Introduction

N is a critical element limiting plant growth in temperate ecosystems (Vitousek and Howarth 1991). Consequently, N dynamics such as N mineralization or plant N uptake determine much of the structure and function of these ecosystems. The assessment of N dynamics is therefore critical for monitoring, understanding and predicting essential functions (biogeochemical cycles and productivity) or structural components (species composition and diversity) of temperate ecosystems. In this respect, the natural abundance of stable 15N isotopes (δ15N) in soils and in particular in plants has been suggested as an efficient and simple non-invasive tool to assess and monitor ecosystem N dynamics (Handley and Raven 1992; Nadelhoffer and Fry 1994; Högberg 1997; Handley et al. 1998; Evans 2001; Robinson 2001; Amundson et al. 2003; Pardo et al. 2006; Houlton et al. 2007).

Variability in plant and soil δ15N values derives primarily from physiological and biogeochemical processes in the N cycle. Soil processes such as N mineralization, nitrification, denitrification or NH3 volatilization discriminate against 15N and lead to soil N pools with different δ15N signatures (Mariotti et al. 1981; Handley and Raven 1992; Nadelhoffer and Fry 1994; Piccolo et al. 1994; Robinson 2001). The δ15N signatures of different soils N pools are further imprinted in the δ15N values of plants that utilize these soil N pools for their N nutrition.

Based on the link between foliar and soil δ15N signals, foliar δ15N values can be used as simple but valuable tools to study ecosystem N dynamics, both as tracers or as integrative signals. For example, foliar δ15N values that reflect the δ15N signatures of the plant’s specific N sources (e.g. NO −3 and NH +4 ) could reveal information on the plant’s N uptake patterns (Houlton et al. 2007). Alternatively, foliar δ15N values have the potential to characterize N turnover in the soil as integrative proxies. High soil N mineralization rates are often correlated with denitrification or leaching of inorganic N forms. Both processes lead to losses of 15N-depleted N2O and N2 or NO −3 , respectively and leave the remaining inorganic N pool enriched in 15N. Foliar δ15N signals that reflect the 15N enrichment of the inorganic soil pool may therefore indicate the magnitude of N fluxes and N losses in ecosystems (Högberg 1990; Garten and van Miegroet 1994; Austin and Vitousek 1998; Emmett et al. 1998; Korontzi et al. 2000; Pardo et al. 2002; Amundson et al. 2003; Pardo et al. 2006; Templer et al. 2007).

Despite the potential to study ecosystem N dynamics, foliar δ15N signatures can be obscured by a variety of factors such as discrimination during N uptake, mycorrhizal status and type, nodulation, and intra-plant isotope partitioning (Handley et al. 1998). As a result, the variability of foliar δ15N values among different non-N2 fixing plant species within an ecosystem remains unclear and complicates the use of plant tissue δ15N to investigate plant N uptake patterns or soil N cycling. For example, experimental evidence that relates variability in foliar δ15N values of different species to differences in the plants’ N sources comes from only two laboratory studies (Miller and Bowman 2002; Falkengren-Grerup et al. 2004). While both studies give credible evidence for a positive relationship between NH +4 uptake and foliar δ15N values, it remains unclear if the direction of this relationship is a universal pattern that can be used in the interpretation of foliar δ15N values in different natural ecosystems.

Given the uncertainties regarding foliar δ15N values, the goal of the study presented here was to mechanistically address the variability of foliar δ15N values among a large number of different plant species within an ecosystem and to test if differences in foliar δ15N values of different plant species are persistent across a range of 18 different grassland ecosystems. Specifically, our study had three objectives: first, we characterized the relationship between soil net mineralization rates and variability of foliar δ15N values from different plant species within and across 18 different grassland sites; second, we determined if a plant’s preference for NO +3 or NH −4 uptake explains the variability in foliar δ15N values among different plant species within grassland sites; and third, we tested edaphic and biotic parameters as drivers of species- or functional group-specific variability in foliar δ15N across 18 grassland sites with different N dynamics.

Materials and methods

Study sites

The study was conducted in the Thüringer Schiefergebirge, Germany (11°00′−11°37′E and 50°21′−50°34′N), using 18 montane grasslands that differed in plant species composition and plant diversity (Table 1). Present-day soils in the area have developed from a carbonate-free, nutrient-poor schist and greywacke bedrock material. All sites were located within an area of 20 by 40 km and were similar in elevation above sea level and exposition. For a detailed description of the sites see Kahmen et al. (2005b). The 18 sites were selected from 78 intensively surveyed grasslands (Kahmen et al. 2005a) based on the following selection criteria: sites had to be ungrazed, cut twice a year (late June and August/September) and must not have received any fertilizer applications for the last 10 years prior to sampling.

Table 1 Means and SDs of edaphic site variablesa for the 18 investigated temperate grasslands. NA Not available
Full size table

Edaphic variables

Soil variables were extensively surveyed 6 times throughout 2002. For each sample, four soil cores (4.3 cm diameter, 10 cm length) were collected at each site, pooled to a single sample and sieved to 2 mm. To determine NO − 3 and NH +4 pools in the soils, one part of each soil sample (∼10 g) was extracted with 50 ml of 1M KCl for 60 min on the same day of sampling. KCl extracts were filtered and then frozen at −20°C and later analyzed using a continuous flow analyzer (SAN Plus; Skalar, Erkelenz, Germany) for NO − 3 and NH +4 concentrations. The remaining soil was dried at 35°C and extracted for 1 h using a 1M calcium acetate-lactate (CAL) solution. CAL extracts were analyzed with ICP-AES (Optima 3300 DV; Perkin-Elmer, Norwalk, USA) for P and K concentrations. Soil pH was measured in water suspensions. For the determination of soil C:N ratios and total soil N (Ntot) and C (Ctot) concentrations, dry soil was ground and analyzed with an elemental analyzer (Vario EL II; Elementar, Hanau, Germany). The same dried and ground material was then used for δ15N analyses of bulk soil with an isotope ratio mass spectrometer (IRMS) (Delta C Finnigan MAT; Bremen, Germany). For statistical analyses, soil variables sampled throughout 2002 were averaged for each plot to a single mean value.

Daily net mineralization rates at each site, i.e., the daily production of plant available NO − 3 and NH +4 , were determined twice in 2003 (mid May to mid June and mid July to mid August) by in situ incubation (Hart et al. 1994). Four subsamples were collected at each site, pooled, and NO − 3 and NH +4 concentrations determined as described above. The remaining pooled soil sample of each site was split into four equal parts, sealed in polyethylene bags and incubated in the soil for 24 or 16 days during the spring and summer campaigns, respectively. After incubation, bags were collected, and soil extracted with 1M KCl and analyzed for NO − 3 and NH +4 concentrations as described above. To calculate daily net mineralization rates, NO − 3 and NH +4 concentrations at the beginning of the incubation were subtracted from the concentrations determined at the end of the incubation, and the difference divided by the number of incubation days. Daily net mineralization rates from both incubation periods were averaged for the statistical analyses as mean daily net mineralization. Mineralization rates and concentrations were calculated on a soil dry weight basis.

Foliar δ15N

At each site, we collected leaf material from different non-N2 fixing plant species for δ15N analyses in mid June 2003. The number of plant species sampled at each site was proportional to the species richness in each plot and covered two thirds of the species at a site. Plant material was dried, ground and analyzed for δ15N with an IRMS (Delta C Finnigan MAT). In total, 20 different herbaceous plant species were sampled in the 18 sites. For a more detailed analysis of plant trait effects on foliar δ15N signatures we separated the investigated plant species into the functional groups “grasses” and “forbs”. The functional group grasses contained all species belonging to the family Poaceae, while the functional group forbs contained all other species. Species names and functional group identity are listed in Fig. 4.

N uptake

To test the effect of N nutrition on leaf δ15N, NO − 3 and NH +4 uptake was determined for 15 plant species (seven forbs and eight grasses) in three of the 18 sites (site 2, site 4 and site 7) using 15N labeled N tracers (see Kahmen et al. 2006). At each site, 15N-labeled NO − 3 and NH +4 was injected at two different soil depths (3 and 8 cm) twice a year (spring, 22–24 May 2003; summer, 4–6 August 2003), each treatment in a separate 1-m2 plot. During each campaign in spring and summer, one grassland was labeled per day. The 15N-labeled NO − 3 and NH +4 compounds were injected (using spine syringes) as separate solutions of 9.66 mM K15NO3 and 9.66 mM 15NH4Cl (>98.9 at% 15N). In order to offer both N compounds in each treatment, we added equal amounts (9.66 mM) 14N of the non-treatment N compound to the 15N solution. Injection points in each treatment were distributed evenly across the 1-m2 plots using a 6.5 × 7.0 cm grid, resulting in 210 injections per plot. Assuming a 2-cm diffusion radius, we injected 2 ml of 15N labeled solution at each injection point, leading to a total of 55.35 mg added 15N m−2. As a result of the large experimental effort, sampling of soil and plant materials was nested within the labeled plots at the three grassland sites. To avoid contamination, the 1-m2 plots where tracers were applied, were located in secure distance to the 5 × 5 m sampling areas where plant leaves were collected for δ15N natural abundance analyses.

Three days after injecting the 15N-labeled N compounds into the soil, the vegetation in each treatment was clipped 2 cm above the ground, sorted to species, dried at 70°C for 48 h and weighed. From the sorted biomass, ten individuals from each species were combined to a single bulk sample. Plant material was ground and analyzed for N concentrations and δ15N values with an elemental analyzer (Vario EL II, Elementar, Hanau, Germany) coupled to an IRMS (Delta C Finnigan MAT).

The recovered 15N tracer in the leaves of the different plant species was used to calculate mean daily NO − 3 and NH +4 uptake of the different species in the individual treatments. In the calculations, background natural abundance 15N and N concentration in the leaves of the respective plants as well as plant available N in the soil were taken into account. For a detailed description of the calculations see Kahmen et al. (2006). For the statistical analyses, NO − 3 and NH +4 uptake was summed up for the two soil depths and averaged over the two temporal treatments.

Data analysis and statistics

To correct leaf δ15N values for site-specific differences in background bulk soil δ15N, Δδ15N was calculated for each species from all sites as the difference between δ15Nplant and δ15Nsoil (Amundson et al. 2003). Foliar Δδ15N values represent the 15N depletion of a plant leaf compared to the soil δ15N background. We tested the effects of NO − 3 and NH +4 uptake on Δδ15N of different species in three sites with ANOVA models (type I SS). The ANOVA models allowed to control for site-specific effects of plant available soil N on foliar Δδ15N across these three sites. Further the ANOVA models allowed us to test if the functional group identity of a species had an effect on the relationship between NO − 3 and NH +4 uptake and foliar Δδ15N (i.e., a significant interaction effect between functional group and N uptake on Δδ15N). In each model, mean daily soil N fluxes were entered first as the block factor (i.e., NO − 3 flux, NH +4 flux and NO − 3 /NH +4 flux in the NO − 3 uptake model, NH +4 uptake model and NO − 3 /NH +4 uptake model, respectively), followed by a fixed hierarchical sequence of functional group as factor, N uptake as covariate, and finally the interaction between functional group and N uptake (Table 2). We also used an ANOVA model (type I SS) to test functional group and species-specific effects as well as effects of the plant’s relative abundance on the relationship between net N mineralization and foliar Δδ15N across all 18 sites. As above, this model allowed testing for direct effects of individual variables and also for interacting effects of variables. Factors (functional group and species identity) and covariates (abundance and net N mineralization) were entered into the model in a fixed hierarchical sequence (Table 3). All ANOVA models were calculated using SPSS 11 for MAC OS X 10.3.

Table 2 15N enrichment of plants compared to soil background values (Δδ15Na) as a function of plant NO − 3 and NH +4 uptake and functional (Funct.) group effects as tested in ANOVAsb (type I SS). The variables for the models were assessed in three sites (site 2, site 4 and site 7)
Full size table
Table 3 ANOVAa (type I SS) for plant 15N enrichment compared to soil background values (Δδ15Nb) for 20 herb and grass species from 18 different grassland sites
Full size table

Results

Soil δ15N values in the 18 grassland sites varied between 3.39 and 6.43‰ (Table 1). NO − 3 and NH +4 pools in the soil were relatively low, ranging from below detection levels (<0.01 μg g−1) to 3.00 μg g−1 and from 3.09 to 6.99 μg g−1, respectively. Net N mineralization was dominated by net nitrification, which exceeded net ammonification in most sites. Net nitrification and net ammonification ranged from 0.08 to 1.53 μg g−1 day−1 and from −0.31 to 0.46 μg g−1 day−1, respectively (Table 1).

Soil δ15N values across the 18 sites increased significantly with net N mineralization and net nitrification in linear regression analyses. However, soil δ15N values showed a negative but non-significant trend with net ammonification (Fig. 1). Plant leaves from both forbs and grasses were all depleted in 15N compared to bulk soils and showed a positive relationship with bulk soil δ15N in linear regression analyses (Fig. 2). Δδ15N values (δ15N leaf − δ15N soil) were negative for all plant species. Despite the correction for the bulk soil background δ15N signal, Δδ15N of forbs and grasses still increased significantly with net N mineralization (Fig. 2).

Fig. 1
figure 1

Simple linear regression analyses of δ15N of bulk soil and daily net mineralization (Nmin), ammonification (NH +4 ) and nitrification (NO − 3 ) rates in 18 different temperate grasslands

Full size image
Fig. 2
figure 2

Simple linear regression analyses between δ15N of bulk soil and leaf δ15N for different forb and grass species as well as the relationship between daily net mineralization (Nmin) and foliar Δδ15N (δ15N plant − δ15N soil) for different forb and grass species in 18 temperate grasslands. d −1 Day−1

Full size image

Δδ15N values of both, forbs and grasses revealed substantial within and across site variability (Fig. 2). In order to explain what drives this observed within-site variability among different plant species and functional groups, we tested the effects of NO − 3 and NH +4 uptake on Δδ15N values of forbs and grasses from sites 2, 4 and 7 in an ANOVA (Fig. 3; Table 2). In all three models site-specific soil N fluxes had a significant effect on Δδ15N matching our observations from all 18 sites (Fig. 2). The analyses also revealed that NO − 3 uptake had no significant effect on Δδ15N of forbs and grasses. More importantly, however, NH +4 uptake had a significantly negative effect on Δδ15N and the ratio between NO − 3 and NH +4 uptake a significantly positive effect Δδ15N of forbs and grasses (Fig. 3; Table 2). Interestingly, the effects of NH +4 uptake and the ratio between NO − 3 and NH +4 uptake were independent of the functional group identity of plants, i.e., the interaction between the factor “functional group” and the covariate “N uptake” were not significant in any of the three models (Table 2).

Fig. 3
figure 3

The relationship between NO − 3 or NH +4 uptake and predicted Δδ15N (δ15N plant − δ15N soil) for 30 plant individuals from seven forb species (14 individuals) and eight grass species (16 individuals) as predicted by the ANOVA models in Table 2. N uptake was determined in site 2, site 4 and site 7

Full size image

Finally, we tested in an additional ANOVA model if the relationship between net N mineralization and Δδ15N of individual species across the 18 sites was significantly affected by functional group identity, species identity, the abundance of a species, or by any interactions of the above parameters. We found that net N mineralization, functional group identity, species identity and the abundance of plants all had significant effects on Δδ15N. More importantly, however, the effects of functional group identity on Δδ15N as well as the effects of species identity on Δδ15N depended on the net N mineralization rate of a site (i.e., significant interaction between the factor “functional group” or “species identity” and the covariate “N mineralization”) and the abundance of a particular species (i.e., significant interaction between the factor “species identity” and the covariate “abundance”).

Discussion

We found substantial variability in soil and plant δ15N values across the 18 investigated grasslands. For soil δ15N values part of this variability was explained by net N mineralization (Fig. 1). Previous studies have suggested that increasing bulk soil δ15N values may reflect increasing rates of soil N cycling which are associated with losses of 15N-depleted mineral N that lead to a gradual 15N enrichment of the remaining bulk soil N (Mariotti et al. 1981; Amundson et al. 2003; Pardo et al. 2006). Although the grasslands investigated here are extensively managed and have not been fertilized for at least 10 years, previous studies have shown that NO − 3 in such grasslands is prone to losses, either via leaching or via microbial denitrification to N2O and N2 (Scherer-Lorenzen et al. 2003; Tilsner et al. 2003b). Our study therefore fits nicely into the context of previous work that has linked increasing bulk soil δ15N values to losses of mineral N and to the openness of the N cycle (Högberg 1990; Högberg 1991; Johannisson and Högberg 1994; Austin and Vitousek 1998; Emmett et al. 1998; Korontzi et al. 2000; Brenner et al. 2001; Pardo et al. 2002, 2006; Amundson et al. 2003; Choi et al. 2003; Houlton et al. 2006; Templer et al. 2007).

In order to compare δ15N values of plants across sites irrespective of soil δ15N, we corrected plant δ15N values for soil δ15N and calculated Δδ15N, the deviation of δ15N plants from soil δ15N (Amundson et al. 2003). Δδ15N was negative across all 18 sites for all species (Fig. 2). Negative Δδ15N values are typically considered to result from mineral soil N uptake, which is depleted in 15N compared to bulk soil N or soil organic N (Mariotti et al. 1981; Robinson 2001). It has been shown that plants can discriminate against 15N during N acquisition, either directly, or indirectly via associations with mycorrhizal fungi (Evans et al. 1996; Michelsen et al. 1996; Hobbie et al. 2000; Hobbie and Colpaert 2003). Direct discrimination against 15N during N uptake is, however, considered not to be relevant in ecosystems where N is a critical and limiting resource such as in the extensively managed temperate grasslands investigated in this study (McKee et al. 2002). Also, isotopic fractionation associated with N uptake via mycorrhizal fungi has typically been observed in ecto- or ericoid mycorrhizal plants (Emmerton et al. 2001). In the grasslands investigated here, only arbuscular mycorrhizal fungi (AMF) infections of roots were found (Börstler et al. 2006), and AMF plants are assumed to discriminate only marginally against 15N during N uptake (Handley et al. 1993; Azcon-G-Aguilar et al. 1998). Consequently, we conclude that Δδ15N values of plant material in our study are in fact driven by the uptake of 15N-depleted inorganic N and that plant material reflects the δ15N signature of plant available soil N.

Across the 18 sites, Δδ15N values of grasses and forbs were not constant, but became less negative (less 15N depleted compared to soils) with increasing net N mineralization in the soil (Fig. 2). Assuming that average plant δ15N of a site reflects the δ15N signature of the plant’s N source, i.e., mineral N, this pattern indicates that soil mineral N becomes increasingly enriched in 15N with increasing net N mineralization. This finding matches our interpretation of bulk soil δ15N, which becomes enriched with increasing soil N mineralization (Fig. 1) since accelerating losses of 15N-depleted NO3, N2O and N2 with increasing net N mineralization result in 15N enrichment of the remaining inorganic N pool. Changes in average Δδ15N values across different ecosystems therefore reflect rates of N mineralization in soils.

Despite net N mineralization explaining some of the variation in overall plant Δδ15N, there remains a large variability among different species within each site (Fig. 2). Based on the assumptions that: (1) the observed Δδ15N values of plants largely reflect the δ15N signature of their N source; and (2) the fact that NO − 3 and NH +4 can vary substantially in their δ15N signature depending on the relative rates of mineralization, nitrification and denitrification (Shearer et al. 1974; Mariotti et al. 1981; Handley and Raven 1992), we tested if differences in NO − 3 versus NH +4 uptake might explain the variability in Δδ15N values among plant species within a site. Interestingly, NO − 3 uptake showed no significant effect on leaf Δδ15N values (Fig. 3; Table 2). Although not specifically tested in this study, the location of NO − 3 assimilation (roots or foliage) can vary among plants. Due to strong isotope effects during plant NO − 3 assimilation, any difference in the location of NO − 3 assimilation could obscure a potential δ15N signal derived from soil NO − 3 in plant tissues, and could thus explain the lack of a direct influence of NO − 3 uptake on leaf δ15N (Pate et al. 1993).

Other than NO − 3 , NH +4 uptake and the ratios of NO − 3 to NH +4 uptake significantly affected Δδ15N of plants, with increasing NH +4 uptake leading to decreased Δδ15N values (Fig. 3; Table 2). We are aware of only two other studies that have directly tested the effects of NO − 3 and NH +4 uptake on foliar δ15N values (Miller and Bowman 2002; Falkengren-Grerup et al. 2004). Interestingly, both studies found that plant δ15N values decreased with increasing NO −3 :NH +4 uptake ratio, a relationship that is opposite in direction to what we found in our experiment (Fig. 3). These contrasting results suggest that 15N enrichment of NH +4 compared to NO − 3 can vary significantly in different ecosystems, depending on the nature of the N cycle. In an ecosystem with a closed N cycle, inorganic N compounds become increasingly depleted in 15N along the N cycle, i.e., NO − 3 is 15N depleted compared to NH +4 , and therefore plants should also become depleted with increasing NO −3 :NH +4 uptake (Mariotti et al. 1981; Nadelhoffer and Fry 1994; Robinson 2001). However, NO − 3 can become enriched in 15N compared to NH +4 in ecosystems where denitrification rates are high (Mariotti et al. 1981; Piccolo et al. 1994; Robinson 2001; Houlton et al. 2006, 2007; Pörtl et al. 2007). For extensively managed temperate grasslands such as investigated in this study denitrification has been shown to be an important path of N loss (Tilsner et al. 2003a, b). Consequently, denitrification causing NO − 3 to become 15N enriched compared to NH +4 could explain why plants with greater NO − 3 over NH +4 uptake are enriched in 15N. Our data therefore confirm the studies of Miller and Bowman (2002) and Falkengren-Grerup et al. (2004) showing that within an ecosystem the variability of foliar Δδ15N values gives important information on the plants’ NO −3 :NH +4 uptake ratio. However, our study also shows that the direction of the relationship between foliar Δδ15N values and the plants’ NO −3 :NH +4 uptake ratio can vary, depending on the dynamics of the N cycle of the ecosystem under consideration.

Interestingly, the effect of NO − 3 to NH +4 uptake on Δδ15N in our study was independent of a plant’s functional group identity (Table 2). This suggests that leaf Δδ15N is not affected by fractionation patterns during N uptake specific to functional groups nor by functional group-specific differences in internal N allocation. Consequently, the δ15N signature of the plant’s N source is a critical driver of variability in Δδ15N among plant species within an ecosystem.

We tested if the observed patterns of within-site variability in Δδ15N among species are consistent across the 18 sites or vary in a functional group- or species-specific manner (Table 3). Both, functional group- and species-specific effects on Δδ15N were detected, but the effects were dependent on net N mineralization rates (Fig. 4; Table 3). In addition, depending on the identity of a species, the abundance of a particular species also had a significant effect on Δδ15N. In summary, functional groups as well as species within these functional groups differed significantly in Δδ15N, but the patterns of these differences varied across sites with N mineralization rates and abundance. There are two possible explanations for changes in a species’ foliar Δδ15N across sites (see also Houlton et al. 2007, addressing similar mechanisms in tropical rainforest species). First: along an environmental gradient, plants might shift their N source as a result of changes in their competitive strength (abundance) compared to other plants and microorganisms. Assuming that the δ15N values of NO − 3 and NH +4 remain constant along the gradient, changes in plants’ foliar Δδ15N values would then reflect a shift in N sources. Second: δ15N values of NO − 3 and NH +4 shift with changing environmental conditions and changes in plants’ foliar Δδ15N values would reflect these changes in soil N dynamics assuming that plants do not change their N source. Unfortunately we cannot distinguish between the two possible explanations in this study since we did not determine the δ15N values of NO − 3 and NH +4 given the uncertainties and complexity of the methods involved in these analyses and the vast number of samples that would have to be processed to account for the seasonal and spatial variability in isotope composition of plant-available N pools for 18 different sites. Method development towards an easier assessment of isotopic ratios in soil NO − 3 and NH +4 (and soluble organic N) are urgently needed so that ecologists can exploit the full potential of contrasting foliar δ15N values among plant species within and across different sites. This may not only allow a more detailed assessment of ecosystem N dynamics but also the determination of shifts in plant–plant and plant–microbe interactions when competing for different soil N sources along environmental gradients.

Fig. 4
figure 4

The relationship between daily net mineralization (Nmin) and Δδ15N (δ15N plant − δ15N soil) for 20 different plant species from 18 different temperate grasslands. For details of the statistical analyses see Table 3. Open symbols represent observed values, closed symbols values predicted by the ANOVA model presented in Table 3

Full size image