Anthropogenic nitrogen (N) enrichment of ecosystems, mainly from fuel combustion and fertilizer application, alters biogeochemical cycling of ecosystems in a way that leads to altered flux of biogenic greenhouse gases (GHGs). Our meta-analysis of 313 observations across 109 studies evaluated the effect of N addition on the flux of three major GHGs: CO2, CH4 and N2O. The objective was to quantitatively synthesize data from agricultural and non-agricultural terrestrial ecosystems across the globe and examine whether factors, such as ecosystem type, N addition level and chemical form of N addition influence the direction and magnitude of GHG fluxes. Results indicate that N addition increased ecosystem carbon content of forests by 6%, marginally increased soil organic carbon of agricultural systems by 2%, but had no significant effect on net ecosystem CO2 exchange for non-forest natural ecosystems. Across all ecosystems, N addition increased CH4 emission by 97%, reduced CH4 uptake by 38% and increased N2O emission by 216%. The net effect of N on the global GHG budget is calculated and this topic is reviewed. Most often N addition is considered to increase forest C sequestration without consideration of N stimulation of GHG production in other ecosystems. However, our study indicated that although N addition increased the global terrestrial C sink, the CO2 reduction could be largely offset (53–76%) by N stimulation of global CH4 and N2O emission from multiple ecosystems.
Anthropogenic nitrogen (N) enrichment and global warming are primary components of global change. The annual input of anthropogenic reactive nitrogen, mainly through intensive N fertilizer application and fossil fuel combustion, has increased more than 10-fold in the last 150 years and N atmospheric deposition rates are predicted to increase another two- or threefold in the coming years (Galloway & Cowling 2002). N enrichment causes a suite of detrimental effects on ecosystem services that are associated with depleting soil minerals, acidifying soil, altering community structure and eutrophying aquatic ecosystems (Vitousek et al. 1997; U.S. EPA 2008). There is mounting evidence that N addition to ecosystems alters physiology of soil microbes and vegetation in a way that leads to altered biogenic flux of the three main greenhouse gases (GHGs) contributing to global warming: CO2, CH4 and N2O (Butterbach-Bahl et al. 1998; Dalal et al. 2003; Bodelier & Laanbroek 2004; Mosier et al. 2006). Therefore, N addition may alter loading of GHGs to the atmosphere and, perhaps, further contribute to global warming. Understanding how N addition affects the biogenic GHG budget is critically important, from both scientific and political perspectives, for understanding global warming under changing environmental conditions and contributing to the scientific foundation for determining allowable limits of anthropogenic N, such as the U.S. National Ambient Air Quality Standards and Water Quality Standards.
The N addition alters fluxes of GHGs through regulating plant and microbial activities that are directly associated with GHG production and consumption processes (Fig. 1). It is extremely important to consider total ecosystem flux because consumption may be offset by production, even within the same ecosystem. For example, N addition stimulates plant growth in most ecosystems (LeBauer & Treseder 2008), which may in turn increase C sequestration in plant biomass. On the other hand, maintenance respiration is positively correlated with tissue N content (Reich et al. 2008), and litter with higher N content also decomposes faster (Berg & Laskowski 2006). Therefore, increased leaf N content under elevated N may result in higher C loss by increasing both autotrophic and heterotrophic respiration. Like CO2, ecosystems may consume and produce CH4 and N2O, and the balance determines their net release.
Methane is produced in anaerobic soils by methanogenic archaea during organic decomposition and consumed in aerobic soils via oxidation by methanotropic bacteria (Le Mer & Roger 2001). The activities of methanogenic coenzymes are optimally active at low redox potentials (≤ 200 mV). Nitrate (NO3−) can decrease CH4 production by increasing redox potentials (Le Mer & Roger 2001). Both CH4 and ammonium (NH4+) can be oxidized by methane monooxygenase (MMO), and NH4+ therefore usually inhibits CH4 oxidation by competing for MMO (Bodelier & Laanbroek 2004).
Nitrous oxide production in the soil is mainly governed by microbial nitrification and denitrification (Dalal et al. 2003). Denitrifying bacteria produce N2O during the reduction of NO3− or NO2− under anaerobic conditions. In an aerobic environment, N2O is released as an intermediate product when nitrifying bacteria oxidize NH4+ to NO3− (Dalal et al. 2003). The increase in N2O emission following NO3− or NH4+ addition was observed in many experiments, mainly attributed to the increased N supply to nitrifying and denitrifying bacteria (Aerts & Toet 1997; Butterbach-Bahl et al. 1998; Aerts & de Caluwe 1999; Keller et al. 2005). Denitrification is assumed to be the major microbial process responsible for N2O consumption by reducing N2O to N2. Low mineral N and low oxygen pressure appear to favour N2O consumption. However, the mechanisms controlling N2O consumption are still not well understood (Chapuis-Lardy et al. 2007). N enrichment could also affect microbial activities positively by increasing C substrates supply or negatively by increasing soil toxicity (Treseder 2008). Because both GHG production and consumption microbes are affected equally, the net effect on net GHG budget is highly dependent on local environmental conditions (Fig. 1).
In this study, we report the results of a meta-analysis that quantitatively synthesizes the available research on changes in GHG flux (CO2, CH4 or N2O) from N addition in multiple terrestrial ecosystems, including agricultural crop, forest, grassland, wetland, tundra, heathland and desert. We investigated whether the direction and magnitude of GHG flux differs by (1) ecosystem type, (2) the level of N loading, (3) chemical species of N addition and (4) experimental condition.
The changes in biogenic CO2, CH4 and N2O fluxes were assessed by different ecosystem response variables. Three response variables were selected to evaluate the effect of N on ecosystem C budget: net ecosystem CO2 exchange (NEE) for non-forest natural ecosystems; ecosystem C content (EC) for forest ecosystems and soil organic carbon (SOC) for agricultural ecosystems. Our analysis of N addition effects on NEE, the net C gain or loss from an ecosystem, is limited to non-forest natural ecosystems because there were insufficient publications available to analyse forest and agricultural ecosystems. Several studies have assessed the effect of N fertilization on forest EC, here defined as the sum of C content of vegetation, forest floor and soil (Johnson et al. 2006). Therefore, we use EC to evaluate forest ecosystems. The average ratio of global annual cereal production to utilization is 1 : 1.005 from 1997 to 2007 (FAO 2008a), suggesting C fixed in cereal is offset by C consumed in the same year. Therefore, we use SOC to estimate C content change in agricultural ecosystems. The changes in CH4 source or sink strength were evaluated separately by including publications that examined the effect of N addition on CH4 emission or CH4 uptake. In contrast, we only evaluated N2O emission, not uptake. Although terrestrial ecosystems can be N2O sinks (Chapuis-Lardy et al. 2007), few publications assessed the effects of N addition on N2O uptake. In total, six response variables, NEE, EC, SOC, CH4 emission, CH4 uptake and N2O emission, were included to characterize the effects of N addition on the GHG flux.
The meta-analysis model requires each observation to be independent (Hedges et al. 1999). For studies which had several measurements through time at the same N addition level, overall means were included as a single observation. Measurements from different N addition levels, N forms, or ecosystems within a single study were considered independent observations (Curtis & Wang 1998). One hundred and nine publications that contained 313 observations across North and South America, Europe and Asia were selected for our analysis. Each observation was categorized by four groups: ecosystem type, N addition level, chemical form of N addition and experimental condition, and experimental length (detailed database structure is described in Appendix S1; table A1). When data were graphically presented, figures were digitized to extract the numerical values using Engauge Digitizer (Free Software Foundation, Inc., Boston, MA, USA). To avoid possible confounding caused by site conditions, we only included studies in which control and treatment sites experienced same climatic, soil and vegetation conditions. Therefore, some studies examining N enrichment effects along N deposition gradients, such as Brumme & Borken (1999), Pilegaard et al. (2006) and Magnani et al. (2007), were not included. Although N2O emission from leguminous crops may be considerable, we could not estimate N2O response ratio caused by biological N fixation because there is no non-N fixation control treatment. The application of organic N fertilizers, such as animal manures, also increases C supply. The effects of added C and N on GHG emissions could not be separated. Therefore, those studies that examined organic fertilizers were excluded from this analysis. Weighted meta-analysis was used to estimate effect size, thus publications which did not report experimental error were not included in our analysis. Because effect size could not be calculated for those observations with zero or a negative value under either control or elevated N treatment, data from studies such as Hall & Matson (1999), Ambus et al. (2006) and Curtis et al. (2006) were not included in the meta-analysis.
The data were analyzed using meta-analysis described by Hedges et al. (1999). The effect size of N enrichment for each individual observation was estimated by response ratio, and its variance is computed by:
Where is the control mean, is the treatment mean, SDc is the control standard deviation, SDT is the treatment standard deviation, nc is the control replication number and nT is the treatment replication number. We use the natural logarithm of the response ratio L = ln (r) to calculate the weighted mean of the log response ratio () and the standard error of the weighted mean. The detailed calculation process was provided by Hedges et al. (1999). The corresponding mean response ratio (μp) and 95% confidence interval (CI) for the mean are obtained by taking antilog. The N effect on a response variable was considered significant if the 95% CI did not overlap 1.
For each categorical group (ecosystem type, N addition level, chemical form of N addition and experimental condition) of each response variable (NEE, EC, SOC, CH4 emission, CH4 uptake and N2O emission), total heterogeneity among group (Qt) can be partitioned into within group heterogeneity (Qw) and between group heterogeneity (Qb), such that Qt = Qw + Qb, where Q-statistic has a chi-square distribution (Curtis & Wang 1998). A significance of Qb indicates that the response ratios are different for the levels of this categorical group. For each response variable, we first calculated the overall response ratio across the whole dataset and then examined Qb for each categorical group. For the categorical group with significant Qb, the data were subdivided and the mean response ratio for each level was calculated (Table 1). Mean of the levels were considered significantly different if their 95% CI did not overlap.
Table 1. Effect of N enrichment on between group heterogeneity (Qb) for each response variable
P-values in bold indicate significance (P ≤ 0.05).
Net ecosystem CO2 exchange
N addition level
Ecosystem C content
N addition level
Soil organic carbon
N addition level
N addition level
N addition level
N addition level
Emission/uptake factor (F)
Adapting from the IPCC guidelines for national GHG inventory methodology (Mosier et al. 1998), N addition-induced emission/uptake factor (F) was estimated for those response variables that were significantly influenced by N addition: F = (GN − GC)/N, where GN is annual flux of GHG from fertilized treatment (kg C or kg N ha−1 year−1); GC is annual flux of GHG from control (kg C or kg N ha−1 year−1); and N is annual N input (kg N ha−1 year−1). Only long-term field studies which measured growing season or annual GHG fluxes were included in this analysis. All statistical analyses were performed using sas (Version 9.1; SAS Institute Inc., Cary, NC, USA).
Global biogenic GHG budget estimation
For countries which do not have their country-specific activity data, the IPCC (2006) recommended Tier 1 methodologies for estimating total national anthropogenic emissions of N2O. The Tier 1 methodologies do not take into account the heterogeneity of site conditions, such as land cover, soil type or climate conditions. Following Tier 1 methodologies of IPCC (2006), we estimated ecosystem-specific emission/uptake factors for the three GHG, and use them to estimate the changes in GHG flux on global scale (Geq).
Geq were calculated in units of CO2 equivalents by: Geq = L × F × S × E × GWP (global warming potential). L is the level of N addition. It was assumed that non-agricultural ecosystems received N mainly from atmospheric deposition and agricultural ecosystems received N from fertilizer application as well as atmospheric deposition. Global average N deposition to terrestrial ecosystem is 3.5 kg N ha−1 year−1 (Phoenix et al. 2006) and mean N application rate for agricultural field is 108.1 kg N ha−1 year−1 estimated from the report of Food and Agriculture Organization of the United Nations (FAO 2008b). These values were used in the calculation. For non-agricultural ecosystem, L is the average N deposition level (3.5 kg N ha−1 year−1). For agricultural ecosystem, L is the summary of N deposition (3.5 kg N ha−1 year−1) and fertilizer application (108.1 kg N ha−1 year−1). The remaining variables are defined as follows: F is the GHG emission/uptake factor; S is the global surface area of the ecosystem; E is the weight conversion factor for CO2-C, N2O-N and CH4-C to CO2, N2O and CH4, which is 3.67, 1.57 and 1.33 respectively; GWP is 1, 296 and 23 for CO2, N2O and CH4 respectively.
Overall, N addition, ranging from 10 to 562 kg N ha−1 year−1, had no significant effect on NEE for non-forest natural ecosystems (Table 2). However, N addition ranging from 25 to 200 kg N ha−1 year−1 applied to forest ecosystems for 6–15 years increased EC by an average of 6%. EC of coniferous forests showed higher response to N addition than deciduous forests, although this difference was not statistically significant (Table 2). N addition ranging from 90 to 550 kg N ha−1 year−1 applied to agricultural fields for 4–50 years marginally increased SOC by an average of 2%. On average, forest ecosystems sequestered 24.5 ± 8.7 kg CO2-C ha−1 year−1 (n = 14) and agricultural soil sequestered 0.53 ± 0.10 kg CO2-C ha−1 year−1 (n = 15) per 1 kg N ha−1 year−1 added to the ecosystems (Table 3). The between-group heterogeneity for N addition level, N chemical form and experiment condition of non-forest NEE, forest EC and agricultural SOC was not significant (Table 1).
Table 2. Effects of N addition on net ecosystem CO2 exchange (NEE) of non-forest natural ecosystems, ecosystem carbon content (EC) of forest ecosystems and soil organic carbon (SOC) of agriculture ecosystem
n, no. observations; R, the mean response ratio; 95% CI, 95% confident intervals.
NEE of non-forest natural ecosystem
EC of forest ecosystem
SOC of agriculture ecosystem
Table 3. Summary of fertilizer-induced emission/uptake factor (F) for C uptake (kg CO2-C ha−1 year−1per 1 kg N ha−1 year−1) CH4 uptake (kg CH4-C ha−1 year−1 per 1 kg N ha−1 year−1) and N2O emission (kg N2O-N ha−1 year−1 per 1 kg N ha−1 year−1)
N addition, ranging from 30 to 400 kg N ha−1 year−1, significantly increased CH4 emission by an average of 95% when averaged across grassland, wetland and anaerobic agricultural system (Fig. 2). This response ratio did not differ among the three ecosystem types, N addition level, N chemical form and experiment condition (Table 1). Methane emission increased by 0.008 ± 0.004 kg CH4-C ha−1 year−1 per 1 kg N ha−1 year−1 added to the anaerobic agricultural system (n = 9; Table 3). Data were not sufficient to estimate an emission factor for non-agricultural ecosystems.
Overall, CH4 uptake was significantly reduced by an average of 38% under N addition, ranging from 10 to 560 kg N ha−1 year−1 (Fig. 3). The mean CH4 uptake decreased by 0.012 ± 0.006 kg CH4-C ha−1 year−1 for aerobic agricultural fields (n = 15) and 0.016 ± 0.004 kg CH4-C ha−1 year−1 for non-agricultural ecosystems (n = 23) per 1 kg N ha−1 year−1 added to the ecosystem (Table 3). In contrast to CH4 emission, there was significant between-group heterogeneity among the ecosystem types and N forms (Fig. 3; Table 1). The mean response ratios of CH4 uptake were < 1 for all ecosystem types, but this inhibition was not significant for grasslands and drained wetlands (Fig. 3). CH4 uptake was significantly reduced by NH4NO3, NH4+, NO3− and urea, but not by urea ammonium nitrate (Fig. 3). The inhibition from short-term studies was more pronounced than that of long-term studies although the difference was not significant (Fig. 3).
Nitrogen addition, ranging from 10 to 562 kg N ha−1 year−1, significantly increased N2O emission by an average of 216% across all ecosystems (Fig. 4). The mean N2O emission increased by 0.0072 ± 0.0012 kg N2O-N ha−1 year−1 (n = 34) for aerobic agricultural system, 0.0127 ± 0.0031 kg N2O-N ha−1 year−1 (n = 14) for anaerobic agricultural system, and 0.0087 ± 0.0025 kg N2O-N ha−1 year−1 for non-agricultural ecosystems (n = 42) per 1 kg N ha−1 year−1 added to the ecosystem (Table 3). The responses of aerobic and anaerobic agriculture systems were similar (Fig. 4). Compared with other ecosystem types, tropical forests emitted more N2O under N enrichment (an average of +739%; Fig. 4). Because only three observations were available for heathlands, the confidence intervals were wide and the inference regarding the mean N effect was limited. Among the five chemical forms of N fertilizer, NO3− showed the strongest stimulation (an average of +493%) of N2O emission. Both short-term and long-term studies showed significantly positive responses although the mean response ratio from short-term studies was significantly higher than that of long-term studies (Fig. 4). N2O emission (Table 1; Fig. 4) was significantly lower at low N addition rate (< 55 kg N ha−1 year−1) compared with high N addition rate (> 150 kg N ha−1 year−1). Compiling ambient N2O emission data from 23 studies, no clear dose–response effect was observed for N deposition and N2O emission (Fig. 5).
Global GHG budget
Our estimations indicated that N increased the GHG sink strength for forest ecosystems. However, agricultural ecosystems were sources for GHG emissions under intensive N application (Table 4). Overall, our calculations suggested that N addition (an average of 3.5 kg N ha−1 year−1 for non-agricultural ecosystems and 111.6 kg N ha−1 year−1 for agricultural fields) increased global terrestrial CO2 sink by 0.35–0.58 PgC per year. However, CO2 reduction could be largely offset by 53–76% because N addition could increase global CH4 emission by 0.29–0.88 Tg CH4-C per year, reduce CH4 uptake 2.92–6.86 Tg CH4-C per year and increase N2O emission by 1.43–1.90 Tg N2O-N per year (Table 5).
Table 4. Estimates of net changes in global biogenic greenhouse gases (GHG) flux caused by N enrichment
N-induced GHG emission/uptake factor (kg C or N ha−1 year−1 per 1 kg N ha−1 year−1)*
−24.5 ± 8.7
0.017 ± 0.005
0.006 ± 0.001
0.006 ± 0.001
0.008 ± 0.004‡
0.036 ± 0.013
−0.53 ± 0.1
0.012 ± 0.006¶
0.008 ± 0.004¶
0.009 ± 0.001
Area (108 ha)§
The changes in CO2 equivalents emission on global scale estimated by emission/uptake factor (Pg CO2 per year)†
*GHG emission/uptake factor (F) = 0 if the meta-analysis indicated that N had no significant effect on GHG flux for this ecosystem. Positive values of F refer to GHG released to the atmosphere and negative values refer to GHG uptake.
†The changes in CO2 equivalents emission on global scale (Geq) for was estimated by: Geq = F × L × S × E × GWP (global warming potential), details about those parameters were provided in the Methods section. 1Pg CO2 = 1012 kg CO2.
‡Data were not sufficient to estimate the emission factor for wetland ecosystem. We assumed that the emission factor of wetland was same as rice paddy because the meta-analysis suggested the response of CH4 emission were not different for the two ecosystems.
§The area of forest is the sum area of tropical, temperate and boreal forests; the area of grassland is the sum area of tropical and temperate grasslands. The source of forest and grassland data is Saugier et al. 2001; the source of wetland data is Finlayson et al. 1999.
¶N-induced CH4 uptake reduction was estimated for upland agricultural field only, the area is 12 × 108 ha. N-induced CH4 emission was estimated for rice field only, the area is 1.5 × 108 ha.
−1.31 ± 0.46
0.098 ± 0.028
0.041 ± 0.007
0.041 ± 0.007
0.014 ± 0.007‡
0.075 ± 0.027
−0.31 ± 0.06
0.055 ± 0.027¶
0.004 ± 0.002¶
0.631 ± 0.070
−1.61 ± 0.35
0.153 ± 0.056
0.018 ± 0.009
0.788 ± 0.111
Global net change induced by N enrichment
−0.655 ± 0.346 PgCO2 per year (−0.179 ± 0.094 PgC per year)
% CO2 uptake offset by N2O and GH4 emission
Table 5. Summary of published estimations of N effects on the global biogenic GHG budget compared with the values from the current study
Calculated from empirical data Data inputs: mean N addition rates; ecosystem specific N-induced CH4 emission factor
Human activities have dramatically increased the atmospheric concentrations of CO2 by 35%, CH4 by 48% and N2O by 18% since pre-industrial times (IPCC 2007). Although atmospheric concentrations of CH4 (1774 p.p.b.) and N2O (319 p.p.b.) are much lower than CO2 (379 p.p.m.), their global warming potential is 23 and 296 times that of CO2 respectively (IPCC 2007). Nitrogen enrichment could have both positive and negative effects on the biogenic production and consumption of these GHGs in ecosystems (Le Mer & Roger 2001; Mack et al. 2004; LeBauer & Treseder 2008). Overall, our meta-analysis indicated N addition likely causes a small increase in forest ecosystem C sequestration, which could be largely offset by increases in CH4 and N2O emissions from multiple types of ecosystems.
The C : N stoichiometry of plants affects N cycling in terrestrial ecosystems. The N cycle is closely coupled to the C cycle by key ecosystem processes such as photosynthesis, plant respiration, C allocation and microbial decomposition (Chapin et al. 2002). There is mixed evidence regarding the effect of N addition on ecosystem C sequestration. Net C uptake for a given ecosystem is determined by the difference between net primary production (NPP) and heterotrophic respiration. A recent meta-analysis of 126 N addition experiments found that N limitation to NPP is globally distributed and N fertilization increased aboveground NPP in all ecosystems except for desert (LeBauer & Treseder 2008). However, N fertilization has been observed to stimulate ecosystem C loss. For example, Bragazza et al. (2006) investigated peatlands across a gradient of N deposition levels and found higher atmospheric N deposition resulted in higher C loss by increasing heterotrophic respiration and dissolved organic carbon leaching. Similarly, Mack et al. (2004) found N fertilization stimulated SOC decomposition more than plant production in a tundra ecosystem, leading to a net loss of ecosystem C. Our meta-analysis indicated that N enrichment had no significant effect on C sequestration in non-forest natural ecosystems. We hypothesize that the nonsignificant response to N addition occurred because C gain via NPP was exceeded by C loss via heterotrophic respiration.
Among terrestrial ecosystems, the response of forests to N availability has been most intensively studied. Several studies suggested that N deposition has negligible or even negative effects on forest C sequestration. For example, Caspersen et al. (2000) found little evidence for growth enhancement due to N deposition after evaluating tree growth rates in five states (Minnesota, Michigan, Virginia, North Carolina and Florida). Nadelhoffer et al. (1999) traced the movement of the added 15N in nine temperate forests and found that 70% deposited N remained in the soil. They concluded that N deposition contributed little to C sequestration because plants were not the primary sinks for N deposition in those forests. Likens et al. (1998) found that chronic N addition increased mortality of sugar maple, probably due to base cation depletion from the soil, and increased soil acidity. Results of our meta-analysis revealed an average of 6% increase in EC after 6–15 years of N addition treatment. This value is derived based on 17 observations from nine studies conducted in United States. Future studies from other regions are needed to better characterize this relationship, given the diversity of forest ecosystems. Although limited data availability caused uncertainty, the general trend of this analysis is robust. Our finding is consistent with other research which has shown N deposition increased the C sink of forests (Magnani et al. 2007; Pregitzer et al. 2008). The increase in EC is likely driven by (1) an increase in N allocation to the plant as the microbial N demand becomes saturated (Magnani et al. 2007), (2) higher photosynthesis rate in response to higher leaf (N) and (3) decline in the activity of soil microbes, such as lignolytic fungi, due to higher litter [N] (Pregitzer et al. 2008). Our results showed that the C uptake factor, the same metric as C : N response ratio, is 24.5 : 1, which is lower but close to C : N response ratio of 40 : 1 reported by Hogberg (2007) and 50–75 : 1 reported by Sutton et al. (2008). Our results also agree with other reports that the NEE response rate to N deposition reported by Magnani et al. (2008), a C : N response of 175–225 : 1, may be overestimated (Sutton et al. 2008; de Vries et al. 2008).
Nitrogen may alter some ecological pathways that increase SOC in agricultural systems and others that may decrease it. Therefore, debate persists on the net effects of N on SOC. Increasing N inputs could increase the C sink in agricultural soils by increasing biomass yields and therefore the return of crop residues (Alvarez 2005), or by increasing N availability and therefore fostering humus formation (Christopher & Lal 2007). However, N addition could have little or even a negative impact on SOC by enhancing the microbial decomposition of crop residues and SOC (Khan et al. 2007). It is also important to note that a recent review by Christopher & Lal (2007) found that manure addition results in higher C sequestration compared with mineral N addition. Due to the confounding effects in SOC formation caused by additional C input of organic N, we only included studies using mineral N fertilizer. Overall, we found that N addition slightly increased SOC of agricultural ecosystems by an average of 2%. We interpret this finding with caution and note several caveats. First, a small change of 2% is likely difficult to detect in the field. Review of the individual studies concurs that although it was often reported that N caused slightly higher SOC, the increase was often not significant. Second, there are a limited number of studies available on N effects on SOC, which likely do not sufficiently represent the spatial and chemical heterogeneity of SOC pool. Considering the caveats of our analysis, we conclude that current data supports a neutral to weakly positive effect of N on SOC. The implications may be important when considering that agricultural usage of land is expanding rapidly to meet food demand, especially in developing countries (IPCC 2007). Our result indicated that the C uptake factor of agricultural land (0.53) was much lower than that of forest ecosystems (24.5), suggesting that converting forest to agricultural land could result in larger C debts than expected because forest is a stronger C sink with N deposition.
Generally, it is accepted that upland soils are sinks for atmospheric CH4, consuming c. 6% of atmospheric CH4, and wetland soils are sources, with natural wetlands accounting for 10–29% and rice paddy accounting for 4–19% of global CH4 emission (Le Mer & Roger 2001). Water table depth is an important factor governing CH4 flux. Water table depth is often dependent on rainfall, and in coastal areas tidal height will also be an important influence. Most publications describe wetlands as either drained or not drained without further comment on water table depth, therefore this analysis evaluated wetlands based on this criteria. Methane emission was observed to increase in the majority of undrained wetlands (Fig 2). However, drained wetlands can be a CH4 sink because water table drawdown results in aerobic soil conditions, therefore a decrease in CH4 production and an increase in oxidation (Crill et al. 1994; Ding et al. 2004; Fig. 3).
Numerous studies have demonstrated that both CH4 production and consumption are influenced by N fertilizers. However, the magnitude and even the direction of this response varied. Our results indicated that N addition caused a general stimulation on CH4 emission and suppression of CH4 uptake across multiple types of ecosystems thereby contributing to the increasing atmospheric CH4 concentration.
Multiple mechanisms could contribute to the effect of N addition on CH4 uptake. NH4+ reduces CH4 oxidation by acting as an inhibitor by competing for MMO. The oxidation of methanotrophic bacteria is optimally active at low osmotic stress. Nitrogenous salts (e.g. KNO3, NH4Cl and NH4NO3) have been shown to suppress the activity of methanotrophic bacteria by increasing osmotic pressure (King & Schnell 1998; Bodelier & Laanbroek 2004). Other mechanisms, such as toxicity of nitrite (NO2−) produced by nitrification or denitrification processes, may also involve in the inhibition of CH4 oxidation by N addition (Schnell & King 1994).
The increase in CH4 emission under N addition probably was caused by the activities of both methanogenic archaea and methanotropic bacteria. Methane production and consumption occur simultaneously in most ecosystems and the balance of the two processes determines the net flux of CH4 (Bodelier & Laanbroek 2004). Higher litter input under N enrichment alleviates C limitation to microbes. As a result, the activities of methanogenic archaea are enhanced and more CH4 is produced. Meanwhile, because less CH4 is oxidized by methanotrophic bacteria under N enrichment, more CH4 is emitted to the atmosphere.
Terrestrial ecosystems are the largest source of N2O, accounting for 60% of global emissions (IPCC 2001). N addition enhances the N2O emission under suitable conditions of temperature and organic C supply, although some studies also observed that N2O emission was decreased by N addition (Skiba et al. 1999; Ambus et al. 2006; Curtis et al. 2006). Overall, we found N2O emission was increased by an average of 215% under elevated N. The response of N2O emission was interactively influenced by ecosystem type, the chemical form and the amount of fertilizer N.
Anthropogenic N emission is increasing in tropical regions and is expected to double or triple its present-day rate in the year 2100 (IPCC 2001). N2O emission from tropical forests showed more pronounced response to N addition compared with other ecosystems. This greater response may be because tropical forests are often phosphorus (P) limited rather than N limited. Hall & Matson (1999) measured N2O emission after adding N fertilizer in two tropical rainforests in Hawaii. They found that N2O emission from P-limited sites was 54 times greater in the short-term N addition experiment and eight times greater in the chronic N addition experiment compared with that from N-limited sites. The P-limited soil had higher inorganic N concentration than the N-limited soil (Hall & Matson 1999), which increased N availability to the nitrifying and denitrificating bacteria. However, climatic conditions, especially temperature and precipitation, could also be key factors that drive N2O emission from tropical forest ecosystems.
We found NO3− caused a higher stimulation of N2O emission than NH4+, suggesting denitrification may be the dominate process contributing to N2O production. It is important to note that the publications in this meta-analysis which reported greater stimulation effects of NO3− were more often in wetlands or in areas receiving intensive precipitation. These environmental conditions may maximize denitrification. Even so, our finding was consistent with several previous field studies (Keller et al. 1988; Lindau et al. 1994; Wolf & Russow 2000; Russow et al. 2008). Keller et al. (1988) found that NO3− fertilization resulted in more than five times higher N2O emission than NH4+ treatment in a tropical forest. Lindau et al. (1994) also found NO3− addition stimulated more N2O emission than NH4+ addition in a forested wetland. They suggested this occurred because very little of the added NH4+ was denitrified due to severe limitation of nitrification by the reduced soil condition. By adding 15N labelled NO3− and NH4+ to soil, Russow et al. (2008) found that the contribution of denitrification to the total N2O production was 54% in soil with normal SOM and 76% in soil with high SOM. High soil NO3− concentration was also shown to inhibit N2O reduction to N2 and results in a high N2O/N2 ratio (Dalal et al. 2003). In contrast, Stehfest & Bouwman (2006) reviewed N2O emission from agricultural fields and did not find any significant difference between NO3− and NH4+ treatment. The discrepancy among published reviews emphasizes that the differential effects of NO3− and NH4+ should be interpreted with caution as they are linked to environmental conditions.
Although N2O emissions were higher at higher N application rates, we found there were no clear dose–response relationships between GHG emission/uptake and the amount of N addition to all non-agricultural ecosystems, a result consistent with observations in agricultural ecosystems [Bouwman 1996; Food and Agriculture Organization/International Fertilizer Industry Association (FAO/IFA) 2001]. However, Butterbach-Bahl et al. (1998) found that increasing NH4+ wet deposition led to a linear increase in N2O emission and decrease in CH4 oxidation at a spruce forest site. The dose–response relationship was observed at small scale characterized by homogenous conditions (such as a specific site), in contrast to the large heterogeneous scale investigated here. This inconsistency is likely caused because GHG production is influenced by multiple interactions of soil, climate and vegetation (IPCC 2001).
Regional and global scale studies which examined the effect of N deposition on climate change mainly focused on the budget of CO2, but not N2O and CH4 (Townsend et al. 1996; Fan et al. 1998; Magnani et al. 2007). However, the impacts of N on the GHG sink may be considerably overestimated because of neglecting N stimulation of N2O and CH4 emission. Our result indicated that CO2 reduction caused by N enrichment was offset by 53–76% because of increasing N2O and CH4 emission from multiple ecosystems. The purpose of our calculation is to illustrate an important concept: the C sink may be largely offset by stimulation of N2O and CH4.
Our approach is a type of static accounting that does not consider the dynamic interactions between N, GHG flux and the suite of environmental factors that influence these relationships. Therefore, it is a coarse brush, using average ecosystem response to N that is intended to illustrate a concept based on current available data. It has several notable uncertainties. First, we only assessed the direct effect of N addition on GHG emissions in terrestrial ecosystems. Increased N addition and deposition also increases N loss to groundwater or surface water through leaching or runoff, which causes indirect impacts on GHG fluxes in aquatic ecosystems. However, the magnitude of this indirect effect is difficult to estimate because of data limitations. Second, the spatial complexity of N deposition and the consequential heterogeneity of ecosystem response were not considered in the estimations. The emission/uptake factors were calculated by combining data from a wide range N addition levels and various environmental conditions that introduced uncertainty. The range of N addition by studies that were included in the emission factor calculation exceeded those which would be caused by deposition to non-agricultural systems. Micrometeorological factors including temperature, soil moisture and precipitation, could vary substantially between ecosystems across the large spatial area of the world. These environmental factors and other human disturbances such as land use and management practices, influence the ecosystem response to N addition and introduce variations into the estimations. Finally, there is limited empirical data for many regions and ecosystems.
Despite the many uncertainties listed above, Table 5 shows that multiple approaches to calculating N effects on C and N2O are beginning to show results that converge. This may indicated that the general direction and magnitude of N effects are robust. Without additional calculations on N effects on CH4 to compare our results, this calculation is more uncertain. In the coming years as more data becomes available that addresses research gaps, it is likely that improved dynamic models will be developed that better represent the diverse regions across the globe and how their response to N may vary. This will ultimately provide more insight as to the effectiveness of a static accounting approach.
Prospects for future research
Our review suggested N enrichment causes the strongest GHG sink in forest ecosystems, mostly due to stimulated C uptake after a few years of N addition. However, it is uncertain if short-term C accumulation can lead to long-term C sequestration. This uncertainly applies to all the types of systems investigated. For example, we found N fertilization slightly increased SOC of agricultural systems. However, other studies suggest that longer term N addition could reduce the capacity of ecosystems to sequester and store decay resistant soil C (Mack et al. 2004; Khan et al. 2007). Further research is needed to determine not only the long-term effect of N enrichment on ecosystem C sequestration, but also its long-term impact on ecosystem C allocation and the mean resident time of different C components. Another area that would benefit from future research is the effects of the frequency of N addition and plot size in experiments. Most available data were from plot-scale studies that applied N in a large pulse at one time; results at larger scales and from chronic low doses of N could be different.
Although many N addition studies on non-agricultural ecosystems have been conducted, most N addition levels were much higher than the levels which ecosystems may experience, and few studies were conducted at multiple levels of N addition. This data gap makes it difficult for evaluation of the N deposition-GHG flux dose–response relationship, which is necessary for quantitative ecological risk assessment and also can inform air pollution control policy. In addition to N addition level, different chemical forms of N could have different impacts on GHG emission. However, the evidence is mixed (Stehfest & Bouwman 2006; current study). Long-term experiments with different N chemical forms and multiple low levels N addition (< 30 kg N ha−1 year−1) are needed for evaluation of N-GHG dose–response relationship. This is particularly critical for N2O and CH4 since long-term and large-scale field measurements on N2O and CH4 are very sparse, compared with the extensive measurements of CO2 flux.
Greenhouse gases fluxes vary greatly over small spatial and temporal scales. An extensive global network is needed for better empirical data from which to estimate the regional and global GHG budget. Most research on N enrichment effects were conducted in North American and European sites. Data are sparse from some regions that receive high N addition through both agricultural N fertilizer input and atmospheric N deposition, such as East Asia and South America. Among all terrestrial ecosystems, the effects of N addition on agricultural lands and forests received more attention than other ecosystems. Ecosystems, such as wetland and tundra, have significant influences on global GHG budget, but our understanding on how N addition affects GHG balance in those ecosystems is insufficient. Measurements from those underrepresented regions and ecosystems are needed to fill the data gap.
The authors thank Dr Jana Compton, Dr Jeffrey Herrick and the anonymous reviewers for their insightful comments. This study was supported by the Research Participation Fellowship Program at the National Center for Environmental Assessment, U.S. Environmental Protection Agency (EPA) administered by the Oak Ridge Institute for Science and Education through an interagency agreement between the U.S. Department of Energy and EPA. This study was reviewed by the National Center for Environmental Assessment, U.S. EPA and approved for publication. Approval does not signify that the contents necessarily reflect the view and policies of the Agency nor mention of trade names or commercial products constitute endorsement or recommendation for use.