Warming and altered precipitation rarely alter N addition effects on soil greenhouse gas fluxes: a meta-analysis

Abstract

Background

Changes in soil greenhouse gas (GHG) fluxes caused by nitrogen (N) addition are considered as the key factors contributing to global climate change (global warming and altered precipitation regimes), which in turn alters the feedback between N addition and soil GHG fluxes. However, the effects of N addition on soil GHG emissions under climate change are highly variable and context-dependent, so that further syntheses are required. Here, a meta-analysis of the interactive effects of N addition and climate change (warming and altered precipitation) on the fluxes of three main soil GHGs [carbon dioxide (CO₂), methane (CH₄), and nitrous oxide (N₂O)] was conducted by synthesizing 2103 observations retrieved from 57 peer-reviewed articles on multiple terrestrial ecosystems globally.

Results

The interactive effects of N addition and climate change on GHG fluxes were generally additive. The combination of N addition and warming or altered precipitation increased N₂O emissions significantly while it had minimal effects on CO₂ emissions and CH₄ uptake, and the effects on CH₄ emissions could not be evaluated. Moreover, the magnitude of the combined effects did not differ significantly from the effects of N addition alone. Apparently, the combined effects on CO₂ and CH₄ varied among ecosystem types due to differences in soil moisture, which was in contrast to the soil N₂O emission responses. The soil GHG flux responses to combined N addition and climate change also varied among different climatic conditions and experimental methods.

Conclusion

Overall, our findings indicate that the effects of N addition and climate change on soil GHG fluxes were relatively independent, i.e. combined effects of N addition and climate change were equal to or not significantly different from the sum of their respective individual effects. The effects of N addition on soil GHG fluxes influence the feedbacks between climate change and soil GHG fluxes.

Background

Carbon dioxide (CO₂), methane (CH₄), and nitrous oxide (N₂O), as the three major greenhouse gases (GHGs) (World Meteorological Organization 2020), have been observed to increase over the past several decades, mainly originating from anthropogenic activities, including fossil fuel combustion, land use change, and chemical fertilizer application (IPCC 2013). Another consequence of the anthropogenic activities is increased nitrogen (N) loads in terrestrial ecosystems (Du et al. 2021; He et al. 2021). N addition can alter the biogeochemical processes of ecosystems and alter soil GHG fluxes (Li et al. 2020; Liu and Greaver 2009), further contributing to changes in both temperature and the hydrological cycle (IPCC 2022). As two key factors regulating terrestrial ecosystem processes, temperature and precipitation influence soil microclimate, microbial activity, and soil substrate availability (Harte et al. 1995; Zhang et al. 2019; Zhou et al. 2018), thus potentially influencing the effects of N addition on soil GHG fluxes (Chen et al. 2021). Therefore, understanding how climate change regulates the feedback between soil GHGs and N addition would improve our capacity to predict future soil GHG fluxes and such potential effects could be incorporated into predictive biogeochemical models.

Globally, researchers have experimentally examined the effects of N addition on soil CO₂, CH₄, and N₂O fluxes. The main underlying mechanisms by which N addition increases soil GHG fluxes are the regulation of plant growth and microbial activities directly associated with soil GHG production and consumption processes (Chen et al. 2021; Niu et al. 2010; Quinn Thomas et al. 2010). In addition, projected shifts in temperature have been documented to enhance microbial activity thereby accelerating soil organic matter decomposition and stimulating soil GHG release (Liu et al. 2020; Zhang et al. 2020). Therefore, warming and N addition can interact to synergistically increase soil GHG fluxes (Chen et al. 2017a; Yang et al. 2022). However, warming can also negate the positive effects of N addition on soil GHGs as warming-induced soil water deficits can suppress microbial activity and plant physiological activity (Zong et al. 2018; Cheng et al. 2022). In addition, the effects of N addition on soil GHG fluxes may be suppressed under decreased precipitation due to reduced plant carbon (C) inputs and limited soluble C substrate availability (Fuchslueger et al. 2014; Aronson et al. 2019). Conversely, higher soil water availability caused by increased precipitation can increase soil labile C availability and microbial activity (Zhou et al. 2016). The findings suggested that stimulated GHG emissions by N addition may particularly occur under high precipitation regimes (Brown et al. 2012).

The effects of N addition in combination with warming or altered precipitation remain ambiguous due to differences in outcomes across ecosystem types, climatic conditions, and experimental methods (Deng et al. 2020; Gong and Wu 2021). Among individual studies, the soil CH₄ fluxes’ responses to N addition in combination with warming vary from potential increase to potential decrease (Chen et al. 2017a, b; Wu et al. 2020; Liu et al. 2015). The directions of soil CH₄ fluxes are highly dependent on ecosystem type and soil condition, as soil CH₄ is produced by methanogens in wet anaerobic soils and consumed by methanotrophs in drier aerobic soils (Le Mer and Roger 2001; Ni and Groffman 2018). In addition, the responses of CO₂ to N addition in combination with climate change are dependent on ecosystem type (Lu et al. 2011). The potential mechanisms that alter CO₂ emissions are (a) transfer of nutrients that stimulate microbial growth and respiration and (b) priming effects on SOM decomposition by microorganisms (Fontaine et al. 2007; Janssens et al. 2010; Oertel et al. 2016). For example, N addition in combination with climate change can have a greater impact on soil CO₂ flux in grasslands than in forests due to higher stimulation of autotrophic respiration in grasslands (Zhou et al. 2014). Climate condition is the another key determinant of differences in reported effects in empirical studies. Previous studies have reported that N addition plus increased precipitation at a dry or mesic site could stimulate GHG emissions, as opposed to a wet site, where responses are neutral or even negative (Shi et al. 2021; Wu et al. 2020). The results may be because dry regions are more sensitive to increased precipitation than moist regions. In addition, experimental methods such as the duration and magnitude of N addition, warming, and altered precipitation can drive differences in responses (Sánchez-Martín et al. 2017; Zhang et al. 2008). No overall conclusions have been reached regarding the combined effects of N addition and warming or altered precipitation, limiting our mechanistic understanding of the responses of soil GHG fluxes to global changes. A meta-analysis across multiple ecosystem types with heterogeneous climatic conditions could facilitate adequate assessment of the effects of global change on GHG emissions at the global scale.

The aims of the present study were to (1) determine how warming and altered precipitation contribute to the effects of N addition on soil GHG emissions; (2) evaluate the interactions between N addition and warming or altered precipitation; and (3) explore the major drivers of the context-dependency observed in empirical studies. Accordingly, a meta-analysis of 2103 observations from published articles that reported both single-driver (N addition, warming, increased precipitation, and decreased precipitation) and corresponding two-driver effects on soil GHG fluxes was conducted. The observations were mainly collected from grassland and forest ecosystems in the Northern Hemisphere, particularly China. We hypothesized that (1) warming and increased precipitation will enhance the positive effect of N addition on soil GHG fluxes, while decreased precipitation will inhibit it and (2) these effects vary across different ecosystems and climatic conditions.

Materials and methods

Data compilation

Peer-reviewed articles published before September 2022 were searched in Web of Science (http://apps.webofknowledge.com/) and China National Knowledge Infrastructure (https://www.cnki.net/) using the following search string: (“nitrogen addition” OR “nitrogen deposition” OR “nitrogen fertilization”) AND (warming OR precipitation OR rainfall OR drought) AND (“greenhouse gas” OR CO₂ OR “carbon dioxide” OR CH₄ OR methane OR N₂O OR “nitrous oxide”). The inclusion criteria were as follows: (1) experiments were conducted in the field and the impacts of N addition and at least one of the climate change factors (warming, increased, or decreased precipitation) were recorded, including both their single-driver and corresponding two-driver effects; (2) at least one variable among CO₂ emissions, CH₄ emissions, CH₄ uptake, and N₂O emissions was measured; (3) except for climate change factors and N addition, the conditions of the control and treatment groups were the same; and (4) the means, standard deviations (or standard errors), and sample sizes of the variables were directly provided or could be calculated.

The experiments manipulating two or more treatments, including different climate change factors or magnitudes with the same control, were divided into different control–treatment pairs. If the individual and combined effects on GHGs in a particular study were observed several times at different experimental stages, we used all observations to facilitate the evaluation of experimental duration effects. Data were directly extracted from the main text, table, or appendices, or from figures using Engauge Digitizer version 11.2 (Free Software Foundation, Boston, MA, USA). The ecosystem types (desert, cropland, wetland, forest, and grassland), experimental methods (experimental magnitude and experimental duration), climate conditions [mean annual temperature (MAT) and mean annual precipitation (MAP)], and edaphic characteristics (pH, water content, C content, and N content) were extracted from the studies to assess the impacts of moderator variables. If the MAT or MAP of the experimental sites was not provided in the studies, we extracted them from the WorldClim database (https://www.worldclim.org/).

We calculated the De Martonne (1926) aridity index for each experimental site using the MAT and MAP data as follows:

$$\mathrm{Aridity\,index}=\frac{\mathrm{MAP}}{\mathrm{MAT}+10}$$

(1)

A total of 2103 paired observations were collected from 57 articles. CH₄ emissions were not included owing to the small number of studies. In these articles, 2, 3, 6, 9, and 37 studies were conducted in deserts, croplands, wetlands, forests, and grasslands, respectively, accounting for 4%, 5%, 10%, 16%, and 65% of the studies, respectively (Additional file 1: Table S1). The form of N addition in 92% of the experiments was NH₄NO₃ and 8% were urea, CaNO₃, and Ca(NO₃)₂. The latitude ranged from 26.32°N to 48.25°N, the altitude from 20 to 4763 m, the MAT from –5.3 °C to 19.1 °C, and the MAP from 200 to 1850 mm (Fig. 1).

Fig. 1

Global distribution of observations derived from 57 articles used in the meta-analysis. The different treatment types are indicated by symbol shape, and the ecosystem types are represented by color

Statistical analysis

Individual and combined effects

To quantify the individual and combined effects of N addition, warming, and altered precipitation on soil GHG fluxes, we calculated the response ratio ($\mathrm{ln}RR$) as a proxy for effect size as follows:

$$\mathrm{ln}RR=\mathrm{ln}\left(\frac{{\overline{X} }_{t}}{{\overline{X} }_{c}}\right)$$

(2)

where ${\overline{X} }_{t}$ and ${\overline{X} }_{c}$ are the mean values of the treatment and control groups, respectively. The variance ($v$) of each $\mathrm{ln}RR$ was calculated as follows:

$$v=\frac{{S}_{t}^{2}}{{n}_{t}{\overline{X} }_{t}^{2}}+\frac{{S}_{c}^{2}}{{n}_{c}{\overline{X} }_{c}^{2}}$$

(3)

where ${\overline{X} }_{t}$, ${S}_{t}$, and ${n}_{t}$ represent the mean, standard deviation, and sample size of the treatment group (t), respectively; and ${\overline{X} }_{c}$, ${S}_{c}$, and ${n}_{c}$ are the mean, standard deviation, and sample size of the control group (c), respectively.

To estimate the overall effect size ($\mathrm{ln}{RR}_{++}$), we performed linear mixed-effects models using the “lme4” package (Bates et al. 2015) in R version 4.2.0 (R Core Team 2022) with the individual $\mathrm{ln}RR$ fitted as the response variable and the identity of the study and plots (nested inside study identity) as the random effect factors (Additional file 2). This approach explicitly accounts for the potential dependence of observations collected from a single study. To investigate whether the ecosystem type, climate (MAT and MAP), edaphic characteristics (moisture, pH, C content, and N content), and experimental methods (N addition rate, warming magnitude, and duration) influence the response of GHGs to N addition and the interaction with climate change, these factors were fitted as continuous or categorical fixed-effect factors. Data were subjected to z-score standardization prior to analysis to allow for comparisons (Ali and Faraj 2014). z-scores were determined by standardization by subtracting the mean value from the raw data and dividing the resulting number by the standard deviation of the mean. The corresponding 95% confidence interval ($\mathrm{CI}$) and percentage change (%) were calculated as follows:

$$95\mathrm{\% }{\mathrm{CI}}_{S}=\mathrm{ln}{RR}_{++}\pm 1.96{S}_{\mathrm{ln}{RR}_{++}}$$

(4)

$$\mathrm{Percentage\,change}=\left({e}^{\mathrm{ln}{RR}_{++}}-1\right)\times 100$$

(5)

The treatment effects were considered non-significant if the 95% $\mathrm{CI}$ of $\mathrm{ln}{RR}_{++}$ overlapped with zero. When the 95% $\mathrm{CI}$ of $\mathrm{ln}{RR}_{++}$ did not overlap with zero, the effect was identified as positive if $\mathrm{ln}{RR}_{++}$ was greater than zero; otherwise, the effect was identified as negative if $\mathrm{ln}{RR}_{++}$ was less than zero.

Furthermore, publication bias was assessed to analyze whether the studies used in the present meta-analysis were representative or not using Egger’s regression tests (Egger et al. 1997). Funnel plots were generated using the response ratios ($\mathrm{ln}RR$) and their standard errors. All results indicated that studies used in this meta-analysis were robust against publication bias (Additional file 1: Fig. S1).

Interactive effects

To further understand the interactive effects of two drivers occurring simultaneously (combined effects), Hedges’ $d$ was used to calculate interactive effect size (${d}_{I}$) as previously described (Gurevitch et al. 2000):

$${d}_{I}=\frac{\left({\overline{X} }_{AB}-{\overline{X} }_{A}\right)-\left({\overline{X} }_{B}-{\overline{X} }_{c}\right)}{s}J\left(m\right)$$

(6)

where ${\overline{X} }_{c}$, ${\overline{X} }_{A}$, ${\overline{X} }_{B}$, and ${\overline{X} }_{AB}$, are the means of the variables in the control group, treatment groups A and B, and their combination (A + B), respectively.

The standard deviation (${s}$), degrees of freedom ($m$), and correction term (${J}\left(m\right)$) were estimated using Eqs. (7–9).

$$\mathrm{s}=\sqrt{\frac{\left({n}_{c}-1\right){\left({s}_{c}\right)}^{2}+\left({n}_{A}-1\right){\left({s}_{A}\right)}^{2}+\left({n}_{B}-1\right){\left({s}_{B}\right)}^{2}+\left({n}_{AB}-1\right){\left({s}_{AB}\right)}^{2}}{{n}_{c}+{n}_{A}+{n}_{B}+{n}_{AB}-4}}$$

(7)

$$m={n}_{c}+{n}_{A}+{n}_{B}+{n}_{AB}-4$$

(8)

$$\mathrm{J}\left(m\right)=1-\frac{3}{4m-1}$$

(9)

where ${n}_{c}$, ${n}_{A}$, ${n}_{B}$, and ${n}_{AB}$ are the sample sizes and ${s}_{c}$, ${s}_{A}$, ${s}_{B}$, and ${s}_{AB}$ are the standard deviations in the control and treatment groups of A, B, and their combination (A + B), respectively.

The variance (${v}_{2}$) of ${d}_{I}$ was estimated as follows:

$${v}_{2}=\frac{1}{4}\left[\frac{1}{{n}_{C}}+\frac{1}{{n}_{A}}+\frac{1}{{n}_{B}}+\frac{1}{{n}_{AB}}+\frac{{d}_{I}^{2}}{2\left({n}_{C}+{n}_{A}+{n}_{B}+{n}_{AB}\right)}\right]$$

(10)

We performed linear mixed-effects models to calculate the weighted mean of ${d}_{I}$ (${d}_{++}$) with ${d}_{I}$ as the response variable and study identity and plots (nested inside study identity) as the random effect factors. All the analyses were conducted in R version 4.2.0 (R Core Team 2022) using the ‘lme4’ package (Bates et al. 2015).

The interactions between treatments were classified into three types: additive, synergistic, and antagonistic (Folt et al. 1999; Crain et al. 2008). Additive effects occurred when their combined effects were not significantly different from the sum of their individual effects; synergistic interactions were observed when the combined effect was significantly greater than the sum of their individual effects and antagonistic interactions when the combined effect was less than the sum of their individual effects. In terms of Hedges’ $d$, interactions were considered additive if the 95% $\mathrm{CI}$ of the weighted mean ${d}_{I}$ overlapped with zero. If the 95% $\mathrm{CI}$ did not overlap with zero, we determined the interaction effect through their individual effects. If the individual effects of the two drivers were both negative or one negative and one positive, interaction effect sizes less than zero were considered synergistic, and those greater than zero were antagonistic. In cases where the individual effects were both positive, interaction effect sizes greater than zero were synergistic, and those less than zero were antagonistic.

Results

Overall responses of soil GHG fluxes

Overall, soil CO₂ emissions did not respond significantly to N addition alone, but it increased by an average of 17.2% and 21.7% under warming alone and warming plus N addition, respectively (Fig. 2a). N addition alone or in combination with warming or altered precipitation minimally affected soil CH₄ uptake, while warming alone significantly increased CH₄ uptake by an average of 28.2% (Fig. 2b). In contrast, N₂O emissions increased significantly by 106.2%, 112.2%, 108.2%, and 105.2% under N addition alone, warming plus N addition, increased precipitation plus N addition, and decreased precipitation plus N addition, respectively. However, individual climate change treatments did not affect soil N₂O emissions, except for decreased precipitation, which significantly decreased N₂O emissions by 40.6% (Fig. 2c).

Fig. 2

Effects of N addition (N), warming (W), increased precipitation (IP), and decreased precipitation (DP) alone or N addition combined with warming (W + N), increased precipitation (IP + N), and decreased precipitation (DP + N), on soil CO₂ emissions (a), CH₄ uptake (b), and N₂O emissions (c). Data are means with 95% confidence intervals ($\mathrm{CI}$). The number of observations is shown along the right axis. Blue, orange, and gray solid dots indicate significantly positive, significantly negative, and non-significant effects, respectively. The vertical dashed lines are zero lines

Variation among ecosystem types

The effects of N addition alone on CO₂ emissions were independent of ecosystem type, whereas the effects of N addition plus warming or altered precipitation were different among ecosystem types, significantly affecting grasslands but not forests (Fig. 3a). The individual and combined effects of N addition and warming positively impacted soil CH₄ uptake in grasslands, but their effects were negative in croplands (Fig. 3b). The responses of soil N₂O emissions to various treatments were independent of ecosystem type (Fig. 3c). Based on the results of linear mixed-effects models, the most important regulator of the differences among ecosystem types was soil moisture rather than soil pH, and initial soil C and N content (Table 1). Soil moisture had significantly negative effects on the $\mathrm{ln}RR$ of CO₂ emissions to N addition plus warming, increased precipitation alone, and N addition plus increased precipitation, and the $\mathrm{ln}RR$ of CH₄ uptake to warming alone and N addition plus warming.

Fig. 3

Influence of ecosystem type on the individual and combined effects of N addition (N), warming (W), increased precipitation (IP), and decreased precipitation (DP) on soil CO₂ emission (a), CH₄ uptake (b), and N₂O emission (c). The p values (p < 0.05) indicate the significance of the differences in the changes in GHG fluxes among ecosystems. Data are means with 95% confident intervals ($\mathrm{CI}$). The numbers of observations are shown on the right. Blue, orange, and gray solid dots indicate significantly positive, significantly negative, and non-significant effects, respectively. The vertical dashed lines are zero lines

Table 1 Linear mixed-effects model analysis of the relationship between local climate, experimental, or edaphic conditions and the effects of N addition (N), warming (W), increased precipitation (IP), decreased precipitation (DP), N addition plus warming (W + N), N addition plus increased precipitation (IP + N), and N addition plus decreased precipitation (DP + N) on soil GHG fluxes

Effects of climate and experimental methods

The $\mathrm{ln}RR$ of GHG fluxes to N addition plus warming or altered precipitation was significantly influenced by climate and experimental methods (Table 1). Aridity index was the most important driver of the combined effects of N addition and increased precipitation on soil CH₄ uptake, showing significant negative effects. The $\mathrm{ln}RR$ of CO₂ and N₂O under warming plus N addition was positively affected by the magnitude of warming. Altered precipitation magnitude, N addition rate (Additional file 1: Fig. S2), and experimental duration had non-significant effects on the $\mathrm{ln}RR$ of GHG fluxes.

Interactions between N addition and climate change

Across all two-driver pairs, the overall interactive effects of N addition and warming or altered precipitation on GHG fluxes were generally additive, except for the antagonistic interaction between decreased precipitation and N addition on N₂O emissions (Fig. 4). Additive interactions accounted for a higher proportion than synergistic or antagonistic effects among individual pairwise observations, representing 26–54%, 60–75%, and 41–67% of the combined effects of warming plus N addition, increased precipitation plus N addition, and decreased precipitation plus N addition, respectively.

Fig. 4

Interactions of N addition (N), warming (W), increased precipitation (IP), and decreased precipitation (DP) on soil CO₂ emission (a), CH₄ uptake (b), and N₂O emission (c) (left), and corresponding frequency distribution of interaction types across individual pairwise observations (right). Solid circles indicate means with ± 95% confidence intervals (CI). The number of observations is shown along the right axis. Interactive effects were additive (gray, where 95% CIs overlap zero); otherwise, effects were synergistic (blue) or antagonistic (orange)

Discussion

Responses of soil CO₂ emissions

The combined effects of N addition and warming or increased precipitation raised CO₂ emissions more than N addition alone, partially supporting our first hypothesis. Soil CO₂ emissions did not respond significantly to N addition alone, but it increased significantly under warming alone and increased precipitation alone. Therefore, the increases in CO₂ emissions were mainly attributed to the effect of warming and increased precipitation, rather than N addition. These results suggested that temperature and precipitation are key drivers regulating soil CO₂ emissions, which is consistent with the previous studies (Deng et al. 2020; Yang et al. 2022). Moreover, the overall interactions between N addition and warming or increased precipitation were additive, indicating that combined N addition and warming or increased precipitation on CO₂ emissions were not substantially different from the sum of their respective individual effects. Therefore, the effects of N addition and climate change factors on CO₂ emissions were relatively independent. The non-significant effects of N addition on CO₂ emissions revealed in our meta-analysis may be attributed to the possibility that the level of N addition reached or exceeded the N critical load due to continuous N enrichment (Zong et al. 2016; Bai et al. 2019). Various studies have suggested that N addition initially stimulated soil respiration but gradually suppressed CO₂ emissions (Xia et al. 2009; Yan et al. 2010). Moreover, Hui et al. (2020) have demonstrated that the absence of other nutrients, such as phosphorus, may suppress the effect of N addition. Therefore, the potential mechanisms by which N addition influenced ecosystem CO₂ emission and the long-term effects of continuous N enrichment should be further explored.

Consistent with our second hypothesis, the effects of N addition in combination with warming or increased precipitation were influenced by climatic condition. Aridity index primarily regulated CO₂ emissions under altered precipitation plus N addition (Table 1). Increased precipitation alleviated water shortages in arid areas and promoted soil respiration (Sierra et al. 2017), while decreased precipitation exacerbated the dry conditions, strongly affecting soil biotic processes by limiting water availability, leading to a decrease in CO₂ emissions (de Dato et al. 2010). Differences in the soil moisture may explain the various responses of CO₂ emissions to N addition plus increased precipitation between forests and grasslands, as forests have higher soil moisture than grasslands (Additional file 1: Fig. S3). Moreover, the combined effect of N addition plus warming on soil CO₂ emissions was positively correlated with warming magnitude, indicating that these combined effects were stimulated by warming (Carey et al. 2016). Although climatic condition has been observed to be a major factor influencing differences in the soil CO₂ emissions of forests and grasslands, the lack of observations from other ecosystems still contributes to our poor understanding of soil CO₂ emissions among different ecosystems globally. Therefore, we propose more multifactorial experiments in various ecosystems in future work.

Responses of soil CH₄ uptake

In contrast to CO₂ emissions, the effects of N addition on soil CH₄ uptake remained unaltered when combined with warming or altered precipitation. Additive interactions between N addition and warming or altered precipitation on soil CH₄ uptake appeared to be much more common compared with synergistic and antagonistic interactions. Warming increased soil CH₄ uptake significantly by increasing the biomass of methane-oxidizing and decreasing the biomass of methanogenic microorganisms (Qi et al. 2022). However, warming in combination with N addition suppressed this significantly positive effect, which may be related to the increase in soil acidity (Bowman et al. 2008; Jamil et al. 2022). Therefore, the response of soil CH₄ uptake to warming plus N addition was negatively related to the N addition rate.

We found that the response of soil CH₄ uptake to N addition combined with warming was negatively correlated with higher soil moisture content. Furthermore, the effect of N addition plus increased precipitation was negatively correlated with aridity index. These findings may be related to the CH₄ production and consumption processes of soil microorganisms. Soil CH₄ is produced by methanogens in wet anaerobic soils and consumed by methanotrophs in drier aerobic soils (Fest et al. 2017; Liu et al. 2019); therefore, warming and increased precipitation could enhance the emissions of soil CH₄ by increasing the activities of methanogenic archaea and methanotrophic bacteria in humid areas (Aronson et al. 2013; Kammann et al. 2001; Zhao et al. 2017). This could also explain the significant differences in soil CH₄ uptake between grasslands and croplands. Due to observational limitations, we could not analyze the effects of altered precipitation plus N addition on CH₄ uptake among different ecosystem types; more experiments conducted in various ecosystems are required to predict CH₄ emissions under future climate change scenarios.

Responses of soil N₂O emissions

The increases in N₂O emissions caused by N addition plus warming or altered precipitation did not differ from that of N addition alone, while warming and increased precipitation alone had non-significant effects. This implies that temperature and precipitation are not major factors affecting soil N₂O emissions (Yang et al. 2021), which was inconsistent with the results of Zhang et al. (2021) and Huang et al. (2014). N addition was an important factor influencing soil N₂O emissions because soil N₂O released as an intermediate product when ammonium (NH₄⁺) is oxidized to nitrate (NO₃^–) by nitrification, or during the reduction of NO₃^– or nitrite (NO₂^–) by denitrification (Snyder et al. 2009; Hube et al. 2017). Therefore, soil N₂O emissions are mainly limited by soil N availability (Gao et al. 2015). Increased N supply can stimulate soil N₂O emissions directly by promoting the proliferation of microbial communities involved in nitrification and denitrification (Pajares and Bohannan 2016). Similar findings were reported by Deng et al. (2020) and Du et al. (2021), they found that N addition increased soil N₂O emissions by 164% and 91%, respectively, which were consistent with our result (+ 106%). In the present study, an overall antagonistic interaction between N addition and decreased precipitation on N₂O emissions was observed. In this case, the negative effects of decreased precipitation on N₂O emissions would be counteracted by N addition in the combined effects (Geng et al. 2017). However, overall interactions of N addition and warming or increased precipitation on N₂O emissions were generally additive, indicating that the effect of N addition, warming, and increased precipitation on N₂O emissions would not be changed in the combined effects. Although synergistic and antagonistic interactions were observed among the individual observations, additive interactions remained predominant. This finding may be attributed to the interactions between multiple region-dependent global change factors, where different climatic conditions and ecosystem types generate various effects (Baah-Acheamfour et al. 2016; Liu and Greaver 2009).

The responses of soil N₂O emissions to individual and combined effects of N addition, warming, and altered precipitation did not vary significantly among ecosystem types. These results indicated that ecosystem type had only a minor impact on soil N₂O emissions (Deng et al. 2020; Li et al. 2020). In addition, the positive effects of N addition combined with warming or altered precipitation on soil N₂O emissions were hardly affected by climatic and experimental conditions. Soil N₂O is consistently released rapidly at the beginning of N addition, irrespective of the experimental duration (Li et al. 2022; Xu et al. 2017; Song and Niu 2022). These observations suggested that N addition is the major driver of increased soil N₂O emissions and the feedbacks between them are not altered by climate change.

Main limitations and future perspectives

We found that soil GHG fluxes could be stimulated to some extent by multiple global change factors, including N addition, warming, and altered precipitation. The responses of soil GHG fluxes to N addition were not significantly affected by warming or altered precipitation, as illustrated by the additive interaction effects. Our study provides a comprehensive analysis of soil GHG fluxes by considering the effects of N addition under joint warming or altered precipitation in multiple terrestrial ecosystems. Nevertheless, we acknowledge some limitations of our study. First, the compiled data were mainly from studies in grassland of Northern Hemisphere, particularly China (Fig. 1), with other regions of the world being poorly represented, which possibly led to a misrepresentation of global change effects on soil GHG fluxes. In addition, multiple subtypes existed in these broad categories of ecosystem types that could not be categorized due to the limited observations, which hampered our ability to draw robust and general conclusions on the effects of global change on soil GHG fluxes among different ecosystems. Second, our results showed that decreased precipitation plus N addition did not affect soil CH₄ uptake, but increased N₂O emissions significantly. However, the sample size of the combined effects was insufficient and only represented grassland ecosystems, which may contribute to the uncertainty of our estimations. Third, in most experiments N addition was applied in the form of NH₄NO₃, yet research has shown that the response of soil GHG fluxes to N addition is influenced significantly by N form (Du et al. 2021). To improve our understanding of how GHG fluxes will respond to global climate change scenarios, future multi-factor experimental studies should focus on underrepresented regions, especially the effects of altered precipitation and N addition on soil CH₄ and N₂O.

Conclusion

This study examined the effects of N addition combined with warming and altered precipitation on GHG fluxes at a multi-continental scale. Warming and altered precipitation rarely influenced the effects of N addition on soil GHG fluxes, both in magnitude and direction. Their overall interactions on GHG fluxes were generally additive (i.e., not differing from the sum of their individual effects) rather than synergistic or antagonistic, suggesting the relative independence of these factors. Soil CO₂ emissions were mainly regulated by temperature and precipitation rather than by N addition, whereas N₂O emissions were overall limited by N addition. The effects of N addition plus increased precipitation on soil CO₂ emissions and the effects of N addition plus warming on soil CH₄ uptake showed contrasting results among different ecosystems, which were mainly related to soil moisture. In contrast, the soil N₂O emissions did not significantly differ among ecosystems. The individual and combined effects of these treatments on soil GHG fluxes were regulated by climate and experimental methods. Overall, our results not only quantitatively synthesized the patterns of N addition and climate change on soil GHG fluxes, but also showed that the effects of N addition, warming, and altered precipitation were relatively independent. These findings advance current understanding of the response of soil GHG fluxes to N addition in combination with warming and altered precipitation and provide insights into C and N cycling under global climate change.