Introduction

Nitrogen is the quantitatively most important plant nutrient. The inorganic N species NO\(_{3}^{-}\) and NH\(_{4}^{+}\) are the major sources of N for plants (Hawkesford et al. 2012), but low-molecular dissolved organic N (DON) such as amino acids can also be taken up by plants without previous microbial mineralization (Bardgett et al. 2003; Weigelt et al. 2005). DON also contains recalcitrant components that plants and microorganisms cannot easily utilize (Neff et al. 2003). The processes of the terrestrial N cycle are mainly governed by land management, soil properties, climatic conditions, and deposition from the atmosphere (Stevenson and Cole 1999). Fertilization and deposition from the atmosphere increase nitrogen availability in soil of grassland systems. Land management activities like mowing with subsequent removal of grassland biomass represent the largest loss of N from hay meadows, followed by N leaching from soil (Oelmann et al. 2007a). In N loss studies, the main focus is usually on inorganic N forms, especially nitrate (van Kessel et al. 2009). Not only nitrate, but also the biologically available part of DON can contribute to eutrophication and acidification of surface and groundwater because of later mineralization to inorganic N by microorganisms (Seitzinger and Sanders 1997; van Kessel et al. 2009). As the contribution of DON to total N leaching from agricultural fields can be high, it should be considered as a significant component of N losses (Jones et al. 2004; van Kessel et al. 2009).

Positive plant species richness-productivity relationships in biodiversity experiments (Hector et al. 1999; Tilman et al. 2001; Spehn et al. 2005; Marquard et al. 2009) are most likely caused by “complementarity effects”, which allow for a more exhaustive use of resources like light, nutrients, and water (Tilman et al. 1996; Spehn et al. 2000; Verheyen et al. 2008). “Complementarity effects” summarize processes like niche differentiation and facilitation, which increase the performance of diverse mixtures above the expected performance of monocultures (Loreau et al. 2001). If the probability that dominating species are present increases in species-rich mixtures, “sampling effects” can arise in biodiversity experiments (Loreau et al. 2001). To prevent “sampling effects” and to allow for the detection and investigation of other mechanisms behind biodiversity-ecosystem functioning relationships, biodiversity experiments need to be carefully designed (Loreau et al. 2001; Roscher et al. 2004). Niche differentiation, or resource partitioning, can reduce competition for resources like N between plants (McKane et al. 2002). Therefore, the observed negative species richness effect on NO\(_{3}\)–N leaching is usually explained by complementary, more exhaustive resource use by diverse plant mixtures (Tilman et al. 1996; Scherer-Lorenzen et al. 2003; Dijkstra et al. 2007; Bingham and Biondini 2011; Leimer et al. 2015). Oelmann et al. (2007b) reported that increasing species richness decreased NO\(_{3}\)–N, DON, and TDN concentrations in soil solution in the first year of a large-scale biodiversity experiment in grassland (Jena Experiment) and attributed this to complementary N acquisition by plants in species-rich mixtures. For leaching of NO\(_{3}\)–N, DON, and TDN from soil, this finding could not be confirmed (Oelmann et al. 2007a), which demonstrates that N fluxes, which consist of the components water flux and N concentration, and N concentrations in soil solution can have independent relationships with biodiversity. Studies investigating effects of plant community composition on NH\(_{4}\)–N concentrations and leaching are rare and cover only short time periods because NH\(_{4}\)–N concentrations in soil solution of grasslands were frequently close to or below the detection limit (Hooper and Vitousek 1998; Niklaus et al. 2001; Scherer-Lorenzen et al. 2003; Dijkstra et al. 2007; Oelmann et al. 2007b). In the Jena Experiment, a significant increase in NH\(_{4}\)–N concentrations in solid soil extracts with increasing species richness appeared in the fourth experimental year (Oelmann et al. 2011). Such an increase in NH\(_{4}\)–N availability together with an increase in organic matter concentrations (Steinbeiss et al. 2008) and microbial activity (Eisenhauer et al. 2013) with increasing species richness might enhance ammonification and lead to higher NH\(_{4}\)–N leaching in species-rich mixtures.

Not only species richness, but also the functional identity of plants included in a mixture can affect ecosystem functions because of specific biological traits (Hooper and Vitousek 1998; Niklaus et al. 2001; Scherer-Lorenzen et al. 2003; Jiang et al. 2007; Oelmann et al. 2007b; Mokany et al. 2008). Such traits comprise the ability of N\(_{2}\)-fixation by legumes which increases the N availability for all plants in a mixture and subsequently biomass production (Temperton et al. 2007; Marquard et al. 2009), the dense and extensive rooting system of grasses which allows for a more efficient exploitation of soil-bound resources like N, or the taproot system of most tall herbs which facilitates resource use from deeper soil depths (Hooper and Vitousek 1998; Scherer-Lorenzen et al. 2003; Nippert and Knapp 2007; Oelmann et al. 2007b; Bessler et al. 2012). Accordingly, the presence of legumes tends to increase NO\(_{3}^{-}\) leaching and grasses decrease NO\(_{3}^{-}\) leaching (Hooper and Vitousek 1998; Scherer-Lorenzen et al. 2003; Dijkstra et al. 2007; Leimer et al. 2015). NH\(_{4}\)–N concentrations in solid soil extracts were higher if legumes were present (Hooper and Vitousek 1998). This might increase NH\(_{4}\)–N leaching if legumes are included in a mixture. In the studies of Oelmann et al. (2007a, b), the presence of grasses decreased and the presence of legumes increased TDN concentrations in soil solution in the first year of the Jena Experiment, but not TDN fluxes in soil. As NO\(_{3}\)–N contributions to TDN concentrations and fluxes were high in previous studies, TDN leaching might be similarly affected by species richness and functional group identity as NO\(_{3}\)–N leaching (Hooper and Vitousek 1998; Scherer-Lorenzen et al. 2003; Dijkstra et al. 2007; Oelmann et al. 2007a).

Plant diversity effects on DON leaching have been rarely studied and only for short time periods because only more recent studies considered DON as an important part of the N cycle in grasslands (Bardgett et al. 2003; Jones et al. 2004; Weigelt et al. 2005). Leaching of DON has been reported to be the dominant form of N leaching in diverse plant mixtures (Dijkstra et al. 2007). In the few studies on species richness effects on DON leaching, DON leaching was found to either increase with increasing species richness (Dijkstra et al. 2007) or was not affected by species richness in a one-year study period shortly after land-use change although DON concentrations in soil solution decreased with increasing species richness (Oelmann et al. 2007a, b). More biomass in species-rich mixtures likely causes greater litter production and subsequently greater DON production which might lead to increased DON leaching (Dijkstra et al. 2007). Decreasing DON leaching in soil solution with increasing species richness could have two different explanations (Oelmann et al. 2007b): First, as amino-acids can be taken up by plants directly, different preferences for specific N forms by different plant species could lead to complementary resource use of DON in more diverse plant mixtures (Bardgett et al. 2003; Weigelt et al. 2005) and therefore decrease DON leaching. Second, higher microbial activity in more diverse mixtures (Eisenhauer et al. 2013) increases the use of DON as a C and N source by microorganisms (Bardgett et al. 2003) and might also accelerate the decomposition of organic N compounds by microorganisms and subsequently the transformation of organic N to inorganic N forms (Oelmann et al. 2007b). The presence of grasses decreased and the presence of legumes increased DON concentrations in soil solution in the first year of the Jena Experiment but not DON fluxes in soil (Oelmann et al. 2007a, b). Direct uptake of amino acids by grasses might be the reason for the negative effect on DON concentrations in soil solution (Streeter et al. 2000). The positive effect of legumes on DON concentrations in soil solution might be attributable to lower soil N demand of plants in mixtures containing legumes because of N\(_{2}\) fixation, reducing the need for DON as a N source for plants (Oelmann et al. 2007b). Furthermore, increased aboveground N storage in mixtures with legumes and subsequently higher N input to soil via litter might lead to a positive legumes-DON relationship (Oelmann et al. 2007b, 2011).

When fertilized arable land is converted to unfertilized grassland, the legacy of the previous land use can persist and affect ecosystem variables for several years (Christian and Riche 1998; Schilling and Spooner 2006; Oelmann et al. 2007b; Başaran et al. 2008). In the first year after land-use change from arable land to grassland, NO\(_{3}\)–N leaching is usually high and considerably lower in the following years (Christian and Riche 1998; Oelmann et al. 2011; Leimer et al. 2015). The contribution of DON to TDN concentrations in soil solution increases with time since land-use change because the effect of former fertilization weakens (Oelmann et al. 2007b). Dijkstra et al. (2007) reported that leaching of dissolved inorganic N from monocultures and 16-species mixtures differed even more if inorganic N fertilizer was applied. Accordingly, plant diversity effects on NO\(_{3}\)–N leaching can change with time since conversion from fertilized arable land to unfertilized grassland (Oelmann et al. 2011; Leimer et al. 2015), which might also be possible for NH\(_{4}\)–N, DON, and TDN leaching. Significant species richness effects on soil extractable NH\(_{4}\)–N appeared only in the fourth year after establishment of a biodiversity experiment (Oelmann et al. 2011), which raises the question if species richness effects on NH\(_{4}\)–N leaching might also appear with some delay after grassland establishment. To our knowledge, temporal variations in plant diversity effects on NH\(_{4}\)–N, DON, and TDN leaching have not yet been investigated. Such temporal variations in plant diversity effects caused by former fertilization or by advancing establishment of plant communities call for long-term observations.

Global biodiversity loss is proceeding and its impact on NH\(_{4}\)–N, DON, and TDN leaching is largely unknown, in contrast to NO\(_{3}\)–N leaching. Therefore, our objective was to investigate plant diversity effects on annual NH\(_{4}\)–N, DON, and TDN leaching from an experimental grassland over 4 years. (1) We hypothesize that species richness, functional group richness, and the presence of grasses decrease and the presence of legumes increases NH\(_{4}\)–N and TDN leaching. (2) We expect that DON leaching decreases with increasing species richness because of complementary use of amino acids by plants and stronger use and degradation of DON by stimulated microbial activity in diverse plant mixtures. (3) We further hypothesize that plant diversity effects on NH\(_{4}\)–N, TDN, and DON leaching change with time since land-use change from fertilized arable land to unfertilized grassland because of fading effects of former fertilization and advancing establishment of the plant communities.

Methods

Study site

This study was conducted as part of the Jena Experiment (www.the-jena-experiment.de), a grassland plant diversity experiment that addresses the role of biodiversity for element cycling and trophic interactions (Roscher et al. 2004). The field site is located in Germany close to the city of Jena (50\(^\circ\)55\(^\prime\)N, 11\(^\circ\)35\(^\prime\)E; 130 m above sea level) on the floodplain of the Saale river. The annual mean air temperature is 9.9 \(^{\circ }\)C and mean annual precipitation amounts to 610 mm (1980–2010; Hoffmann et al. 2014). In the studied years, annual precipitation amounted to 470 mm (2003), 570 mm (2004), 420 mm (2005), and 490 mm (2006). The soil developed from up to 2 m thick loamy fluvial sediments, almost free of stones, and can be classified as Eutric Fluvisol. The texture ranges from sandy loam near the river to silty clay with increasing distance from the river because of the fluvial dynamics. The study site was converted from grassland to an arable field in the 1960s and consequently fertilized and plowed for crop production until the beginning of the grassland plant diversity experiment in 2002 (Roscher et al. 2004).

The entire experimental design is described in Roscher et al. (2004). Briefly, the main experiment comprises 82 plots (20 m \(\times\) 20 m). To consider the systematic variation in soil texture in the experimental design, the plots were arranged in four blocks with approximately homogeneous soil texture, parallel to the river Saale. In May 2002, the plots were established from seeds with different levels of plant species richness (1, 2, 4, 8, 16, or 60) and plant functional group richness (1, 2, 3, or 4 of the functional groups: grasses, small herbs, tall herbs, and legumes) in a factorial design. The design of the Jena Experiment allows for controlling the sampling effect. Each block contains four plots with 1, 2, 4, and 8 sown species, respectively (n = 16 per richness level), covering the whole range of number of functional groups. The 16-species mixtures were replicated 3 to 4 times per block (n = 14). One 60-species plot was included in each block (n = 4). The species were chosen by the random replacement method from a species pool of 60 species from the Molinio-Arrhenatheretea meadows, Arrhenatherion community (Ellenberg 1996). The successful establishment of the species richness gradient can be assumed because the number of sown species and the realized species richness correlated strongly (\(R^{2} > 0.9\) in each of the years 2003–2007; Marquard et al. 2009). The management of all plots was adapted to meadows that are managed with low intensity and used for hay production by mowing twice a year (June and September). To maintain the sown species composition, the plots were weeded regularly. The plots were not fertilized during the experimental period. According to data availability, we investigated the 62 plots from blocks 1–3 in this study (12 plots with 1, 2, 4, and 8 plant species, respectively, 11 plots with 16 plant species, and 3 plots with 60 plant species).

Data processing and statistical analyses

Data for NO\(_{3}\)–N, NH\(_{4}\)–N, and TDN concentrations in soil solution were available between January 2003 and December 2006 for the 62 plots from blocks 1, 2, and 3. Sampling and measurement of soil solution is described in Oelmann et al. (2007b). Briefly, soil solution was collected at the 0.3 m soil depth using suction plates with permanent vacuum (UMS, Munich, Germany, sintered glass, diameter 0.12 m, pore size 1—1.6 \(\mu\)m) to ensure a continuous collection of soil solution. To prevent changes in the natural flow field in direct vicinity of the suction plates, which can cause differences in solute concentrations (Weihermüller et al. 2007), the applied vacuum was chosen similar to the ambient soil matric potential and regularly adjusted. We preferred suction plates over zero-tension lysimeters because suction plates do not disturb the aboveground management, as do the frames of zero-tension lysimeters, and cause less structural disturbance when installed. Nevertheless, suction plates yield a similar solution quantity and quality. In zero-tension lysimeters, in contrast to suction plates, soil saturation and exposition of the soil to atmospheric conditions at the upper lysimeter boundary can alter solute concentrations (Weihermüller et al. 2007). Suction plates do not allow for measurement of drainage volumes unless the vacuum exactly compensates the current soil matric potential at all times, which is difficult to achieve. In zero-tension lysimeters, the measurement of drainage volumes is also difficult because of edge-flow effects (Weihermüller et al. 2007). Samples were collected every second week and subsequently stored frozen until the analysis to prevent transformations by biological processes. Concentrations of NO\(_{3}\)–N, NH\(_{4}\)–N, and TDN (after UV digestion) were measured photometrically with a Continuous Flow Analyzer (CFA). Nitrite-N concentrations in soil solution were expected to be small and included in the NO\(_{3}\)–N concentrations because of the analysis method. Not detected NO\(_{3}\)–N, NH\(_{4}\)–N, and TDN concentration values were replaced by half the detection limit of the CFA for NO\(_{3}\)–N, NH\(_{4}\)–N, and TDN, i.e. 0.01, 0.015, and 0.01 mg l\(^{-1}\), respectively (Oelmann et al. 2007b). Concentrations of NO\(_{3}\)–N, NH\(_{4}\)–N, and TDN higher than \(mean + 2 \cdot standard \,deviation\) were assumed to be outliers and set to not available (this was the case for 48 of 1972 available fortnightly NO\(_{3}\)–N concentration values, 6 of 1901 NH\(_{4}\)–N values and 40 of 1951 TDN values). DON concentrations in soil solution were calculated by subtracting NO\(_{3}\)–N and NH\(_{4}\)–N from TDN concentrations. Calculated DON concentrations lower than −0.02 mg l\(^{-1}\) were regarded as measurement errors and removed from the data set (72 of 1822 available fortnightly DON concentrations). Calculated DON concentrations between −0.02 and 0 mg l\(^{-1}\) were set to 0 mg l\(^{-1}\) (30 of 1822 available fortnightly DON concentrations) because the lowest detection limit of NO\(_{3}\)–N, NH\(_{4}\)–N, and TDN concentrations was 0.02 mg l\(^{-1}\) and could have caused this error in the calculated DON concentrations. If the concentration of at least one N species was not available for a sampling date (after the described procedure), this sampling date was removed from the data set for all N species (resulting in 1750 valid fortnightly concentration values per N species). Subsequently, monthly mean NO\(_{3}\)–N, NH\(_{4}\)–N, DON, and TDN concentrations in soil solution were calculated (Fig. 1c).

Fig. 1
figure 1

Temporal course of monthly a precipitation, b downward water flux, and c NO\(_{3}\)–N, NH\(_{4}\)–N, dissolved organic N (DON), and total dissolved N (TDN) concentrations and d leaching. Points and error bars indicate mean and standard error of b all 62 plots and c, d plots for which data was available (number given at the bottom of d per timestep)

Full size image

Downward water fluxes [mm = L m\(^{-2}\)] per plot from the 0–0.3 m soil layer were simulated for the years 2003–2006 in weekly resolution with a water balance model (Leimer et al. 2014a). The model allows for calculation of downward water fluxes (DF), actual evapotranspiration (ETa), and capillary rise (CR) using the input variables precipitation (P), potential evapotranspiration, and volumetric soil water content (S). It has been successfully applied for our study site (Oelmann et al. 2007a; Leimer et al. 2014a). The model is based on the water balance equation (Eq. 1) where \(\Delta S\) is the change in volumetric soil water content between two subsequent observation dates (\(\Delta S = S_{t_{1}} - S_{t_{2}}\)).

$$\begin{aligned} P + CR = DF + ETa - \Delta S \end{aligned}$$
(1)

Meteorological data were obtained in high resolution from a central field station. Potential evapotranspiration was calculated from meteorological data using the Penman-Wendling equation (DVWK 1996). 9240 volumetric soil water content measurements were performed on the experimental plots between July 2002 and January 2006. To obtain a complete and equidistant soil water content data set, despite data gaps and variations in sampling intervals in the plot-scale soil water content data set, a Bayesian hierarchical linear regression model using soil water content measurements from the central meteorological station as explanatory variable was applied. The Bayesian model for soil water contents in the 0–0.3 m soil layer performed well with a Bayesian coefficient of determination R\(_{B}^{2}\) of 0.68. Subsequently, the complete soil water content data set was used as input for the water balance model to obtain downward water fluxes. Finally, downward water fluxes were aggregated to monthly sums per plot (Fig. 1b). A detailed description of the simulation of downward water fluxes, including simulation of plot-scale soil water contents, is given in Leimer et al. (2014a).

To cope with data gaps in the monthly NO\(_{3}\)–N, NH\(_{4}\)–N, DON, and TDN concentrations in the soil solution data set, we calculated annual flux-weighted mean concentrations of NO\(_{3}\)–N, NH\(_{4}\)–N, DON, and TDN on the basis of the monthly downward water flux data set. Subsequently, the annual flux-weighted mean concentrations in soil solution were multiplied with the annual sum of downward water flux to obtain annual NO\(_{3}\)–N, NH\(_{4}\)–N, DON, and TDN leaching per plot from the 0–0.3 m soil layer, which can be regarded as the main rooting zone (Bessler et al. 2009).

In May 2006, soil samples for the measurement of soil microbial biomass were taken with a steel corer (5 cores per plot, depth 5 cm, diameter 5 cm) and sieved (2 mm; Anderson and Domsch 1978). Microbial biomass C of approximately 5 g soil (fresh weight) was measured using an O\(_{2}\)-microcompensation apparatus (Scheu 1992). Substrate-induced respiration was calculated from the respiratory response to D-glucose for 10 h at 22\(^{\circ }\)C (Anderson and Domsch 1978). Glucose was added according to preliminary studies to saturate the catabolic enzymes of microorganisms (4 mg g\(^{-1}\) dry weight solved in 400 \(\mu\)l deionized water). The mean of the lowest three readings within the first 10 h was taken as maximum initial respiratory response (MIRR; \(\mu\)l O\(_{2}\) h\(^{-1}\) g\(^{-1}\) soil dry weight) and microbial biomass (\(\mu\)g C g\(^{-1}\) soil dry weight) was calculated as 38 x MIRR (Eisenhauer et al. 2010).

In May and August 2006, aboveground plant biomass was harvested on all plots within a frame (0.2 \(\times\) 0.5 m, height 0.03 m) at four randomly located sites per plot. Plant biomass of mixtures was separated into species. After ovendrying (70\(^\circ\)C) to constant weight, the samples were weighed. N concentrations in aboveground biomass were determined from pooled samples of living plant material per plot which were ground with a Cyclotec 1093 Sample Mill (Foss Tecator, Hoganas, Sweden). Twenty mg of the ground plant material was analyzed for plant N concentration with an elemental analyzer CE 1110 (Carlo Erba Instruments, Milan, Italy). The amounts of N harvested from the plant communities in May and August 2006 were then calculated using aboveground biomass and N concentrations (Oelmann et al. 2011). To obtain the annual N harvest per plot, which can be regarded as a surrogate for annual N uptake by plant communities, data from May and August were summed.

Annual NO\(_{3}\)–N, NH\(_{4}\)–N, DON, and TDN leaching from the 0–0.3 m soil layer of the 62 plots in 2003–2006 was analyzed for plant diversity effects using linear mixed-effect models. Block and plot identity were treated as random factors in a nested sequence. The analysis was started with a constant null model and the fixed terms were fitted in a sequence following the hypotheses of the biodiversity experiment and previous findings. Year (factor) was fitted first, followed by log-transformed species richness (continuous), presence of legumes, presence of grasses, presence of tall herbs, and finally presence of small herbs. N leaching was expected to vary strongly with time since conversion from fertilized arable land to unfertilized grassland and therefore time was fitted first. The order of the functional groups was chosen because legumes have a strong effect on the N cycle as a consequence of their N\(_{2}\)-fixing ability and legumes increase downward water fluxes from the 0–0.3 m soil layer, grasses decrease N concentrations in soil solution and downward water fluxes, tall herbs decrease downward water fluxes in the 0.3-0.7 m soil layer, and small herbs frequently show the smallest or no effect. Furthermore, the interactions between year (factor) and the previously mentioned treatment variables were added to the linear mixed-effect model to analyze for possible temporal changes in diversity and compositional effects. In an additional linear mixed-effect model, we investigated if functional group richness (continuous) affects NO\(_{3}\)–N, NH\(_{4}\)–N, DON, or TDN leaching by fitting the fixed terms in the order: year, log-transformed species richness, functional group richness, and the interactions between year and each of the two treatment variables. The maximum likelihood method and likelihood ratio tests were used to assess the statistical significance of model improvement. To meet the prerequisites of the statistical analyses, NO\(_{3}\)–N, NH\(_{4}\)–N, DON, and TDN leaching had to be log-transformed.

We calculated the Pearson correlation for DON leaching versus species richness (both log-transformed) in each year to investigate the reason for the significant effect of the interaction between year and species richness on DON leaching. Structural equation modeling (SEM) was performed to test the hypotheses that (log-transformed) species richness effects on DON leaching in 2006 can be explained by soil microbial biomass or by the amount of N harvested that depends on plant N uptake and would indicate complementary use of DON by the plant communities. SEM allows testing direct and indirect relationships between variables in a multivariate approach (Grace 2006). The adequacy of the model was determined via \(\chi ^{2}\) tests, Aikaike information criterion (AIC), and root mean square error of approximation (RMSEA).

Calculations and statistical analyses were done with the R 3.1.0 software package (R Core Team 2014), the linear mixed–effect modeling with the function lme() in the package nlme (Pinheiro et al. 2014), and Pearson’s correlation with the function cor.test(). SEM was performed with IBM SPSS Amos 21.0.0 (Amos Development Corporation; Meadville, PA,USA).

Results

Leaching of all N species varied significantly between the years (Table 1). Mean NH\(_{4}\)–N leaching of all plots was lowest in the first (0.06 kg ha\(^{-1}\)) and highest in the fourth year (0.10 kg ha\(^{-1}\)) after land-use change from fertilized arable land to unfertilized grassland. As similarly reported for NO\(_{3}\)–N leaching (Leimer et al. 2015), mean leaching of TDN was highest in the first year, lowest in the third year, and increased again in the fourth year (Fig. 2a, b). Mean leaching of DON varied little between the first two years (2003: 2.05 kg ha\(^{-1}\), 2004: 2.06 kg ha\(^{-1}\)), was lower in the third year (0.66 kg ha\(^{-1}\)), and increased again in the fourth year (1.44 kg ha\(^{-1}\)). Dissolved inorganic N leaching in this grassland was dominated by NO\(_{3}\)–N (Fig. 2c, d). NO\(_{3}\)–N contributed, averaged over the annual mean contributions, 31% (±SD 5% over time) and NH\(_{4}\)–N 4% (±3%) to TDN leaching. The DON contribution to TDN leaching, averaged over the annual mean contributions, was 64% (± 4%; Fig. 2c, d).

Table 1 Results of the linear mixed-effects model analysis of log-transformed annual NO\(_{3}\)–N, NH\(_{4}\)–N, DON, and TDN leaching [mg m\(^{-2}\)] from the 0–0.3 m soil layer in 2003–2006
Full size table
Fig. 2
figure 2

Mean annual leaching of total dissolved N (TDN) from plots a with and b without legumes. Mean contribution of NO\(_{3}\)–N, NH\(_{4}\)–N, and dissolved organic N (DON) to TDN leaching from plots c with and d without legumes

Full size image

Leaching of NH\(_{4}\)–N was not significantly influenced by any of the plant community-treatment variables (Table 1). The interaction between year and species richness had a significant effect on DON leaching (Table 1). In 2003–2005, there was no significant correlation between DON leaching and species richness. Only in 2006, DON leaching significantly decreased with increasing species richness (r = −0.28, p = 0.028; Fig. 3a). Species richness decreased TDN leaching significantly (Table 1; Fig. 4). Adding functional group richness to the linear mixed-effects model after year and species richness did not show a significant effect of functional group richness on NH\(_{4}\)–N, DON, or TDN leaching. The presence of legumes significantly increased DON leaching (Table 1). The presence of grasses and the presence of small herbs decreased TDN leaching whereas the presence of legumes increased TDN leaching significantly (Table 1; Fig. 2). Apart from the significant interaction between year and species richness for DON leaching, none of the interaction terms was significant for NH\(_{4}\)–N, DON, or TDN leaching (Table 1). The same statistical results as for annual TDN leaching in this study were found by Leimer et al. (2015) for monthly NO\(_{3}\)–N leaching.

Fig. 3
figure 3

Dissolved organic N (DON) leaching versus a species richness level (both axis log-scaled; Pearson’s correlation: r = −0.28, p = 0.028), c microbial biomass (r = −0.35, p = 0.007) and d N harvest (r = 0.33, p = 0.011) in 2006. b Microbial biomass per species richness level in 2006 (r = 0.47, p < 0.001). Trend lines are shown for illustration purpose only

Full size image
Fig. 4
figure 4

Mean total dissolved N (TDN) leaching per year and species richness level. Error bars indicate standard error

Full size image

According to a \(\chi ^{2}\) test, the data on DON leaching in 2006 did not significantly deviate from the structural equation model (\(\chi ^{2} = 1.7\), \(p = 0.43\)). The SEM indicated that species richness significantly increased microbial biomass (Fig. 3b) and microbial biomass significantly decreased DON leaching in 2006 (Table 2; Figs. 3c, 5). Furthermore, species richness significantly increased the amount of harvested N and harvested N significantly increased DON leaching in 2006 (Table 2; Figs. 3d, 5). A direct path from species richness to DON leaching was not significant, indicating that the significant decrease in DON leaching with increasing species richness in 2006 can be explained through variations in soil microbial biomass and that the negative influence of soil microbial biomass overly compensated the positive influence of increased N harvest on DON leaching with increasing species richness (Fig. 5).

Table 2 Unstandardized path coefficients of the structural equation model (SEM) of the species richness effect on microbial biomass and N harvest and subsequently on dissolved organic N (DON) leaching in 2006
Full size table
Fig. 5
figure 5

Structural equation model (SEM) of the species richness effect on microbial biomass and N harvest and subsequently on DON leaching in 2006. Arrows indicate significant relations. Numbers on arrows indicate standardized path coefficients (for unstandardized path coefficients see Table 2) and percentages indicate the variance explained by the model

Full size image

Discussion

Species richness and functional group effects on NH\(_{4}\)–N, DON, and TDN leaching

Leaching of NH\(_{4}\)–N played a minor role in this grassland because dissolved inorganic N leaching was dominated by NO\(_{3}\)–N (Fig. 2). The absence of significant species richness and functional group effects on NH\(_{4}\)–N leaching is related with the low NH\(_{4}\)–N availability for plants as reflected by the low NH\(_{4}\)–N concentrations in soil solution. The low NH\(_{4}\)–N availability is attributable to rapid microbial immobilization and nitrification masking potential plant diversity effects. The finding of Hooper and Vitousek (1998), who reported higher NH\(_{4}\)–N concentrations in solid soil extracts if legumes were present, was not reflected by NH\(_{4}\)–N leaching in our study (Table 1).

Dissolved organic N leaching contributed most of all considered N species to TDN leaching, which complies with the findings of Dijkstra et al. (2007) and Oelmann et al. (2007a). Our analysis of the data until the year 2006 revealed that species richness significantly decreased DON leaching in 2006 for the first time. This is in line with findings of Oelmann et al. (2007a, b) that DON concentrations in soil solution were lower in species-rich mixtures between spring 2003 and spring 2004, however not DON leaching. The result of Dijkstra et al. (2007) that DON leaching is higher in 16-species mixtures than in monocultures could not be confirmed in our study. This disagreement might be caused by the large differences in soil properties between the study of Dijkstra et al. (2007) (nutrient-poor sandy outwash plain) and the Jena Experiment (nutrient-rich loamy floodplain). Soil properties can modify DON concentrations, downward water fluxes, and DON leaching. In coarse textured soils, downward water fluxes are usually larger than in fine textured soils and DON loss likely increases with sand content (van Kessel et al. 2009). Besides the differences in soil properties, more litter and subsequently more DON production and less biodegradation of DON in 16-species mixtures compared to monocultures might have caused the increasing DON leaching with increasing species richness in the study of Dijkstra et al. (2007). In the Jena Experiment, a positive relationship between ammonification and species richness was found in an in-situ incubation experiment (Rosenkranz et al. 2012), which might, in contrast to the study of Dijkstra et al. (2007), lead to decreasing DON leaching in diverse mixtures. The reasons for lower DON leaching in species-rich mixtures in our study are discussed in detail in the section “Why does species richness decrease DON leaching?”. Functional group richness did not significantly influence DON leaching, which is in line with the results of Oelmann et al. (2007a) for April 2003 to March 2004 in the Jena Experiment. The positive effect of the presence of legumes on DON leaching (Table 1) is in accordance with previous findings regarding DON concentrations in soil solution (Oelmann et al. 2007b). The positive legume effect on DON leaching could be explained by the fact that plots with legumes produced significantly more aboveground biomass (Marquard et al. 2009) and had a higher aboveground N storage, which even strengthened with time (Oelmann et al. 2011). More biomass implies more litter production (i.e., more dead organic material that is not removed during harvest, but reaches the soil) and subsequently more DON production which can lead to increased DON leaching (Dijkstra et al. 2007). Lower net ammonification rates in mixtures with legumes might have further enhanced DON leaching in legumes-containing mixtures (Rosenkranz et al. 2012).

Leaching of TDN decreased with increasing species richness (Fig. 4) likely because of the high contribution of NO\(_{3}\)–N to TDN leaching and the negative relationship between species richness and NO\(_{3}\)–N leaching (Table 1; Scherer-Lorenzen et al. 2003; Bingham and Biondini 2011; Leimer et al. 2015). Complementary and therefore more exhaustive use of resources like NO\(_{3}\)–N by plants is usually suggested as reason for lower NO\(_{3}\)–N concentrations and leaching in species-rich mixtures (Scherer-Lorenzen et al. 2003; Oelmann et al. 2007b; Roscher et al. 2008). Another possible explanation would be that the dominance of productive species and facilitation by legumes leads to increased N uptake in species-rich mixtures (Bessler et al. 2012). The negative species richness effect on DON leaching in 2006, together with the high contribution of DON to TDN leaching, possibly also contributed to decreasing TDN leaching with increasing species richness. Functional group richness did not significantly influence TDN leaching, which is in line with the results of Oelmann et al. (2007a) for NO\(_{3}\)–N and TDN and with Leimer et al. (2015) for NO\(_{3}\)–N leaching. The finding that the presence of legumes increases TDN leaching (Table 1; Fig. 2a, b) is probably related with the strong positive effect of legumes on its components DON leaching as described above and NO\(_{3}\)–N leaching (Table 1; Leimer et al. 2015). NO\(_{3}\)–N leaching increases if legumes are present because legumes have the ability to fix N\(_{2}\) from the atmosphere (Scherer-Lorenzen et al. 2003; Leimer et al. 2015). Additionally, downward water fluxes from the 0–0.3 m soil layer are higher if legumes are present in a plant mixture (Leimer et al. 2014a), which further contributes to increased TDN leaching. The presence of grasses decreased TDN leaching (Table 1) likely because of the dense and extensive rooting system of grasses which allows for a more exhaustive use of resources (Scherer-Lorenzen et al. 2003). Grasses have been shown to reduce both resources that contribute to TDN leaching, N and especially NO\(_{3}\)–N concentrations in soil solution (Hooper and Vitousek 1998; Scherer-Lorenzen et al. 2003; Oelmann et al. 2007b; Leimer et al. 2014b) and soil water content and subsequently downward water fluxes (Leimer et al. 2014a). The lower TDN leaching in mixtures in which small herbs were present (Table 1) is not reflected in soil solution (Oelmann et al. 2007b). Therefore, the small herb effect on TDN leaching might be solely attributable to the decreasing effect of small herbs on downward water fluxes (Leimer et al. 2014a).

Why does species richness decrease DON leaching?

The results of the SEM suggest that, in contrast to the hypothesis, DON leaching in 2006 increased with increasing N harvest (Fig. 3d) which was positively related with species richness (Fig. 5). This finding suggests that complementary resource use of plant-available components of DON (Bardgett et al. 2003; Weigelt et al. 2005), which would lead to decreasing DON leaching with increasing N harvest, plays no or only a minor role for DON leaching in the studied grassland. The positive species richness effect on DON leaching via N harvest can be explained by higher aboveground (from the first year on; Marquard et al. 2009) and belowground (from the fourth year on; Ravenek et al. 2014) biomass production of species-rich mixtures that causes more above- and belowground litter reaching the soil and subsequently a higher DON production (Dijkstra et al. 2007).

As a direct path from species richness to DON leaching was not significant and the path via N harvest showed a positive correlation with DON leaching, the decrease in DON leaching with increasing species richness in 2006 can only be explained by enhanced biodegradation because of elevated microbial activity (Fig. 5). Positive effects of species richness on soil microorganisms became visible in the fourth year after the establishment of the Jena Experiment (2006) and were consistently present in the following years (Eisenhauer et al. 2010, 2011, 2012; Strecker et al. 2016), which explains why species richness decreased DON leaching only in 2006. The mechanism behind the microbial biomass-DON leaching relationship is probably a higher microbial use of DON as a C and N source with increasing microbial biomass (Bardgett et al. 2003). Higher microbial biomass might also increase the decomposition of organic N compounds by microorganisms and subsequently the transformation of organic N to inorganic N forms (Oelmann et al. 2007b). Accordingly, Rosenkranz et al. (2012) reported a positive correlation between species richness and in-situ net ammonification rates and Oelmann et al. (2011) between plant species richness and KCl-extractable NH\(_{4}\)–N concentrations in the fourth year of the Jena Experiment. In both studies, the observed results were attributed to increased microbial activity with increasing species richness resulting in accelerated organic matter decomposition in species-rich mixtures. A weaker effect of species richness on DON leaching via microbial biomass than via N harvest might have resulted in a positive species richness effect on DON leaching in the study of Dijkstra et al. (2007). At our study site in 2006, the negative effect of species richness via microbial biomass on DON leaching exceeded the positive effect of species richness via the amount of harvested N on DON leaching (Fig. 5).

Temporal variations

The high TDN leaching in 2003 (Fig. 2a, b) can be attributed to former fertilization with mineral N which strongly affected NO\(_{3}\)–N leaching in the same direction (Leimer et al. 2015). In 2004, TDN leaching slightly decreased (Fig. 2a, b), although annual precipitation was highest in this year. The considerably lower TDN leaching in 2005 is likely related to the lowest annual precipitation and lowest annual downward water fluxes in this year compared to the other studied years (Fig. 1; Leimer et al. 2014a). In 2006, TDN leaching increased again, while annual precipitation was similar to the year 2003 and lower than in 2004. The temporal course of precipitation cannot solely explain TDN leaching over time and the increase in TDN leaching in 2006 might also be related to a similar temporal course of soil organic matter concentrations (Steinbeiss et al. 2008). Higher soil organic matter concentrations in 2006 probably increased DON leaching, but also contributed to enhanced NH\(_{4}\)–N release by mineralization (Oelmann et al. 2011). Except for the species richness effect on DON leaching, no temporal variations in the effects of species richness, functional group identity, or functional group richness on annual N leaching rates were observed (Table 1). The evaluation of a simulated monthly NO\(_{3}\)–N leaching data set, based on the available measurements from the study site, revealed seasonal variations in species richness and functional group identity effects on NO\(_{3}\)–N leaching (Leimer et al. 2015). Unfortunately, only a statistical evaluation of annual NH\(_{4}\)–N, DON, and TDN leaching was possible in this study because of data gaps. Like the positive effect of species richness on soil organisms (Eisenhauer et al. 2011, 2012), the effect of species richness on DON leaching became visible in the fourth year after the establishment of the Jena Experiment. The fact that DON contributes most to TDN leaching increases the importance of this finding in terms of resource and environmental protection. Furthermore, the appearance of such delayed species richness effects calls for long-term observations in biodiversity experiments.

Conclusions

Leaching of NH\(_{4}\)–N played a minor role in this grassland and was not influenced by plant diversity. Species richness and functional group identity, but not functional group richness, contributed significantly to the explanation of variations in TDN leaching. Species richness decreased DON leaching only in 2006, probably because of a delayed soil biota effect that increased microbial use and mineralization of organic N in species-rich communities. There was no indication of complementary use of DON by plants. Our results demonstrate that plant species richness reduces TDN leaching during the first four years and DON leaching in the fourth year after land-use change from fertilized arable land to unfertilized grassland.