanceolata, and most strongly to that of L. corniculatus (β = 0.78, B = 0.03, < 0.0001) and T. repens (β = 0.48, B = 0.05, < 0.0001). The presence of specific species in the plant communities also affected total N in vegetation: the presence of L. corniculatus related positively (β = 0.17, B = 2.67, < 0.001) and that of A. odoratum negatively (β = −0.15, B = −2.30, < 0.05) to total N in vegetation, but again there was no significant relationship between plant species richness or functional group richness per se and vegetation N.

Soil C and N pools

Despite the strong effects of species richness and soil fertility on C and N in vegetation, we did not detect any effects of plant species richness (F4,115 = 0.08, > 0.05) or soil fertility (F1,115 = 1.65, > 0.05) on changes in the pool of soil C of planted mesocosms, which on average declined with 121 ± 21 g C m−2 in 2 years. Also, the size of the soil N pool declined independently of plant species richness (KW-H = 2.25, > 0.05), while the reduction in the soil N pool over 2 years was marginally greater in fertile (13 ± 3 g N m−2) than in less fertile soil (5 ± 2 g N m−2) (Mann–Whitney U-test; = 1.98, = 0.05).

In monocultures there was a significant effect of species identity on the changes in the pool of soil C (KW-H = 13.04, = 0.023) and to a lesser extent in the pool of soil N (KW-H = 10.13, = 0.072) (Fig. 4a,b). In most species monocultures there was a net loss of soil C and N after 2 years, especially in the grass L. perenne and the forb A. millefolium, while the legume T. repens caused a gain in soil C (59 ± 94 g C m−2) and N (10 ± 4 g N m−2), and soil fertility had no effect on these measures (Mann–Whitney U-test = 1.22 for C and −1.67 for N, > 0.05).

Figure 4.

 Effect of monoculture species and of functional group (FG) richness and composition within two-species mixtures on soil carbon (C) and nitrogen (N) change 2008–2006 (g m−2 2 year−1). Abbreviation for monocultures see Table 2 and for FG see Fig. 3. Bars represent mean ± 1 SE. Different letters denote significant differences between groups at < 0.05.

In mixtures of two plant species functional group composition significantly affected soil C (KW-H = 12.50, = 0.029) and N (F5,33 = 5.72, < 0.001) (Fig. 4c,d), but only in mixtures of two legumes was there a gain in soil C (199 ± 139 g C m−2) and N (19 ± 13 g N m−2) after 2 years. The loss in soil C and N was largest in communities that contained forb species (Fig. 4c,d). In two-species mixtures, soil C declined less or increased more in high- than in low-fertility soil (Mann–Whitney U-test; = 2.06, < 0.05), while soil fertility did not affect change in soil N (F1,33 = 0.43, > 0.05).

The importance of plant species richness and biomass of specific plant species for changes in soil C and N content was verified using multiple regressions, whereby plant species richness (including unplanted mesocosms) and above-ground biomass of each species was used as a predictor and changes in soil C and N content as response variables (Table 3). A strong positive relationship was found between the change in soil C content and above-ground biomass of L. corniculatus and T. repens, indicating that these species increased soil C content. A negative relationship was found with above-ground biomass of A. millefolium, and there was no significant relationship with species richness (Table 3) or total vegetation biomass (R= 0.021, = 0.09, = 128) and change in soil C content. Changes in soil N content did not relate significantly to total vegetation biomass (R= 0.004, = 0.48, = 128) or species richness but this measure was negatively related to above-ground biomass of A. millefolium and P. lanceolata (Table 3). Stepwise multiple regression showed that changes in soil C were positively related to above-ground biomass of L. corniculatus (B = 0.22, < 0.05) and presence of T. repens (B = 92.25, < 0.05) but negatively to above-ground biomass of A. millefolium (B = −1.06, < 0.05) and presence of L. perenne (B = −88.95, < 0.05) (Table 3). Changes in soil N related positively to the presence of T. repens (B = 19.29, < 0.05) and L. corniculatus (B =8.66, = 0.07) but negatively to above-ground biomass of A. millefolium (B = −0.10, < 0.05) (Table 3).

Table 3.   Regression coefficients of multiple regressions of plant species richness (SR) and species above-ground biomass (g m−2) on soil carbon (C) and nitrogen (N) pool change over 2 years (g m−2), leaching volume (mL mesocosm−1) and DIN (mg mesocosm−1) loss in spring across planted and unplanted mesocosms. For species names see Table 1. Coefficients in italics indicate significance of species presence, underlined ones indicate significance of species biomass or species richness (SR) following stepwise forward parameter selection
Regression parameters
Response variableInterceptSRLcTrAoLpAmPIR2F7,120
  1. DIN, dissolved inorganic N; Lc, Lotus corniculatus L; Tr, Trifolium repens; Ao, Anthoxanthum odoratum L.; Lp, Lolium perenne L.; Am, Achillea millefolium L.; Pl, Plantago lanceolata.

  2. Significance: (*)< 0.10, *< 0.05, **< 0.01, ***< 0.001, ****< 0.0001, NS, not significant.

Soil C20082006−103.64*−11.73, NS±0.24*+0.60(*)−0.04, NS−1.07, NS−1.11(*)−0.26, NS0.122.39*
Soil N2008–2006−4.21, NS+1.88, NS0.01, NS+0.02, NS−0.07, NS−0.08, NS−0.18**−0.16*0.132.54
Volume leached+1663****−73.46**−0.26(*)−0.44, NS0.29, NS+3.05**−2.71***−2.85**0.369.51*
DIN leached+10.38****−0.53, NS−0.01***−0.02*−0.04****−0.06**−0.06****−0.07**0.307.50**

We also measured soil microbial biomass N and C after two growing seasons and found that across all plant communities, microbial biomass N was not affected by plant species richness (F4,103 = 1.54, = 0.20), soil fertility (F1,103 = 0.79, = 0.38) or their interaction (F4,103 = 0.19, = 0.94). In contrast, microbial biomass C was affected by plant species richness (F4,113 = 3.25, = 0.01) and soil fertility (F1,113 = 14.60, < 0.001) but there was no interaction between these two factors (F4,113 = 1.38, = 0.24). Post hoc analysis revealed that microbial biomass C was greater in high fertility (419 ± 27 μg C g−1 soil dry wt.) than in low-fertility soil (311 ± 19 μg C g−1 soil dry wt.). Moreover, in high-fertility soil microbial biomass C was significantly greater in mixtures of six plant species (607 ± 171 μg C g−1 soil dry wt.) than in unplanted soil (276 ± 84 μg C g−1 soil dry wt.).

Ecosystem CO2–C fluxes

Sampling month had strong effects on net ecosystem CO2–C exchange rates in both soils when testing effects of species richness (F4,208 = 28.04, < 0.0001) and plant community compositions (F4,132 = 40.73, < 0.0001). Overall, net ecosystem CO2–C exchange rates were most negative (indicating net C assimilation) in May, least negative in April and intermediate in February, March and June (Fig. 5a). The effect of sampling time is most likely due to a combination of differences in PAR and the cover of photosynthetically active vegetation, while differences in soil temperature and respiration activity of soil biota and plants may have also played a role. Average PAR was highest in May (1548 ± 27 μmol m−2 s−1) and so was average soil temperature (22.6 ± 0.3 °C); in June, average PAR was 915 ± 57 μmol m−2 s−1 and average soil temperature 17.5 ± 0.4 °C, so that both photosynthetic and respiratory rates were reduced. The interaction between plant species richness and sampling month was significant (F4,208 = 2.37, = 0.05; Fig. 5a). Only in May did mixtures of six plant species have greater net ecosystem CO2–C exchange rates than the average of the monocultures (F1,49 = 5.76, = 0.02), and this was not affected by soil fertility (F1,49 = 0.49, = 0.49). The effects of plant community composition were also significant in May (F6,39 = 5.94, < 0.001), with six-species mixtures and legume species having greatest, forbs intermediate and grasses lowest net ecosystem CO2–C exchange rates (Fig. 5b).

Figure 5.

 Effect of (a) species richness on net ecosystem CO2–C exchange rate in February–June and (b) species richness and identity on net ecosystem CO2–C exchange rate in May. Bars represent mean ± 1 SE. Asterisk and different letters denote significant differences between plant communities at < 0.05. For species names see Table 2.

Loss of C and N through leaching

The amount of organic C lost through leaching (DOC) was not affected by plant species richness, even though unplanted mesocosms were included in the analysis (F4,115 = 1.23, > 0.05). However, the amount of DOC lost in leachates was positively correlated with the volume of drained leachate (R= 0.13, < 0.0001), which in turn was negatively related to plant species richness (R= 0.13, < 0.0001). Soil fertility (F1,115 = 4.84, = 0.03) as well as plant species richness significantly affected the total volume of leachate produced (F4,115 = 6.97, < 0.0001), and there was no interaction between soil fertility and species richness (F4,115 = 1.59, = 0.18). In both soils, plant species richness also had strong effects on DON (KW-H = 20.29, < 0.001) and DIN (KW-H = 33.41, < 0.0001) leaching (Fig. 6). The presence of plants strongly reduced DON and especially DIN loss. Moreover, in high-fertility soil mixtures of six species had lower DON and DIN loss compared with the average loss from the monocultures. Multiple regression analysis across planted and unplanted mesocosms in both soils revealed that leaching volume significantly declined with increasing species richness and with above-ground biomass of L. corniculatus, P. lanceolata and especially A. millefolium. However, leaching volume increased with larger above-ground biomass of L. perenne, while loss of DIN declined with increasing above-ground biomass of all plant species (Table 3). Leaching loss was not only affected by plant species richness and total plant biomass but also by the presence of specific species. Leached volume was positively related to the presence of L. perenne (B = 274.65, < 0.001), negatively to above-ground biomass of A. millefolium (B = −3.81, < 0.01) and marginally negatively to plant species richness (B = −101.26, = 0.06) (Table 3). Loss of DIN was positively related to the presence of P. lanceolata (B = 3.60, = 0.07) but decreased significantly with increasing species richness (B = −1.91, < 0.05) and with increasing above-ground biomass of A. millefolium (B = −0.08, < 0.05), P. lanceolata (B = −0.10, < 0.001), A. odoratum (B = −0.03, < 0.05) and L. perenne (B = −0.02, < 0.05).

Figure 6.

 Effect of species and functional group richness (FG) on organic (DON) and inorganic (DIN) nitrogen leaching in spring. Bars represent mean ± 1 SE. Different letters denote significant differences between groups within soil fertility at < 0.05.

Discussion

In this study we investigated in model plant communities the potential benefits of vegetation composition for the delivery of multiple ecosystem services related to C and N cycling, and the dependency of these benefits on resource availability. We found that soil fertility, plant species and functional group richness and plant identity all had significant effects on C and N pools in vegetation and the loss of water and N in leachates, while changes in soil C and N content were most responsive to the biomass or presence of specific plant species. Although the absolute mass of C and N was generally greater in high- than in low-fertility soil, the response patterns to species richness and composition were similar in both soils. Remarkably, the strong increase in C and N pools in vegetation with higher plant species richness did not translate into increases or lower losses of soil C and N. Rather, changes in soil C and N pools were related more to the above-ground biomass or presence of specific plant species than to plant species richness or total community biomass per se. However, we expect that over time the effects of functional groups will become more pronounced, and hence sequestration of soil C and N will increase with their richness, especially given the increase in primary productivity that was found with increasing plant diversity. In contrast to our expectation, higher species richness and primary productivity did not translate to increased loss of DOC, but as expected, the loss of DIN and DON in leachates was strongly reduced by higher species richness and biomass. The most productive species in monoculture, Lotus corniculatus, consistently dominated mixed plant communities and consequently contributed most to the total C and N pool in vegetation across levels of species richness. The positive relation between species richness and C and N in vegetation can in part be attributed to a sampling effect (Huston 1997), given that the number of communities with versus without L. corniculatus increased with increased levels of species richness (i.e. L. corniculatus was present in a sixth of all monocultures, a third of all two-species mixtures, half of the three-species mixtures and in all six-species mixtures). However, when only considering communities without L. corniculatus, C and N storage in vegetation also increased with increased levels of species richness, indicating that the positive relationship is due to more than a sampling effect. The rates of C and N accumulation in soil, as well as leaching losses, were not primarily driven by the most dominant species (L. corniculatus), and C and N pools in soil did not change with species richness. The relationship between these response variables and plant community composition can thus not be explained by a sampling effect.

Vegetation C and N pools

Our finding of larger C and N pools in above-ground and below-ground vegetation with increased plant species and functional group richness corresponded with earlier studies (Spehn et al. 2002; Hooper et al. 2005; Fornara & Tilman 2008). However, we also found that the C and N pools in vegetation were mostly determined by the biomass of the most productive species in monoculture, namely, the legume L. corniculatus. Overall, the mass of C and N in vegetation was larger in plant communities grown in high- than in low-fertility soil, but the level of enhancement of these pools by species richness was comparable in both soils. This contrasts with the results of Reich et al. (2001) and Fridley (2002) who found stronger effects of species richness at higher levels of soil fertility and may be due to the different identity of plant species used in these experiments (Fridley 2002; Wacker et al. 2009).

Soil C and N pool changes

The few studies that have examined effects of plant species richness on soil C and N sequestration have reported positive effects which have been attributed to associated increases in below-ground biomass (Fornara & Tilman 2008; Steinbeiss et al. 2008). In contrast, we did not find promotion of soil C and N pools by species or functional group richness, despite their strong positive effects on C and N pools in vegetation. This could be due to the relatively short duration of our experiment and initial C and N losses after soil disturbance (Steinbeiss et al. 2008). Plant community composition, however, did affect soil C and N content; both of these measures were promoted by the presence of L. corniculatus and T. repens, which are both common legume species of grassland. The role of key functional groups for soil C and N sequestration was also stressed by Fornara & Tilman (2008). In that experiment, the combination of legumes and C4 grass species had overruling effects on C and N accumulation rates, while in the study of Steinbeiss et al. (2008) tall herbs were responsible for greater accumulation of C in soil. In our experiment, there were no C4 grasses and C3 grasses did not appear to perform a similar role in terms of complementing legumes, although we did find a main effect of the legumes on C and N storage in soil. In contrast to Steinbeiss et al. (2008), we found a negative effect of the tall herb A. millefolium (following the grouping of Roscher et al. 2004) on soil C and N content, but this species did store more C and N in vegetation than the other non-leguminous species. Overall, our results point to the consistent key role that legumes play in delivering ecosystem services of soil C and N sequestration in soils of contrasting fertility, and indicate that the biomass and/or presence of specific species rather than species richness per se or total community biomass can promote soil C and N sequestration.

Loss of C and N through leaching

Retention of nutrients and water are important ecosystems services which underpin sustainable productivity and diversity conservation (Stevens et al. 2004; Drinkwater & Snapp 2007; Schlesinger 2009). Several studies have reported reduced DIN losses with increased plant species richness (Hooper et al. 2005), but our study is unique in that we also considered a potential interaction with soil fertility. Moreover, we also considered the loss of DON and DOC simultaneously, which both constitute a significant and often neglected source of C and N loss from soil (Bardgett 2005). In unplanted soils, the low-fertility soil leached smaller volumes of water than the high-fertility soil, despite it being sandier and having lower organic matter content. We speculate that the unplanted low-fertility soil may have dried quicker and had larger water deficits before rainfall events than did the more fertile soil. Also, potential differences in edge effects by soil shrinking when drying may have played a role. However, in both soils, plant presence strongly reduced the amounts of DIN and DON loss through leaching, while in the high-fertility soil also higher species and functional group richness decreased DIN and DON loss. Especially for DIN leaching, this result may be attributed to the increased amount of N immobilized in vegetation (Dijkstra et al. 2007; Oelmann et al. 2007). In contrast, losses of DON can increase when N pools in vegetation and their decomposition become larger as with increased species richness (Dijkstra et al. 2007) or increased biomass of legumes (Scherer-Lorenzen et al. 2003; Oelmann et al. 2007). Unlike in the aforementioned studies (Scherer-Lorenzen et al. 2003; Dijkstra et al. 2007; Oelmann et al. 2007), we found reduced DIN and DON loss with increased species richness, which was likely to be due to the reduced leaching volume. Loss of DIN was especially lower with higher biomass of the forbs A. millefolium and P. lanceolata and the grasses A. odoratum and L. perenne, while for DON losses there was no particular relationship with specific plant species.

In our experiment, several functional groups contributed to reduced N leaching, which contrasts the findings of Phoenix et al. (2008) who found an overruling role for grasses. Our results may especially apply to young communities establishing on bare soil, while the results of Phoenix et al. (2008) were obtained in mature communities without legumes. Species richness and functional group richness did not increase DOC loss by soil leaching, despite their strong positive effect on plant biomass and C assimilation rate, and hence on C supply to soil. Two co-occurring responses may explain our findings. First, the volume of soil water that leached out over the sampling period declined significantly with increasing plant species and functional group richness and the associated higher plant biomass, and second, the most diverse and productive communities also stimulated C immobilization in microbial biomass and potentially altered its functioning. Overall, more species-rich communities provided a buffer for losses of inorganic N by reducing soil water leaching and N immobilization in vegetation and this did not trade off against high losses of organic N and C. We have to bear in mind that our results of leaching losses cover losses that incurred during spring and we cannot exclude different outcomes if losses occurred over a different season or over an entire year. Clearly, longer-term experiments are warranted to determine whether and how effects of community composition on N and C losses change over time during plant community development.

Mass ratio, plant species richness and plant traits

In line with the mass ratio hypothesis (Grime 1998), we found that the absolute mass of C and N stored in vegetation was determined by the dominant species (L. corniculatus) rather than by species or functional group richness per se. The increase in the mass of C and N stored in soil caused by vegetation did not directly relate to total above-ground or below-ground biomass, in contrast to findings of Fornara & Tilman (2008), but rather specifically to the presence and above-ground biomass of legumes in mixed communities. This discrepancy may be due to the absence of C4 grasses in our system as well as to the shorter duration of our experiment. Given that both legume species had lower tissue C:N ratios than other species, the promotion of soil C sequestration by the legumes contrasts the expectation that species with high tissue C:N ratio would be generally more beneficial for this ecosystem function than plant species with low tissue C:N ratios (De Deyn, Cornelissen & Bardgett 2008). This result can be attributed to the N-limited status of the soil and the overruling effect of the N-fixing capability for promoting plant growth and hence C and N input quantity. Effects of poor litter quality on soil C and N sequestration may only become apparent in the longer term in more mature communities. Within legumes, the trait of N-fixation alone is unlikely to explain the overall dominance of L. corniculatus because T. repens is a more efficient N fixer (Spehn et al. 2002). Other important plant traits of L. corniculatus may be deeper rooting, better drought resistance and less proneness to shading given its higher stature compared with T. repens (Burdon 1983; Jones & Turkington 1986).

Conclusions

Overall, our results demonstrate that plant community composition can exert strong effects on multiple ecosystem functions of grassland in a consistent way in soils of contrasting fertility. Importantly, we found that while plant species richness positively affected stocks of C and N in vegetation, the storage of C and N in soil did not relate strongly to total C and N in vegetation or to plant species and functional group richness. However, storage of C and N in soil was strongly dependent on the above-ground biomass and/or the presence of legume species, while reduction of water loss and inorganic N in leachates was mostly governed by the above-ground biomass of the forb and grass species. These findings indicate that several different species contributed to the provision of a multitude of ecosystem functions, and thus their combined presence and biomass is of fundamental importance to the maintenance of the multifunctionality of grasslands, as suggested by Hector & Bagchi (2007) and Mokany, Ash & Roxburgh (2008). Our results indicate that further understanding of the relationship between community composition and ecosystem multifunctionality, including soil C and N storage, could be provided by using a functional trait and trade-off based approach (Diaz et al. 2007; De Deyn, Cornelissen & Bardgett 2008; Mokany, Ash & Roxburgh 2008).

Acknowledgements

This work was supported by funds from the BBSRC Agri-Food Committee, awarded to RDB. We thank William Taylor for collection of the soils and Kate Harrison and Will Mallott for help with setting up the experiment. We also thank two anonymous referees and the Handling Editor for their constructive comments on an earlier version of this manuscript.

Ancillary