Affiliations Institute of Ecology, Friedrich-Schiller-University Jena, Jena, Germany, Terrestrial Ecology, Department of Ecology and Ecosystem Management, Center of Life and Food Sciences Weihenstephan; Technische Universität München, Freising, Germany ⨯
Affiliation Geographic Institute, University of Berne, Berne, Switzerland ⨯Affiliation Institute for Chemistry and Biology of the Marine Environment, Carl von Ossietzky University of Oldenburg, Wilhelmshaven, Germany ⨯
In the course of the biodiversity-ecosystem functioning debate, the issue of multifunctionality of species communities has recently become a major focus. Elemental stoichiometry is related to a variety of processes reflecting multiple plant responses to the biotic and abiotic environment. It can thus be expected that the diversity of a plant assemblage alters community level plant tissue chemistry. We explored elemental stoichiometry in aboveground plant tissue (ratios of carbon, nitrogen, phosphorus, and potassium) and its relationship to plant diversity in a 5-year study in a large grassland biodiversity experiment (Jena Experiment). Species richness and functional group richness affected community stoichiometry, especially by increasing C:P and N:P ratios. The primacy of either species or functional group richness effects depended on the sequence of testing these terms, indicating that both aspects of richness were congruent and complementary to expected strong effects of legume presence and grass presence on plant chemical composition. Legumes and grasses had antagonistic effects on C:N (−27.7% in the presence of legumes, +32.7% in the presence of grasses). In addition to diversity effects on mean ratios, higher species richness consistently decreased the variance of chemical composition for all elemental ratios. The diversity effects on plant stoichiometry has several non-exclusive explanations: The reduction in variance can reflect a statistical averaging effect of species with different chemical composition or a optimization of nutrient uptake at high diversity, leading to converging ratios at high diversity. The shifts in mean ratios potentially reflect higher allocation to stem tissue as plants grew taller at higher richness. By showing a first link between plant diversity and stoichiometry in a multiyear experiment, our results indicate that losing plant species from grassland ecosystems will lead to less reliable chemical composition of forage for herbivorous consumers and belowground litter input.
Citation: Abbas M, Ebeling A, Oelmann Y, Ptacnik R, Roscher C, Weigelt A, et al. (2013) Biodiversity Effects on Plant Stoichiometry. PLoS ONE 8(3): e58179. https://doi.org/10.1371/journal.pone.0058179
Editor: Andrew Hector, University of Zurich, Switzerland
Received: October 29, 2012; Accepted: January 31, 2013; Published: March 4, 2013
Copyright: © 2013 Abbas et al. This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Funding: The Jena Experiment is funded by the German Science Foundation, FOR 456,1451 (www.dfg.de). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.
Competing interests: The authors have declared that no competing interests exist.
Recent years have seen the rise of a strong body of literature examining the effects of biodiversity on ecosystem functioning (BDEF), which has been triggered by increasing concerns about potential consequences of loosing species in ecosystems worldwide. Recent syntheses of BDEF research in experimental ecosystems concluded that the loss of biodiversity reduces ecosystem process rates and stability [1]–[3]. However, there is substantial concern that many previous studies underestimate the strength of BDEF relationships because biodiversity effects become stronger in more complex settings and many of the previous experiments were restricted to a low maximum diversity [4]. Most BDEF studies addressed single ecosystem processes (e.g. primary production), but the loss of biodiversity might impact single processes less than “ecosystem multifunctionality”, which is defined as the composite of multiple ecosystem processes [5]. Different species might contribute to different processes in ecosystems, and recent work has shown that the loss of species is more likely to influence multiple processes rather than single processes (e.g., [6]–[8]).
Ecological stoichiometry (ES) [9] ties multiple processes in ecosystems together as the relation between organisms demand for multiple elements and the availability of these elements in their resources has profound impact on process rates and the relative importance of different processes. The nutrient stoichiometry of plant tissue can be decisive for species interactions with other trophic levels (herbivory, pathogen infestation) and nutrient recycling [10], [11]. Consequently, the ratios of carbon:nutrients can be used as a main predictor of the relative role of herbivory and detritus pathways in ecosystems [12], [13]. Different plant species may significantly affect the usage of different elements, and processes connected to these elements such as carbon (C) - based total primary productivity or cycling of nitrogen (N) or phosphorus (P), because plants separately consume anorganic elemental resources. Therefore, plants can show trade-offs in resource uptake and storage efficiency for different elements, which leads to higher plasticity in their elemental composition compared to animals, which consume resource packages [9], [14].
Unifying ES and BDEF research potentially creates new insights in how communities process available nutrients depending on the number of species involved. However, BDEF research has largely ignored stoichiometric considerations of ecosystem processes, whereas the analysis of ecological stoichiometry has rarely involved biodiversity because ES research often focused upon single species per trophic group or large-scale analyses in certain vegetation types or biomes (e.g., [15], [16]). In a pioneering study using algal microcosms, phytoplankton diversity was shown not only to alter primary productivity and P use, but also C:P ratios [17]. Ptacnik et al. [18] suggested on a more general level that plant elemental composition should vary with plant diversity (hypothesis H1), with different outcome depending on how much plant diversity affects resource use efficiency or storage for different elements. If species are highly complementary in their C-acquisition (e.g. different strategies in light acquisition through morphological or physiological traits), but not for mineral nutrients, then C should increase more rapidly with richness than N in community-wide chemical composition and C:nutrient ratios should increase (H1a). In contrast, if plant species show complementarity mainly for the uptake of organic or mineral nutrients but not for C-fixation, then increasing plant diversity should decrease C:nutrient ratios (H1b). The null hypothesis to both is that stoichiometry of community-wide chemical composition is independent of diversity because acquisition of different elements is so strongly coupled that no stoichiometric change is observed or because complementarity in resource acquisition traits is lacking.
It is important to test these potential diversity effects over time as nutrient availability and light limitation might change during community development, leading to time-dependent effects of diversity on stoichiometry (hypothesis H2). In addition to shifting mean ratios, plant diversity will also reduce the variance in chemical composition (hypothesis H3). This can be due to multiple mechanisms underlying diversity effects on resource uptake: More species increase the chance for selection or complementarity effects maximizing nutrient incorporation for each element across the assemblage and thereby lowering the variability of elemental concentrations. Alternatively, the variance in chemical composition can decrease with richness by a statistical averaging effect [19], where more species mask the signature of the stoichiometry of single species.
Here, we tested these three hypotheses on community-wide chemical composition using a grassland biodiversity experiment comprising communities of different species richness (1, 2, 4, 8, 16, and 60) and functional group richness and composition (1 to 4; legumes, grasses, small herb, tall herbs), the Jena Experiment [20]. We analyzed the relationship between plant diversity (species richness, functional group richness and functional composition) and the stoichiometry of plant chemical composition (C, N, P and K) over the first five years of the experiment.
The study was conducted in the Jena Experiment, a large biodiversity experiment established in 2002. The experimental site is located on the floodplain of the river Saale in Jena (Thuringia, Germany, 50°55′ N, 11°35′ E, 130 m a.s.l.) [20]. The area around Jena has a mean annual temperature of 9.3°C and an average annual precipitation of 587 mm [21].
The experimental design is described in detail in Roscher et al. [20]. Briefly, the main experiment comprises 82 plots of 20 m×20 m size. The soil of the experimental site is an Eutric Fluvisol. Due to flooding dynamics, the soil texture ranges from sandy loam close to the river Saale to silty clay with increasing distance from the river. Species were randomly drawn from a pool of 60 perennial species characteristic for Central European semi-natural, species-rich mesophilic grassland communities (Molinio-Arrhenatheretea [22]). According to the results of a cluster analysis of a literature-based matrix of functional traits, plant species were divided into 4 functional groups (16 grasses, 12 small herbs, 20 tall herbs, and 12 legumes). The experimental design ensures that the presence/absence of each functional group is minimally confounded with species number. Plant communities were established with different levels of species richness increasing on a logarithmic scale (1, 2, 4, 8, 16 and 60). Each species-richness level was replicated with 16 plots with different species composition, only species mixtures with 16 and 60 species were replicated on 14 and 4 plots, respectively. In addition to the main experiment, each experimental species was sown in replicated monocultures resulting in 120 plots of 3.5 m×3.5 m size.
To account for the gradient in soil characteristics, a block design was used with blocks arranged parallel to the river Saale. The plots were mown twice a year, in June and September, and mown material was removed. Additionally, plots were weeded at the beginning of the growing season and after first mowing to maintain the sown species combinations. Weeding was done mostly by hand [23].
Aboveground biomass was harvested from 2003 to 2007 at estimated peak standing biomass in late May prior to mowing. Plants were clipped at 3 cm above ground level in four rectangles of 20×50 cm size. In May 2005 only three samples were harvested. Biomass on small monoculture plots was sampled with two replicates. Sample location was selected randomly for each harvest leaving out the outer 70 cm of the plot. After harvest, plant material was sorted into sown species, species which were not sown at a particular plot and detached dead material. Biomass was dried at 70°C for at least 48 hours [23].
Biomass samples of the entire plant community per plot were shredded and milled for chemical analyses. Thus, all stoichiometric analyses were done on the mixture of pooled plot biomass. N and C concentrations were determined by an Elemental Analyzer (EA, Vario EL III, Elementar, Germany). Plant material were digested with HNO3 at 200°C using a microwave system (MARS5Xpress, CEM, Germany) to analyze P photometrically after irradiation with UV and oxidation with K2S2O8 with a Continuous Flow Analyzer (AutoAnalyzer, Bran&Luebbe, Germany) and K using atomic absorption spectrometry (AAS 240 FS, Varian, Germany) [24].
Bivariate molar ratios of C, N, P and K in the plant community were analyzed across the gradient of plant diversity with different levels of species and functional group richness. We present molar ratios as these are standard in ecological stoichiometry [9]. Since the six different molar ratios (C:N, C:P, C:K, N:P, N:K, P:K) were not independent of each other, a multivariate analysis of variance (MANOVA) was performed with the following factors: block, sown species richness, functional group richness, legume presence, grass presence. Legumes and grasses were explicitly tested because of potential strong impact on N (N2-fixing) and C (C-storage) concentration. We used the Pillai’s trace statistic, which is recommended to test for significant effects on interdependent response variables [25]. In cases of significant effects we used additional univariate tests for each ratio to analyze which ratios responded significantly to the factor. We opted for testing years separately in order to avoid a repeated measurement MANOVA which would be difficult to interpret.
We made two sensitivity analyses to test the robustness of our results. First, the MANOVA was repeated without the 60 species mixture to test for effects of this less replicated treatment. The results were comparable to those including the 60 species mixtures (Table S1), and changes were restricted to a reduced significance level reflecting the smaller statistical power (6 out of 25 results) and changes in the bivariate ratios becoming significant in the univariate tests of factors significant in the MANOVA (8 out of 25 results). Second, we changed the order of effects and tested functional group richness before testing species richness (see Table S2). Here we found that functional group richness often replaced species richness in importance, reflecting that changes in stoichiometry depended on functional and species richness in a comparable way. We present this alternative model in the supporting online material for comparison.
To analyze how increasing diversity levels effect the bivariate ratios and their variance, we used the software package generalized additive model for location, scale and shape (GAMLSS [26]). We preferred GAMLSS models over ordinary least square regression since they allow for fitting trends in mean and variance simultaneously. Models were fitted assuming normal distribution in the dependent variable. For both, mean and variance, two possible responses (none or linear) result in four possible combinations. The null model assuming no effect on either mean or variance was rejected in every case, such that we present the following models for comparison: Trend in mean, but no trend in variance (model m1), trend in mean and variance (m2) and no trend in mean but trend in variance (m4). The best fitting model per year and ratio was selected by AIC (Akaike’s Information Criterion) [27].
The results from MANOVA and GAMLSS were not always congruent for the different bivariate ratios. Generally, the MANOVA detects more significant diversity effects than the GAMLSS, which reflects the fact that the latter tested the direct association between richness and stoichiometry only, whereas the MANOVA extracts additional variation based on block and functional group presence and richness. Thus, in interpreting these outcomes, the MANOVA is the more powerful test for diversity effects on mean ratios, whereas the GAMLSS is the more powerful test for simultaneous changes in the variance.
Additionally, we tested how species diversity affects multi-element stoichiometry and the predictability of chemical composition. For each year, we used a Principal Components Analysis (PCA) on the concentrations of C, N, P, and K, producing two orthogonal axes explaining between 73% and 80% of the total variance in the different years. The first principal component was loaded by C concentration and opposed by P and K concentrations, whereas the second principal component was loaded by N concentrations alone. Factor loadings for PCA analysis (Table S3) were calculated with Statistica 8.0 (Statsoft, Tulsa Oklahoma). We used the origin of the PCA (0;0) as average stoichiometric composition and calculated the stoichiometric distance (SDist) for each of the 82 plots in each year. SDist is a multivariate expression on how deviant the chemical composition of the plant community was from the average across plots and years. An analysis of variance (ANOVA) was performed on SDist as described above.
In order to test whether plant diversity effects on stoichiometry could be explained by mixing plant species only, we compared expected to observed ratios. For the expected ratios, the biomass contribution of each species on the large plots from 2003 to 2006 [23] was multiplied by their element concentrations in monoculture, which were obtained by averaging species replicates from small monoculture plots. Because biomass of the small plots was only analyzed for C and N concentrations from 2003–2006, we lack monoculture information on P-content and thus, the calculation was only done for the C:N ratio in those years. The predicted ratios were plotted against the observed ratios of the years 2003–2006 for plant monocultures and diversity levels of 2, 4, 8 and 16 species mixtures. In addition to perform an ordinary least square regression (OLS), we are assessing the performance of the predicted nutrient ratios by an orthogonal regression (also called ‘total least squares’). In OLS, the resulting coefficient is not independent from the choice of predictor and regressor. Conversely, orthogonal regression makes no assumptions regarding the source of the error. The orthogonal regression was estimated using function ‘princomp’ in R [28]. Confidence intervals for the parameters were estimated from 1000 bootstrap iterations. Additionally, we compared the coefficient of variation (CV) observed in mixtures with a predicted CV derived from species relative cover in mixtures and their monoculture CN ratios.
Plant diversity effects on plant stoichiometry started to become significant after three years (from 2005 onwards) for both species and functional group richness (MANOVA, Table 1). The significance of species richness and functional richness in the model partly depended on the sequence of terms (Table 1, Table S1 and S2), but we still found significant influence of functional diversity after accounting for species richness and marginally significant richness effects after accounting for functional group richness.