Introduction

Owing to human activities, global nitrogen (N) deposition has dramatically increased and is expected to reach 200 Tg N yr−1 by 2050 (Galloway et al. 2008). N fertilization in agricultural ecosystems has steadily increased as well and is expected to reach 103 Tg N yr−1 by 2050 (Bouwman et al. 2013). Excess N inputs could relieve N limitation in some ecosystems, but may lead to an imbalance of the N cycle, causing many negative effects. N cycling is complex, with multiple transformation processes, feedbacks, and interactions with other biogeochemical cycles. The natural abundance of stable N isotope ratios (15N:14N expressed as δ15N) has been used as an indicator of the N status of an ecosystem or an N transformation process (Bai and Houlton 2009; Craine et al. 2009; Kalcsits et al. 2014) and could be a convenient method to study how increasing reactive N would affect N cycling.

The variation of δ15N value among different components in terrestrial ecosystems is mainly caused by isotope fractionation and isotope mixing. For a reaction or a biogeochemical process, light isotope 14N is usually more abundant in the product, and heavy isotope 15N is usually enriched in the substrate (Evans 2001). For an ecosystem N pool, its δ15N value is determined by both input processes and output processes (Brenner 2001; Amundson 2003). Atmospheric N deposition and biological N fixation are generally isotopically depleted sources compared to δ15N values of soils and plants (Amundson 2003). Therefore, if N output rates remain constant, the increase of N input rates would decrease δ15N values of soils and plants. However, the N output processes such as denitrification and leaching generally have strong isotope fractionation against 15N (Houlton et al. 2006; Bai and Houlton 2009; Brookshire et al. 2012). If the rates of these output processes change under increasing N deposition, the responses of plant and soil δ15N values would be more complex (Högberg et al. 2014; Kalcsits et al. 2014; Kriszan et al. 2009). For instance, long-term N fertilization was found to enhance N losses through gaseous pathway and to significantly increase soil δ15N value (Kriszan et al. 2009). Therefore, soil and plant δ15N signatures could provide information on how ecosystem output processes respond to increasing N inputs. In addition, foliar δ15N value has been widely recognized as a useful integrator to characterize the N status of an ecosystem and higher foliar δ15N value generally indicates higher N availability (Evans 2001; Robinson 2001; Craine et al. 2009; Craine et al. 2015). Thus, examination of ecosystem δ15N change under N addition treatment could provide valuable information on how N fluxes, assimilation, and allocation would respond to increasing N inputs into terrestrial ecosystems.

Over the past three decades, a series of individual relevant studies on the effects of N addition on the natural abundance of δ15N value have been conducted. However, inconsistencies among results have hampered our ability to draw general conclusions. Two primary reasons are probably responsible for this problem: (1) these studies were conducted in various ecosystems (e.g., forest, shrubland, grassland, and agriculture) and different types of plants were tested (e.g., tree, shrub, grass, crop, moss, and lichen) and (2) the experimental setup differed among different studies. For example, several kinds of fertilizers, different rates of N addition, and varying length of treatment duration were implemented in these studies. To improve the utilization of δ15N signature as an indicator or a predictor of N cycling change under increasing N deposition or fertilization, here we synthesized the results across these individual studies and conducted a meta-analysis following previous suggestions (Hedges et al. 1999).

By compiling data from 48 individual studies, we conducted a meta-analysis to identify the general patterns of the responses of plant and soil δ15N signatures to simulated N deposition and to investigate the differences among different experimental scenarios (e.g., ecosystem/plant types, plant functional types, treatment duration, N addition rates, and N fertilizer types), which could help us better predict how ecosystem N status would respond to increasing N deposition or fertilization.

Materials and methods

Data compilation

In total, 48 papers with 951 observations (an observation is composed of both control and N addition treatment data for one examined variable) were selected from the Web of Science resource (Supporting Information, Reference S1 and Table S1). The search terms “N isotope” and “nitrogen addition,” “N-15” and “nitrogen input,” “nitrogen natural abundance” and “deposition, ” “nitrogen addition” and “isotope signature, ” and “nitrogen loss” and “15N” were used. Any selected articles should meet the following criteria: at least one of our selected variables was reported; the means and sample sizes for both control and N addition treatments had to be provided; and for multifactorial studies, only control and N addition treatment data were selected and the interacting effects were excluded.

For each study, we noted the ecosystem type (or plant type), N addition rate, fertilizer type, treatment duration, the response variables, and other background information (e.g., soil type and pH, mean annual precipitation, and ambient N deposition). Furthermore, specific name and plant functional types (e.g., broadleaf/needle, evergreen/deciduous, perennial/annual, C3/C4, N2-fixing/non-N2-fixing) were noted for the sampled plants. Meanwhile, the change of corresponding N content was also extracted if reported. We used Getdata software to extract the data which were expressed in figures.

In order to better understand the responses of δ15N signature to experimental treatments, we grouped our data according to ecosystem/plant type (forest, shrubland, grassland, cropland/trees, shrubs, herbs, crops, moss and lichen), N addition rate (<100, 100–200, and ≥200 kg N ha−1 yr−1), fertilizer type (urea, NH4NO3, NH4Cl/(NH4)2SO4, KNO3/NaNO3, and manure), and treatment duration (<3, 3–10, and >10 years). We also grouped plants into different organs (leaf, wood, grain, and root) and functional types.

Statistical analyses

Due to the negative values of δ15N, the commonly used “effect size”—the natural log of the response ratio (Hedges et al. 1999), cannot be applied. We used the Hedges’ d as the effect size instead (Hedges and Olkin 1985) to quantify the response of δ15N signatures to N addition. We assumed that other factors were constant and only N input rates were different between the treatments (N addition and control). The d value was calculated as Eq. 1.

$$ d\kern0.5em =\kern0.5em \frac{\left({X}_e\kern0.5em -\kern0.5em {X}_c\right)}{S}J $$
(1)
$$ S\kern0.5em =\kern0.5em \sqrt{\frac{\left({N}_e\kern0.5em -\kern0.5em 1\right){\left({S}_e\right)}^2\kern0.5em +\kern0.5em \left({N}_c\kern0.5em -\kern0.5em 1\right){\left({S}_c\right)}^2}{N_c\kern0.5em +\kern0.5em {N}_e\kern0.5em -\kern0.5em 2}} $$
(2)
$$ J\kern0.5em =\kern0.5em 1\kern0.5em -\kern0.5em \frac{3}{4\left({N}_c\kern0.5em +\kern0.5em {N}_e\kern0.5em -\kern0.5em 2\right)\kern0.5em -\kern0.5em 1} $$
(3)

where X e and X c are the mean values of δ15N (‰) under N addition treatment and control, respectively; S is the pooled standard deviation (Eq. 2); and J is used to correct the bias when sample sizes are small (Eq. 3). N e and N c are the sample sizes for the N addition treatment and control groups, respectively. S e and S c are the standard deviations (SDs) for the N addition treatment and control groups, respectively. If only standard error (SE) was reported, we calculated SD by multiplying SE by the square roots of sample size N (SD = SE \( \sqrt{\mathrm{N}} \)). If neither SD nor SE was reported, the arithmetic mean SD calculated from each data set was used (Wiebe et al. 2006).

The variance of d (v) was approximated using the following equation:

$$ v\kern0.5em =\kern0.5em \frac{N_c\kern0.5em +\kern0.5em {N}_e}{N_c{N}_e}+\frac{d^2}{2\left({N}_c\kern0.5em +\kern0.5em {N}_e\right)} $$
(4)

A non-parametric weighting function was used to weight individual studies (Hedges et al. 1999). We calculated the weighting factor w of each observation by the inverse of the pooled variance (1/v). Because most studies contained two or more observations for one variable, we adjusted the weight by the total number of observations per study in order to reduce the weight from the same study (Bai et al. 2013) and used the final weight (w′) to calculate the mean effect sizes (\( \overline{d}^{\prime } \)) for all observations (Eqs. 5–7).

$$ {w}^{\prime}\kern0.5em =\kern0.5em w/n $$
(5)
$$ {d}^{\prime}\kern0.5em =\kern0.5em {w}^{\prime}\kern0.5em \times \kern0.5em d $$
(6)
$$ {\overline{d}}^{\prime }=\frac{{\displaystyle {\sum}_i{d}_i^{\prime }}}{{\displaystyle {\sum}_i{w}_i^{\prime }}} $$
(7)

where n is the total number of observations in the same study, d′ is the weighted effect size, and \( {d}_i^{\prime } \) and \( {w}_i^{\prime } \) are the d′ and w′ of the ith observation, respectively.

In order to determine whether N addition had a significant effect on a variable, we used a fixed-effect model completed by Metawin 2.1 software (Rosenberg et al. 2000). Confidence intervals (CIs) of the weighted effect size were generated using bootstrapping with 60,000 iterations. If the 95 % CI values did not overlap with 0, we deemed that N addition had a significant effect on a variable. Otherwise, the effect was not considered to be significant.

The difference of plant and soil δ15N values between N addition treatment and control under different fertilize types (Δ t − c , ‰) was calculated according to the following equation:

$$ {\Delta}_{t-c}\kern0.5em =\kern0.5em {X}_e\kern0.5em -\kern0.5em {X}_c $$
(8)

At last, a continuous randomized-effect model was used to find out if there was any correlation between the effect size of N addition on δ15N signature and the duration and addition rate of N addition treatment. The total heterogeneity of d′ among studies (Q T ) was partitioned into the difference among group cumulative effect sizes (Q M ) and the residual error (Q E ) (Rosenberg et al. 2000).

Results

Effects of N addition on plant δ15N signatures

Plant δ15N value was significantly increased by N addition, and the mean effect size and actual 15N change were 1.1093 (95 % CI 0.0527∼1.6321) and 1.25‰, respectively (Fig. 1a and Table 1). When different plant types were considered, δ15N values of trees, shrubs, herbs, and moss/lichen were still higher under N addition treatment, and the mean effect sizes were 0.4372 (95 % CI 0.0356∼0.871), 0.7003 (95 % CI 0.1055∼1.3464), 0.7922 (95 % CI 0.4648∼1.1603), and 1.0913 (95 % CI 0.5072∼1.7823), respectively (Fig. 1b). However, δ15N value of crops showed a non-significant response to N addition (Fig. 1b).

Fig. 1
figure 1

The mean effect size of nitrogen addition on the natural abundance of plant δ15N and the difference among different plant types (b), nitrogen addition rates (kg ha−1 yr−1) (c), fertilizer types (d), and duration of treatment (year) (e). The error bars represent the 95 % confidence intervals (CIs). The solid points are the variables with >20 observations, and the hollow points are the variables with <20 observations. The dashed line was drawn at mean effect size = 0. The effect of nitrogen addition was considered to be significant if the 95 % CI of the effect size did not cover zero. The sample size for each variable was shown next to the point. AS ammonium salt ((NH4)2SO4/NH4Cl), NS nitrate salt (KNO3/NaNO3)

Full size image
Table 1 The mean plant and soil δ15N values under N addition treatment with different fertilizer types and control and their differences (δ15Ntreatment − δ15Ncontrol, Δ t − c ) and δ15N values of the source fertilizer (δ15Nfertilizer, based on values in Table S2)
Full size table

Different experimental methods showed different effects on plant δ15N signature (Fig. 1c, d, e). Plant δ15N value was most increased when the ammonia salt (NH4Cl or (NH4)2SO4) was applied, and the mean effect size was 1.9843 (95 % CI 0.1108∼2.0403) (Fig. 1d). NH4NO3, nitrate salt (KNO3 or NaNO3), and manure addition treatments all had a positive effect on plant δ15N value. Whereas, urea showed a negative effect on plant δ15N value (mean effect size = −0.4812, 95 % CI −0.8889∼0.0107) (Fig. 1d). The continuous randomized-effect model showed that inorganic N addition rate had a significantly positive correlation with the effect size of N addition on all plant δ15N values (Table 2). However, when the rates of N addition were divided into three levels (<100, 100–200, and ≥200 kg N ha−1 yr−1), we found that plant δ15N value was significantly increased under the lowest level of N addition (<100 kg N ha−1 yr−1), and non-significant change was found under the other two levels (Fig. 1c). On the other hand, the continuous randomized-effect model showed that the duration of inorganic N addition treatment had a significantly positive correlation with the effect size of N addition on δ15N value of shrubs (Table 2). We further partitioned duration into short (<3 years), medium (3–10 years), and long terms (>10 years) and found that the increase of plant δ15N value was highest when long-term duration (>10 years) of N addition was applied (mean effect size = 1.1452, 95 % CI 0.8996∼1.4620). No significant correlation was found between organic N addition rate and plant δ15N value (Table 2).

Table 2 Relationships between the effect size of nitrogen addition on plant (or soil) δ15N and N addition rate (or treatment duration)
Full size table

Different plant organs showed different responses to N addition (Fig. 2a). N addition had a significantly positive effect on δ15N values of leaf, wood, and root and a significantly negative effect on δ15N value of grain (mean effect size = −1.0029, 95 % CI −1.4772∼−0.0296). Similarly, different plant functional types also showed different responses to N addition (Fig. 2b, c, d). Trees and shrubs were classified into broadleaf and needle or evergreen and deciduous functional types (Fig. 2b). Plant δ15N value in the needle group was significantly increased by N addition with a mean effect size of 1.1243 (95 % CI 0.8821∼1.7542), but no significant change was found for the broadleaf group. Both the evergreen and the deciduous groups showed positive change of plant δ15N value under N addition, and the mean effect sizes were 0.5468 (95 % CI 0.0920∼1.0525) and 0.2502 (95 % CI −0.0304∼0.5429), respectively. Grasses and crops were classified into perennial and annual functional groups, and we found that δ15N value of perennial plants was significantly increased by N addition (mean effect size = 0.7483, 95 % CI 0.4038∼1.3110), and δ15N value of annual plants had no significant change (Fig. 2c). When plants were divided into C3 and C4 groups according to their photosynthesis pathways, the effect size of N addition on δ15N value of C3 plants was obviously higher than that of C4 plants, although neither was significant (Fig. 2d). When plants were divided into non-N2-fixing and N2-fixing functional groups, the change of δ15N value by N addition was higher for non-N2-fixing plants than for N2-fixing plants (Fig. 2d).

Fig. 2
figure 2

The mean effect size of nitrogen addition on plant δ15N natural abundance of different plant organs (a) and functional types (b–d). The error bars represent the 95 % confidence intervals (CIs). The effect of nitrogen addition was considered to be significant if the 95 % CI of the effect size did not cover zero. The sample size for each variable was shown next to the point

Full size image

Effects of N addition on soil δ15N signatures

The δ15N values of soil total N and soil NO3 − were significant increased by N addition with mean effect sizes of 0.6685 (95 % CI 0.3894∼0.9626) and 0.6841 (95 % CI 0.1742∼1.2628), respectively. On the contrary, there was a negative effect of N addition on δ15N values of soil NH4 + (mean effect size = −0.2007, 95 % CI −1.2467∼0.6482) and soil N2O (mean effect size = −1.5639, 95 % CI −2.3180∼−0.9896) (Fig. 3a). There were only two observations for δ15N value of soil leachate, which were not enough for evaluation (Fig. 3a).

Fig. 3
figure 3

The mean effect size of nitrogen addition on soil δ15N natural abundance and the difference among different ecosystems (b), nitrogen addition rates (kg ha−1 yr−1) (c), fertilizer types (d), and duration of treatment (year) (e). The error bars represent the 95 % confidence intervals (CIs). The solid points are the variables with >20 observations, and the hollow points are the variables with <20 observations. The dashed line was drawn at mean effect size = 0. The effect of nitrogen addition was considered to be significant if the 95 % CI of the effect size did not cover zero. The sample size for each variable was shown next to the point

Full size image

All of the four ecosystem types (forest, shrubland, grassland, and cropland) showed a positive effect of N addition on δ15N value of soil total N (Fig. 3b). The δ15N value of soil total N was significantly increased by urea, NH4NO3, and manure fertilization but was significantly decreased by (NH4)2SO4 addition (Fig. 3d). The mean effect size and actual 15N discrimination were 1.7046 and 2.01‰ when manure was applied, reaching the highest level (Fig. 3d and Table 1). The continuous randomized-effect model and classified N addition rate analysis both showed that the effect of N addition on δ15N value of soil total N was independent of N addition rate (Table 2 and Fig. 3c). Although the continuous randomized-effect model showed that inorganic and organic N addition duration had little correlation with the effect size of N addition on δ15N value of soil total N (Table 2), long-term duration (>10 years) did have a significantly more positive effect on δ15N value of soil total N with a mean effect size of 0.9524 (95 % CI 0.5861∼1.4946) compared to the two treatments with short duration (Fig. 3e).

Discussion

Effects of N addition on soil δ15N signatures

In this meta-analysis, δ15N value of soil total N was significantly increased by N addition (Fig. 3a). The δ15N value of soil total N is regulated by both N input and N output fluxes. The δ15N value of inorganic N fertilizer generally is close to zero (Table S2), which is lower than soil δ15N value in this meta-analysis (Table 1). N addition would have caused a decrease of soil δ15N value if output fluxes did not change; so the increasing trend of soil δ15N value under N addition we observed must be caused by output fluxes. Our speculation was based on the mass balance model by Brenner et al. (2001), which showed that at steady state, soil δ15N is determined by the weighted mean isotopic ratio of inputs divided by the proportion of 15N to 14N leaving the soil (Eq. A11 in Brenner et al. 2001). Because N addition generally decreases the weighted mean isotopic ratio of inputs (Bateman and Kelly 2007), higher soil δ15N value under N addition must be caused by lower proportion of 15N to 14N leaving the soil. Therefore, we speculate that the losses of 15N-depleted pools through leaching and gaseous emissions increased following N addition. However, studies on δ15N values of gaseous losses and soil leachate are extremely rare and difficult and we could not find enough observation data for this meta-analysis. A previous study suggested that the δ15N value of volatilized ammonia ranged from −28 to −0.3‰, which was more depleted than δ15N value of the originally added urea-N (1.2‰) (Frank et al. 2004). Koba et al. (2012) found that both NH4 + in stream water and soil-emitted N2O were extremely 15N depleted (the δ15N values were −9.2 and −14.3‰, respectively) compared to soil δ15N value (+3.9‰) in an N-saturated subtropical forest. These N output fluxes usually lead to an isotope fractionation against 15N, leaving heavier 15N in soil (Robinson 2001). In addition, plant N uptake also could cause isotope discrimination, enriching soil δ15N value in a short term (Evans 2001). Therefore, this result suggested that soil δ15N value was more affected by N loss fluxes and plant N uptake than by N inputs.

The δ15N value of soil NO3 − was significantly increased by N addition, while δ15N value of soil NH4 + was not significantly changed (showing a decreasing trend) by N addition (Fig. 3a). When the integrated δ15N value of input fluxes is higher than the integrated δ15N value of output fluxes for an N pool, its δ15N value is increased and vice versa. The input fluxes (nitrification and direct N addition (including N deposition)) and the output fluxes (immobilization, denitrification, leaching, and plant uptake) of soil NO3 − together determine the size of this pool. The integrated δ15N value of nitrification and N addition derived NO3 − should generally have been reduced by N addition (Robinson 2001), so the increase of δ15N value of soil NO3 − by N addition was most likely caused by the output fluxes. This result indicated that N addition decreased the integrated δ15N value of output fluxes for NO3 −, which was most likely caused by the increase of 15N-depleted outputs such as denitrification and leaching (Robinson 2001). Although N addition increased δ15N value of soil NO3 − (3.412‰), it was still lower than soil δ15N value (4.082‰). Therefore, the increased losses of this pool would still cause the enrichment of soil δ15N value under N addition. In addition, the observation number is too low for studies on N addition effects on δ15N value of soil NO3 − and more studies are needed to evaluate its responses. In our meta-analysis, the two observations of N addition on soil leachate were too few, which were not enough to evaluate (Fig. 3a). We suggest that more studies should be carried out on this topic. On the other hand, the decrease of δ15N value of N2O by N addition in the meta-analysis could further support the result that the increased δ15N value of soil NO3 − was mainly caused by denitrification (Fig. 3a). N losses via NOx and N2 also contribute to the enrichment of soil δ15N value, whose effects could be much bigger than the effects of N2O emission (Robinson 2001). However, due to the low observation numbers on NOx and N2 losses under N addition, we did not include these fluxes in our study.

For soil NH4 +, the input fluxes are mineralization and direct N addition (including N deposition), while the output fluxes are immobilization, nitrification, ammonia volatilization, and plant uptake. As discussed above, the increase of output fluxes would have enriched δ15N value of soil NH4 +, so the unchanged soil NH4 + under N addition was most likely the balanced result between input and output fluxes. We speculate that the increase of N mineralization under N addition (Lu et al. 2011) produced more 15N-depleted NH4 +, reducing δ15N value of soil NH4 +. Therefore, our results suggest that N addition has accelerated many N-cycling processes such as denitrification and mineralization, which are reflected by the change of natural abundance of nitrogen isotopes.

The δ15N value of soil total N was significantly increased by N addition in forest, grassland, and cropland but was not affected by N addition in shrubland (Fig. 3b). There were only four observations in shrubland, which may be the potential reason of the discrepancy. More studies should be carried out in shrubland to better understand the responses of nitrogen cycling to N addition in this ecosystem type.

Our results indicated that δ15N value of soil total N was increased most when manure was applied among all types of fertilizer (Fig. 3d). Manure is more enriched in 15N than inorganic fertilizer in this meta-analysis (Table S2), which is due to the isotope fractionation during digestion processes and ammonium volatilization (Kerley and Jarvis 1996; Choi et al. 2002). Urea and NH4NO3 application also increased δ15N value of soil total N, but application of (NH4)2SO4 showed a different effect (Fig. 3d). Previous studies indicated that the growth of ammonia-oxidizing bacteria was restrained, and thus, nitrification was inhibited when (NH4)2SO4 was applied (Tong and Xu 2012). This inhibition of microbial activities by (NH4)2SO4 might be the potential reason of the decrease of soil δ15N value under (NH4)2SO4 treatment.

The continuous randomized-effect model showed no correlation between the effect size of N addition on soil δ15N value and N addition rate or treatment duration (Table 1). However, when treatment duration was classified into three categories, δ15N value was significantly higher under longer-term (>10 years) treatment compared to the other two shorter terms. These results indicated that increasing N addition rate had minor effects on soil δ15N value, but with increasing treatment duration, soil δ15N value may be more affected by N addition.

It should be noted that other factors, such as water availability, nutrient status, microbial community composition, and soil physical properties, can also affect plant and soil 15N signatures (Högberg 1997; Robinson 2001; Pardo and Nadelhoffer 2010). While N addition treatment changed N input rates, those above-mentioned factors may also be affected, changing the effect size of N addition treatment. Therefore, it is important to examine if other factors also changed when N addition treatment was applied and caution should be taken when explaining the responses of plant and soil δ15N to N inputs due to the complexity of natural 15N abundance.

Effects of N addition on plant δ15N signatures

N addition significantly increased plant δ15N value (Fig. 1). The source of plant N of non-N2-fixing plants is soil N; therefore, the increase of plant δ15N value was directly linked to the rise of soil δ15N value (Fig. 3). In addition, the isotope fractionation during plant N uptake (Evans 2001) might have been decreased by N addition, enriching plant δ15N value. When soil N availability was lower, plants might depend more on mycorrhizal fungi, which usually transfer 15N-depleted N to plants (Hobbie and Colpaert 2003). Therefore, N addition increased N availability, weakening the role of mycorrhizal fungi, which could increase plant δ15N value. Different plant types did not show different response, except for the big variation of the effect size of N addition on crop δ15N value (Fig. 1). The effect size of inorganic N addition on δ15N value of shrubs showed a positive correlation with treatment duration, and the effect size of inorganic N addition on δ15N value of all types of plants showed a positive correlation with N addition rate (Table 1), which suggested that cumulative enrichment of plant δ15N value might happen with increasing inorganic N addition duration and rate. However, no significant correlation was found between organic N addition rate (or treatment duration) and plant δ15N value. One reason was because studies on organic N addition were mostly done in croplands with high N addition rates. Therefore, the lack of data under low N addition levels and long-term treatments caused the insignificant relationship between organic N addition rate (or treatment duration) and plant (or soil) δ15N values. In addition, the mean δ15N value of manure was 10.19‰ with high-standard variation (Table S2), which made the variations of plant or soil δ15N values more complex and uncertain.

It is worth noting that under urea addition, plant δ15N value was decreased (Fig. 1), although soil δ15N value was significantly increased (Fig. 3). We speculate that the following three potential reasons were responsible for this phenomenon. First, substantial 15N-depleted NH3 would be volatilized when urea was applied (Elliot and Fox 2014), and studies suggested that plant shoot and leaves can directly absorb NH3 from the air when its partial pressure exceeded a certain threshold (Farquhar et al 1980; Ping et al. 2000; Frank et al. 2004; Boaretto et al. 2013). Second, during the hydrolysis of urea, some 15N-depleted NH4 + was quickly used by plant roots (Bateman and Kelly 2007), while the rest NH4 + underwent N transformation processes which enriched soil N pools. Lastly, many plants can take up urea intactly (Witte 2011), making urea different from other N forms as N sources.

The δ15N value of different plant organs showed various responses to N addition (Fig. 2a). Both leaf and root δ15N values were significantly increased by N addition with similar mean effect sizes. Therefore, although there could be intra-plant isotope discrimination (Evans 2001), the effect of N addition on this discrimination was minimal based on our meta-analysis. The effect size of N addition on wood δ15N value was also positive (Fig. 2a), although wood δ15N value reflects cumulative N with plant growth (Hietz et al. 2010). Interestingly, we observed that δ15N value of crop grains was significantly reduced by N addition. Grain δ15N value is generally more negative than shoot and root δ15N values (Yun et al. 2011). Therefore, the reduction of grain δ15N value by N addition hinted increased isotopic discrimination during N transport to grain. The 15N fractionation processes such as assimilation, reallocation, and transformation to grain may be affected by N addition (Serret et al. 2008; Zhou et al. 2013).

For trees and shrubs, δ15N value of needle plants was increased most by N addition, while δ15N value of broadleaf plants was unresponsive to N addition (Fig. 2b). In general, broadleaf plants are more 15N enriched than needle plants (Martinelli et al. 1999; Fang et al. 2011). So, our results indicated that needle δ15N value may be more sensitive to N addition. A previous study also demonstrated that leaf δ15N values showed site-specific and species-specific responses to N addition (Silva et al. 2015). Of course, other factors, such as N preference, may also contribute to this phenomenon (Evans 2001).

For herbs and crops, we classified them into perennial and annual function groups. We found that δ15N value of perennial plants was significantly increased while δ15N value of annual plants was unchanged by N addition (Fig. 2c). This was probably due to their different N utilization strategies. A meta-analysis pointed out that the increase of N in annual plants was higher than that in perennial plants under N addition, probably because annual plants had advantages of absorbing added N (Xia and Wan 2008). Because added N was generally more 15N-depleted N compared to soil N, N addition effects on δ15N value of annual plants were thus weakened.

The effect size of N addition on δ15N value of C3 plants was bigger than that on δ15N value of C4 plants (Fig. 2d). The underlying mechanism is still unclear. Similar to non-N2-fixing plants, δ15N value of N2-fixing plants was significantly increased by N addition (Fig. 2d). Plant δ15N values of N2-fixing plants are generally close to 0‰, because of their N source from atmospheric N2. However, N2 fixation is a process with high energy consumption (Gutschick 1981; Fisher et al. 2010), and with N addition, N2-fixing plants would swift a proportion of their N sources from atmospheric N2 to soil N to avoid high energetic cost. Previous studies have found that nodule formation and nitrogenase activity were inhibited by N addition, and the function of biological N2 fixation could be reduced (Streeter and Wong 1988; Vessey and Waterer 1992). Therefore, δ15N value of N2-fixing plants was also significantly changed by N addition.

Conclusion

Based on our meta-analysis of 48 studies, the conceptual model of plant and soil δ15N signatures to N addition is presented in Fig. 4. This is only a conceptual model, which reflects general patterns but may not fit all cases equally well. In summary, N addition generally increased both soil and plant δ15N values, which suggested that N addition would increase N losses via gases or leaching pathways. The δ15N value of soil NO3 − was significantly increased by N addition, while δ15N value of soil NH4 + showed a decreasing trend under N addition, although it was not significant. The δ15N value of needle plants was increased the most among all types of trees and shrubs. The δ15N value of perennial grasses showed higher responses to N addition than that of annual grasses. The significantly positive relationships between inorganic N rate and the effect size of inorganic N addition on δ15N value of all types of plants indicated that a cumulative enrichment of plant δ15N value might happen with increasing N addition duration. Our study revealed a comprehensive picture of the effects of N addition on δ15N signatures in terrestrial ecosystems, which could help us to understand how plant and soil δ15N values respond to N addition and to better use δ15N signature as an indicator to predict how ecosystem N status would respond to increasing N deposition or fertilization. Ecosystem δ15N signatures may provide information on thresholds of atmospheric N levels and should be more explored in global climate change studies.

Fig. 4
figure 4

Conceptual model of δ15N response to nitrogen addition based on our meta-analysis. Min mineralization, AS assimilation, Vol ammonia volatilization, Den denitrification. The plus sign means a significant positive effect, the minus sign means a significant negative effect, the question mark means not clear at present, and n.s. means not significant

Full size image