Soil δ¹⁵N is a better indicator of ecosystem nitrogen cycling than plant δ¹⁵N: A global meta-analysis

The nitrogen-15 (¹⁵N) natural abundance composition (δ¹⁵N) in soils or plants is a useful tool to indicate the openness of ecosystem N cycling. This study aimed to evaluate the influence of the experimental warming on soil and plant δ¹⁵N. We applied a global meta-analysis method to synthesize 79 and 76 paired observations of soil and plant δ¹⁵N from 20 published studies, respectively. Results showed that the mean effect sizes of the soil and plant δ¹⁵N under experimental warming were −0.524 (95 % CI (confidence interval): −0.987 to −0.162) and 0.189 (95 % CI: −0.210 to 0.569), respectively. This indicated that soil δ¹⁵N had negative response to warming at the global scale, where warming had no significant effect on plant δ¹⁵N. Experimental warming significantly (p<0.05) decreased soil δ¹⁵N in Alkali and medium-textured soils, in grassland/meadow, under air warming, for a 4–10-year warming period and for an increase of >3 ^∘C in temperature, whereas it significantly (p<0.05) increased soil δ¹⁵N in neutral and fine-textured soils and for an increase of 1.5–3 ^∘C in temperature. Plant δ¹⁵N significantly (p<0.05) increased with increasing temperature in neutral and fine-textured soils and significantly (p<0.05) decreased in alkali soil. Latitude did not affect the warming effects on both soil and plant δ¹⁵N. However, the warming effect on soil δ¹⁵N was positively controlled by the mean annual temperature, which is related to the fact that the higher temperature can strengthen the activity of soil microbes. The effect of warming on plant δ¹⁵N had weaker relationships with environmental variables compared with that on soil δ¹⁵N. This implied that soil δ¹⁵N was more effective than plant δ¹⁵N in indicating the openness of global ecosystem N cycling.

1 Introduction

Nitrogen (N) is one of the most important nutrient elements for plant growth and the key limiting factor for vegetation productivity (McLay et al., 2001; Zhu et al., 2018; Lu et al., 2020). On the one hand, if the available N in the soil is insufficient, it will damage and weaken the ecosystem service function, including the supply of primary material products, water conservation, climate regulation, etc. (Averill and Waring, 2018). On the other hand, if the available N in the soil is over supplied, it will also damage the structure and function of the ecosystem, resulting in a series of environmental problems such as soil acidification and imbalance of ecosystem nutrient (Schrijver et al., 2008). The intermediate products of the N cycling processes, such as nitrate nitrogen (${\mathrm{NO}}_{\mathrm{3}}^{-}$-N), nitrous oxide (N₂O) and nitric oxide (NO), may also cause eco-environmental pollution such as eutrophication of water bodies and aggravation of climate-related issues (Liao et al., 2019). Therefore, it is of great significance to reveal the openness of the ecosystem N cycle process for understanding the plant N fixation and long-term trend of N cycling and protecting the eco-environment (Wang et al., 2014; Wu et al., 2019). Openness is a measure of both N inputs and outputs relative to internal cycling and determines both the potential rate of N accumulation in the ecosystem and the potential for N losses following a disturbance (Rastetter et al., 2021).

The ¹⁵N natural abundance composition (δ¹⁵N) in soils or plants (leaves, shoots, fine roots, and litter) is often used to indicate the openness of ecosystem N cycling (Robinson, 2001). This is because the lighter isotope of ¹⁴N is always preferentially lost from the ecosystem. Thus, the isotopic fractionation effect results in gradual ¹⁵N enrichment in the ecosystem (Aranibar et al., 2004). The larger the δ¹⁵N value, the higher degree of openness of N cycling. In addition, soil δ¹⁵N also appears to reflect the degree of decomposition of the organic matter, showing that δ¹⁵N increases with processing (Craine et al., 2015). A large number of studies have confirmed that climate was the main factor regulating the soil and plant δ¹⁵N (Craine et al., 2015; Soper et al., 2015). Previous studies have demonstrated that precipitation had a negative effect on soil and plant δ¹⁵N from in situ evidence to cross-site syntheses (Swap et al., 2004; Soper et al., 2015). However, the influence of temperature on soil and plant δ¹⁵N remains controversial. Some studies have shown that soil and plant δ¹⁵N increased with temperature (Amundson et al., 2003; Craine et al., 2015), while others have indicated that δ¹⁵N decreased with temperature (Cheng et al., 2009; Sheng et al., 2014) or that they were not correlated (Yang et al., 2013). The various studies suggested that the responses of soil and plant δ¹⁵N to warming were very complex and not well understood. In addition to climate factor, soil and plant δ¹⁵N are affected by a variety of other environmental factors, such as vegetation type, topography, soil properties and management practices (Gurmesa et al., 2017; Wang et al., 2019). However, we know little about the influences of environmental factors on the warming effect on ecosystem N cycling, in terms of soil and plant δ¹⁵N.

Soil and air warming experiments have often been conducted to study the effect of warming on the ecosystem N cycling at the site scale (Schindlbacher et al., 2009). At present, the effect of experimental warming on soil and plant δ¹⁵N has not been studied on a global scale. The objectives of this study were to (i) detect the effect of experimental warming on the soil and plant δ¹⁵N based on a global meta-analysis of 20 studies and (ii) identify the main factors influencing the warming effect on the soil and plant δ¹⁵N. In addition, previous studies (e.g., Liu and Wang, 2009; Wang et al., 2014) have found that the correlation between soil δ¹⁵N and environmental factors was stronger than that for plant, which may be due to the fact that soil samples represented a long-term average for a given location, while plant samples were affected by the microenvironment or the short-term environmental fluctuations. Therefore, we specifically hypothesized that soil δ¹⁵N is a better indicator of ecosystem N cycling than plant δ¹⁵N.

Table 1 Site characteristics from a global meta-analysis of 20 studies.

^a Soil organic matter content; ^b Mean annual temperature; ^c Mean annual precipitation. If the soil organic carbon content was provided in the literature, soil organic matter content was determined by multiplying the organic carbon content by a coefficient of 1.724.

2 Materials and methods

2.1 Source of data and selection criteria

Peer-reviewed journal articles and dissertations related to soil and plant δ¹⁵N under experimental warming were searched using Web of Science and China National Knowledge Infrastructure (CNKI; http://www.cnki.net) last access: 31 March 2020 (Table 1). The keywords used for the literature search were related to “nitrogen isotope composition”, “experimental warming”, and “ecosystems nitrogen cycling”.

Our criteria were as follows: at least one of the target variables was contained, including soil (different fractions, e.g., sand, silt, clay, aggregate, and bulk soil) and plant (leaves, shoots, roots, and litter) δ¹⁵N; studies with climate gradients (space-time substitution) were excluded and only field warming experimental studies were included; only data from control and warming treatments were applied for multifactor experiments; means, standard deviation (SD) (or standard error (SE)), and sample sizes were directly provided or could be calculated from the studies; if one article contained soil or plant δ¹⁵N in multiple years, only the latest results were applied since the observations should be independent in the meta-analysis (Hedges et al., 1999).

2.2 Data extraction and statistical analysis

In total, 20 published papers were selected from 54 published papers. The locations of warming experiments were presented, and their site information is listed in Table 1. For each study, the means, the statistical variation (SE or SD), and the sample size values for treatment and control groups were extracted for each response variable (δ¹⁵N). In addition to δ¹⁵N, the latitude, longitude, altitude, soil pH, organic matter content, vegetation type, mean annual precipitation (MAP), and mean annual temperature (MAT) were also extracted if they were provided (Table 1). All data were extracted from tables or digitized from graphs with the software GetData v2.2.4 (http://www.getdata-graph-digitizer.com, last access: 17 October 2020). A total of 79 and 76 paired observations for soil and plant δ¹⁵N were obtained, respectively.

The MetaWin 2.1 software (Sinauer Associates Inc., Sunderland, MA, USA) (Rosenberg et al., 2000) was used to perform meta-analysis in this study. The Hedges' d value was used as the effect size (Hedges et al., 1999). The absolute d value indicated the magnitude of the treatment impact. Positive or negative d values represented an increase or decrease effect of the treatment, respectively. Zero meant no difference between treatment and control groups. Resampling tests were incorporated into our meta-analysis using the bootstrap method (999 random replicates). The mean effect size (calculated from 999 iterations) and 95 % bootstrap confidence interval (CI) were then generated. If the 95 % CI values of d did not overlap zero, the effects of experimental warming on δ¹⁵N were considered significant at p<0.05. We used a random effects model to test whether warming had a significant effect on δ¹⁵N. To examine whether experimental conditions alter the response direction and magnitude of soil and plant δ¹⁵N, observations were further divided into subgroups according to the soil acidity–alkalinity (acid (pH < 6.5), neutral (6.5 < pH < 7.5), and alkali (pH > 7.5)), vegetation types (forest/shrub, moss/lichen, and grassland/meadow), warming treatments (soil warming, air warming, and both soil and air warming), soil texture (fine-, medium-, and coarse-textured soil), length of warming (< 4, 4–10, and > 10 years), and increase in temperature (< 1.5, 1.5–3, and > 3 ^∘C). A random effects model with a grouping variable was used to compare responses among different subgroups. Linear regression analyses were applied to assess the relationships between the Hedges' d values and environmental factors (i.e., latitude, altitude, MAT, and MAP).

Figure 1 Effect sizes of the experimental warming on soil and plant δ¹⁵N from a global meta-analysis of 20 studies. The error bars indicate effect sizes and 95 % bootstrap confidence interval (CI). The warming effect was statistically significant if the 95 % CI did not bracket zero. The sample size for each variable is shown next to the bar.

3 Results

Across all sites, the mean effect sizes of the soil and plant δ¹⁵N under experimental warming were −0.524 (95 % CI: −0.987 to −0.162) and 0.189 (95 % CI: −0.210 to 0.569), respectively (Fig. 1). Experimental warming significantly (p<0.05) decreased soil δ¹⁵N in Alkali (mean effect size = −2.484; 95 % CI: −2.931 to −2.060) and medium-textured (mean effect size = −0.676; 95 % CI: −1.153 to −0.249) soils, in grassland/meadow (mean effect size = −0.609; 95 % CI: −1.076 to −0.190), under air warming (mean effect size = −0.652; 95 % CI: −1.081 to −0.273), for a 4–10-year warming period (mean effect size = −0.652; 95 % CI: −1.081 to −0.273) and for an increase of > 3 ^∘C in temperature (mean effect size = -0.652; 95 % CI: −1.081 to −0.273). However, it significantly (p<0.05) increased soil δ¹⁵N in neutral (mean effect size = 0.359; 95 % CI: 0.078 to 0.620) and fine-texture soils (mean effect size = 2.394; 95 % CI: 1.770 to 3.735), and for an increase of 1.5–3 ^∘C in temperature (mean effect size = 0.409; 95 % CI: 0.070 to 0.707) (Fig. 2). Experimental warming did not significantly (p > 0.05) change soil δ¹⁵N under other experimental conditions.

Figure 2 Factors influencing the effect sizes of the soil and plant δ¹⁵N under experimental warming from a global meta-analysis of 20 studies, including (a–b) soil acidity–alkalinity, (c–d) vegetation types, (e–f) warming treatments, (g–h) soil texture, (i–j) length of warming, and (k–l) increase in temperature. The error bars indicate effect sizes and 95 % bootstrap confidence interval (CI). The warming effect was statistically significant if the 95 % CI did not bracket zero. The sample size for each variable is shown next to the bar.

In addition, experimental warming significantly (p<0.05) increased plant δ¹⁵N in neutral (mean effect size = 3.157; 95 % CI: 1.529 to 6.967) and fine-textured soils (mean effect size = 1.202; 95 % CI: 1.042 to 1.360), whereas it significantly (p<0.05) decreased plant δ¹⁵N in alkali soil (mean effect size = −1.930; 95 % CI: −2.325 to −1.573) (Fig. 2). Experimental warming did not significantly (p > 0.05) change plant δ¹⁵N under other experimental conditions.

Figure 3 Relationships between the Hedges' d values of soil and plant δ¹⁵N with the latitude, altitude, mean annual temperature (MAT), and mean annual precipitation (MAP) under experimental warming.

For soil and plant δ¹⁵N, their responses to experimental warming did not correlate well with latitude (p=0.268 and p=0.160, respectively) (Fig. 3a, b). However, the Hedges' d values of soil δ¹⁵N decreased significantly with altitude (p<0.001) (Fig. 3c) and increased significantly with MAT (p<0.001) and MAP (p<0.001) (Fig. 3e, g). In addition, the Hedges' d values of plant δ¹⁵N were also found to increase significantly with MAP (p<0.001) (Fig. 3h). However, the responses of plant δ¹⁵N to experimental warming did not correlate well with altitude (p=0.109) and MAT (p=0.002) (Fig. 3d, f).

4 Discussion

A significant decreasing trend in soil δ¹⁵N and no significant trend in plant δ¹⁵N were found in this study. This is somewhat inconsistent with previous findings. Chang et al. (2017) observed that soil and plant δ¹⁵N values decreased under warming in the Tibetan permafrost. However, Zhang et al. (2019) found that the warming treatment significantly increased soil and plant δ¹⁵N in a subtropical forest. The various studies suggest that soil and plant δ¹⁵N are controlled by interactive effects of N fixation and mineralization. At the global scale, δ¹⁵N of N input (∼ 0) is generally lower than that of soil, so greater N fixation or higher N input (deposition and fertilization) under warming can result in a lower soil δ¹⁵N (Sorensen and Michelsen, 2011; Rousk and Michelsen, 2017; Wang et al., 2018).

Soil pH has an important influence on nitrification, denitrification, and N₂O emissions from soils (Kyveryga et al., 2004). The results in this study showed that when the soil was alkaline the mean effect sizes of soil and plant δ¹⁵N under warming were negative, and when the soil was neutral they were positive (Fig. 2a, b). Compared with alkaline condition, the near-neutral conditions are more suitable for the biological activities of heterotrophic denitrifying bacteria (Simek and Cooper, 2002). Therefore, the denitrification activity is usually higher under neutral conditions, resulting in an enrichment of soil and plant N pools with ¹⁵N (Kyveryga et al., 2004). Vegetation type had limited effects on δ¹⁵N under warming, except for soil δ¹⁵N in grassland/meadow (Fig. 2c, d). This may be related to the differences in altitude, MAP, and MAT among three vegetation types (Table 1). The type of warming treatment was found to have a substantial effect on soil δ¹⁵N, showing that the mean effect size of soil δ¹⁵N under air warming was negative and less than that under soil warming (Fig. 2e, f). Salmon et al. (2016) have found that soil warming can increase N availability by stimulating mineralization of organic matter in the warmed active layer. In addition, air warming directly impacts aboveground temperatures and has an indirect effect on soil δ¹⁵N (Pardo et al., 2006). From Fig. 2g and h, the finer the soil texture, the more significant the positive effect of warming on soil and plant δ¹⁵N. One possible reason is that the finer the soil texture, the stronger the adsorption of various ions on the soil and the smaller the leaching loss of the soil, resulting in greater residual amount of ¹⁵N in the soil (Webster et al., 1986). In addition, the longer warming period and the greater increase in temperature resulted in a more negative effect of warming on soil δ¹⁵N (Fig. 2i, k). Chang et al. (2017) deduced that N fixation was greater under warming and consequently resulted in a lower soil δ¹⁵N.

In the study of Mayor et al. (2015), soil and plant δ¹⁵N values were significantly (p<0.001) and negatively correlated with latitude at the global scale. However, the Hedges' d values of soil and plant δ¹⁵N had weak correlations with latitude in this study (Fig. 3). The warming effect on soil δ¹⁵N was significantly (p<0.001) influenced by altitude, MAT and MAP. Among these, the strongest correlation was observed for MAT. It is possible that soil δ¹⁵N increased with increasing MAT when the MAT exceeded a certain threshold (e.g., 9.8 ^∘C as proposed by Craine et al., 2015). In this case, the increase in MAT can enhance the positive effect of experimental warming on soil δ¹⁵N. In addition, the MAT can also affect ecosystem N cycle by influencing soil texture. Craine et al. (2015) reported that hot sites had greater clay concentrations than cold sites. As depicted in Fig. 2g, the finer the texture of the soil, the more significant the effect of experimental warming on soil δ¹⁵N. High d values of soil δ¹⁵N corresponded to MAT of about 20 ^∘C, which was the most suitable temperature for nitrification and denitrification. However, warming had a substantial negative impact on soil δ¹⁵N when MAT decreased to around −5 ^∘C. Recently, Rousk et al. (2018) also found that the increase in temperature in the Arctic promoted the biological N fixation, which can decrease the soil δ¹⁵N. The decrease of d values of soil δ¹⁵N with increasing altitude and decreasing MAP in this study might be caused by the positive response of d values to MAT.

The relationships between the d values and environmental variables for plant δ¹⁵N were weaker than those for soil δ¹⁵N (Fig. 3). One possible reason is that several other factors (e.g., plant N concentrations and species richness) might co-regulate plant δ¹⁵N (Wu et al., 2019). This is consistent with the study of Craine et al. (2009), who found different inflection points in soil and plant δ¹⁵N relationships with MAT. In addition, plants are generally depleted in ¹⁵N relative to soils. Above results implied that soil δ¹⁵N was more efficient in indicating the openness of ecosystem N cycling than plant δ¹⁵N at the global scale. Although the present study provided a global meta-analysis of the responses of δ¹⁵N to experimental warming, the magnitude of these responses might be uncertain. For example, a small number of observations were obtained in moss/lichen under soil warming and both soil and air warming treatments, which would affect the results of the meta-analysis. Future research should take more experimental data into account in order to better investigate the warming effects on δ¹⁵N.

5 Conclusions

Our global meta-analysis indicated a significant decreasing trend in soil δ¹⁵N and no significant trend in plant δ¹⁵N under experimental warming. Latitude did not affect the warming effects on δ¹⁵N. However, the warming effect on δ¹⁵N was related to soil acidity–alkalinity; texture; vegetation type; warming treatment and period; and increase in temperature, altitude, MAT, and MAP. The effect of warming on soil δ¹⁵N was better correlated with environmental variables compared with that on plant δ¹⁵N. Our findings should be useful for understanding the underlying mechanisms of the response of ecosystem N cycling to global warming.