Effectiveness of soil management strategies for mitigation of N2O emissions in European arable land: A meta‐analysis

Soil management strategies involving the application of organic matter (OM) inputs (crop residues, green and livestock manure, slurry, digestate, compost and biochar) can increase soil carbon storage but simultaneously lead to an increase in non‐CO2 greenhouse gas (GHG) emissions such as N2O. Although multiple meta‐analyses have been conducted on the topic of OM input impacts on GHG, none has focused specifically on European arable soils. This study plugs this gap and can assist policymakers in steering European agriculture in a more sustainable direction. The objective of this meta‐analysis was to quantify how OM inputs of different nature and quality, but also the application strategy, can mitigate soil N2O emissions in different pedoclimatic conditions in Europe. We quantitatively synthesised the results of over 50 field experiments conducted in 15 European countries. Diverse arable crops, mainly cereals, were cultivated in monoculture or in crop rotations on mineral soils. Cumulative N2O emissions were monitored during periods of 30–1070 days in treatments, which received OM inputs, alone or in combination with mineral N fertiliser; and in controls fertilised with mineral N. The overall effect of OM inputs had a slight tendency to reduce N2O emissions by 10% (n = 53). With the increasing carbon‐to‐nitrogen ratio of the OM inputs, this mitigation effect became more pronounced. In particular, compost and biochar significantly reduced N2O emissions by 25% (n = 6) and 33% (n = 8) respectively. However, their effect strongly depended on pedoclimatic characteristics. Regarding the other types of OM inputs studied, a slight N2O emission reduction can be achieved by their application alone, without mineral N fertiliser (by 16%, n = 17). In contrast, their co‐application with mineral N fertiliser elevated emissions to some extent compared to the control (by 14%, n = 22). We conclude that amongst the seven OM inputs studied, the application of compost and biochar are the most promising soil management practices, clearly demonstrating N2O emission reduction compared to mineral N fertiliser. In contrast, other OM inputs had a small tendency to mitigate N2O emissions only when applied without mineral N fertiliser.

meta-analysis was to quantify how OM inputs of different nature and quality, but also the application strategy, can mitigate soil N 2 O emissions in different pedoclimatic conditions in Europe.We quantitatively synthesised the results of over 50 field experiments conducted in 15 European countries.Diverse arable crops, mainly cereals, were cultivated in monoculture or in crop rotations on mineral soils.Cumulative N 2 O emissions were monitored during periods of 30-1070 days in treatments, which received OM inputs, alone or in combination with mineral N fertiliser; and in controls fertilised with mineral N. The overall effect of OM inputs had a slight tendency to reduce N 2 O emissions by 10% (n = 53).With the increasing carbon-to-nitrogen ratio of the OM inputs, this mitigation effect became more pronounced.In particular, compost and biochar significantly reduced N 2 O emissions by 25% (n = 6) and 33% (n = 8) respectively.However, their effect strongly depended on pedoclimatic characteristics.Regarding the other types of OM inputs studied, a slight N 2 O emission reduction can be achieved by their application alone, without mineral N fertiliser (by 16%, n = 17).In contrast, their co-application with mineral N fertiliser elevated emissions to some extent compared to the control (by 14%, n = 22).We conclude that amongst the seven OM inputs studied, the application of compost and biochar are the most promising soil management practices, clearly demonstrating N 2 O emission reduction compared to mineral N fertiliser.In contrast, other OM inputs had a small tendency to mitigate N 2 O emissions only when applied without mineral N fertiliser.

K E Y W O R D S
climate change mitigation, effect size, EJPSOIL, nitrous oxide, organic matter inputs, pedoclimatic characteristics

| INTRODUCTION
Approximately 6.2 Tg N 2 O-N a À1 , or 35% of the annual worldwide emission, comes from agricultural soils (Kroeze et al., 1999).Following the application of chemical or organic fertiliser to the field, microbial processes (known as nitrification and denitrification) are the primary sources of nitrous oxide (N 2 O) in soils.Soil management strategies involving the application of organic matter (OM) inputs such as crop residues, green manure, livestock manure, slurry, digestate, compost and biochar have been shown to increase soil carbon (C) storage, as documented by several global meta-analyses and reviews (Bai et al., 2019;Bolinder et al., 2020;Gross et al., 2021;Jian et al., 2020;Siedt et al., 2021;Tiefenbacher et al., 2021;Xia et al., 2018).However, the mitigation effects of agricultural practices enhancing C sequestration can be offset if N 2 O emissions increase (Lugato et al., 2018;Zhou et al., 2017).The complex processes of N 2 O loss from soil can also be affected by many environmental and crop management factors such as soil organic C (SOC) content, nitrate and ammonium concentrations in soil solution, N application rate, fertiliser type and application technique, soil oxygen status, microbial abundance and activity, soil pH, soil drainage and moisture, and crop species (Butterbach-Bahl et al., 2013;Thangarajan et al., 2013).In contrast, soil management practices enhancing C sequestration and reducing N 2 O emissions may imply a "double-win" situation, with synergetic mitigation effects on climate change.

| Crop residues
The return of crop residues to the soil is an agricultural nutrient-conserving practice that effectively increases soil fertility and crop nutrition.When added to the soil, they are subject to microbial N mineralisation and nitrification, which results in N 2 O production.This process, however, relies on the N content of crop residues (Frimpong & Baggs, 2010;Garcia-Ruiz & Baggs, 2007;Millar & Baggs, 2005;Miller et al., 2008).Crop residues also stimulate microbial N assimilation through the development of an organic C substrate for microbial growth.Hence, heterotrophic microorganisms may compete with autotrophic nitrifiers for NH 4 + (Burger & Jackson, 2003), leading to a reduction of N 2 O production.However, crop residues may serve as an energy provider for denitrifiers, enhancing denitrification and consequently N 2 O production under anaerobic conditions.Moreover, it is after the incorporation of crop residues that peaks in N 2 O emissions are expected, also driven by the soil compaction status (Pulido-Moncada et al., 2022).
The impact of crop residues on N 2 O emissions was studied in five global meta-analyses.However, in four, most of the studies originated from China and India, and the number of European studies included was scarce (Table 1).Field studies demonstrated a large variation in response to crop residue inputs, depending on the soil water regime: either (i) emission stimulation for upland crops; or (ii) emission reduction and a negligible impact for paddy crops.Furthermore, crop residue effects on soil N 2 O emissions depended greatly on soil properties, specifically soil moisture content and soil texture (Chen et al., 2013), and clay content (Xia et al., 2018).Crop residue quality such as biochemical and physical characteristics is also an important factor controlling N 2 O emissions (Olesen et al., 2023).These quality factors control the balance between N mineralisation and immobilisation (due to microbial assimilation) during decomposition (Mary et al., 1996), as well as residue C dynamics and partitioning between mineralisation and stabilisation (Lashermes et al., 2016).Although Chen et al. (2013) highlighted the necessity of connecting the quantity and quality of crop residues with soil properties for predicting soil N 2 O emissions, in two later meta-analyses, the authors did not T A B L E 1 Global meta-analyses on the impact of organic matter inputs on N 2 O emissions, the main effect sizes, number of independent studies and main moderators studied.Weighted meta-analysis by the inverse of variance (Hedges et al., 1999).

Author
d OM inputs were compared to mineral fertiliser.consider these moderators (Fan et al., 2023;C. Liu et al., 2014).

| Cover crop/green manure
Green manure is a term used to describe crops that are grown in and incorporated into the soil to improve its fertility and OM content.Green manuring can have different impacts on N 2 O emissions, depending on the type of crop, the method of incorporation and the rate of mineral N fertilisation.Legume crops such as clover or alfalfa, tend to increase N 2 O emissions compared to non-legume crops such as grass or rye because they fix more N from the atmosphere and release it into the soil (Carter et al., 2014).Non-legume cover crops that mediate the trapping of nitrate in arable fields can reduce the indirect N 2 O emissions (Constantin et al., 2010).
Composting green manure with straw before soil application can reduce N 2 O emissions compared to ensiling or fresh incorporation because composting reduces N availability and increases the carbonto-nitrogen (C/N) ratio of the OM.Incorporating green manure by ploughing can increase N 2 O emissions compared to harrowing because ploughing promotes anaerobic conditions near the decomposing OM, which favours denitrification (Carter et al., 2014).Green manure can therefore have both positive and negative effects on N 2 O emissions, depending on how it is managed.
Four meta-analyses have studied the effect of cover crops/green manure on N 2 O emissions based on field data (Basche et al., 2014;Han et al., 2017;Li et al., 2023;Muhammad et al., 2019).There was a clear difference between non-legume and legume cover crops in terms of their effect on N 2 O emissions: nonlegume cover crops had either a negligible effect or reduced the emissions, whilst legumes typically stimulated N 2 O emissions (Table 1).Han et al. (2017) found that with increasing N inputs from cover crops, N 2 O emissions also increased.Both environmental and farm management factors modified the impact of cover crops on N 2 O emissions, including fertiliser N rate, precipitation and the period of measurement (Basche et al., 2014).Furthermore, the incorporation of cover crop residues contributed to an increase in N 2 O emissions, whilst a surface-placed residue resulted either in an emission reduction or had a negligible effect (Basche et al., 2014;Muhammad et al., 2019).However, when measured for periods of 1 year or longer, cover crops on average lead to a small or negligible increase in N 2 O emissions.
1.3 | Livestock manure, slurry, compost and digestate Globally, the application of animal manure to arable land as organic fertiliser enhances SOC stocks compared to synthetic N fertiliser alone (e.g.Maillard & Angers, 2014).However, the potential of manure application for climate change mitigation by increasing SOC stocks can be attenuated by enhanced N 2 O emissions.The magnitude and duration of N 2 O emissions from manure depend on its composition and quality, including its total N, ammonium-N, organic N, C/N ratio, pH and water content (Zhou et al., 2017).
The method and timing of manure application can also influence N 2 O emissions.For example, the surface application or incorporation of manure can reduce N 2 O emissions compared to injection or band spreading because it diminishes anaerobic conditions and denitrification potential in the soil (Thorman et al., 2020).Matching manure application with crop N demand can also reduce N 2 O emissions, for example, in the spring or summer (the beginning of the crop season) compared to the autumn (end of season) to reduce N losses (Thorman et al., 2020).
Two global meta-analyses have summarised the effects of manure and slurry on N 2 O emissions in comparison with mineral fertilisers in fields (Han et al., 2017;Zhou et al., 2017), and two meta-analyses have included mostly studies in China (Wei et al., 2020) or in China and India (Fan et al., 2023).Zhou et al. (2017) showed that the application of raw manure strongly increased soil N 2 O emissions, but with increasing soil clay content and pH, the effect sizes of manure application on N 2 O emissions decreased.Three other meta-analyses demonstrated overall small and non-significant effects of manure and slurry, ranging from À15.3% to 12% (Fan et al., 2023;Han et al., 2017;Wei et al., 2020) but also showed a negative correlation between clay content and the effect sizes (Han et al., 2017).In coarser soils, manure and slurry slightly increased N 2 O emissions compared to mineral fertiliser, whilst in finer soils, they tended to have similar or lower N 2 O emissions than mineral fertiliser (Han et al., 2017).The replacement rate of mineral fertiliser with organic fertiliser was another crucial factor, explaining the variability in effect sizes.The full substitution of mineral fertiliser with organic fertiliser decreased N 2 O emissions by 25% (Wei et al., 2020).
Two meta-analyses showed small and statistically non-significant effects of compost and digestate on N 2 O emissions (Kong et al., 2023;Zhou et al., 2017).The effect sizes of N 2 O emissions depended primarily on the replacement ratio, and the full substitution of mineral fertiliser with digestate slurry increased N 2 O emissions by 26% (Kong et al., 2023).

| Biochar
Biochar, the product obtained after pyrolyzing biomass, potentially contributes to climate change mitigation (Smith, 2016;Woolf et al., 2010).During pyrolysis, biomass is subjected to high heat and low O 2 conditions, resulting in about a quarter of its mass being converted to biochar.The process rearranges C into strongly bonded aromatic molecules which are more difficult for microbes to access and metabolise.Biochar C can remain unmineralised in the soil for hundreds to thousands of years, depending on the feedstock from which it is made, the temperature at which it is produced, and the soil and climate systems to which it is added (Lehmann et al., 2015;Lehmann et al., 2006).
In addition to its C sequestration potential, many studies have reported that biochar can help reduce soil N 2 O emissions.Four global meta-analyses synthesised the effect of biochar on N 2 O emissions in laboratory, greenhouse and field experiments with various crops (Borchard et al., 2019;Cayuela et al., 2014;Liu et al., 2018;Verhoeven et al., 2017).Cayuela et al. (2014) included studies conducted mainly in laboratories and greenhouses and demonstrated the largest overall reduction in N 2 O emissions by about 50% (Table 1).The biochar feedstock, pyrolysis conditions and C/N ratio were shown to be key factors influencing N 2 O emissions, whilst a direct correlation was found between the biochar application rate and N 2 O emission reductions.Indeed, Liu et al. (2018) demonstrated that along with the biochar addition rate, the magnitude of emissions reduction increased, reaching the maximum when the biochar addition rate was higher than 40 t ha À1 .The chemical form of N fertiliser applied with biochar and the interaction between soil texture and biochar were also found to have a major influence on soil N 2 O emissions.
Another global meta-analysis by Verhoeven et al. (2017) showed much lower reductions of N 2 O emissions due to biochar under field conditions (Table 1).This meta-analysis consisted of 43 studies, of which only 6 were conducted in Europe.Liu et al. (2018) showed the intermediate effects of biochar addition to soils resulting in a reduction of N 2 O emissions by 32%.The authors pointed out that the effect of biochar was largest in loam soils, but small and non-significant for soils with low organic carbon content (≤5 g kg À1 ).Biochar made from manure or pyrolyzed at temperatures lower than 350 C showed a weak and insignificant reduction of soil N 2 O emissions.Verhoeven et al. ( 2017) demonstrated there was a trend of reduced N 2 O mitigation in longer-term studies.Borchard et al. (2019) also stressed that although the overall effect of biochar was statistically significant (À38%), N 2 O emission reductions tended to be negligible after 1 year.The use of biochar reduced N 2 O emissions in arable farming and horticulture, but not in grassland or perennial crops.
To summarise, during the last decade, several metaanalyses have been published to quantify the impact of OM inputs on N 2 O emissions at the global scale (Table 1).These global meta-analyses showed inconsistent results even for the same OM type.In addition, none of these global meta-analyses focused on European agricultural soils, with most studies originating from China, India and United States, or in some cases, the countries of origin were not reported.These facts may prompt scepticism concerning the straightforward applicability of the results for European pedoclimatic zones.Moreover, many EU science policy practitioners may not be convinced when recommendations are based on the synthesis of non-European studies, which may inadequately reproduce agricultural management practices and pedoclimatic conditions in Europe.
Although there were some attempts to harmonise the results of European field experiments on the effect of crop residues, green manure and slurry on N 2 O emissions (Lehtinen et al., 2014;Sandén et al., 2018), the scarce data collected were a barrier to conducting a reliable research synthesis.Our study is the first European meta-analysis to summarise the effect of soil management strategies, involving the application of a wide range of OM inputs (crop residues, green manure, livestock manure, slurry, digestate, compost and biochar), on N 2 O emissions in arable land.
We hypothesised that soil management strategies involving the application of OM inputs would mitigate soil N 2 O emissions in European arable land.The objective of this meta-analysis was to quantify the effectiveness of OM inputs with contrasting nature and quality, as well as the effectiveness of the application strategy for mitigating soil N 2 O emissions in different pedoclimatic conditions in Europe.

| MATERIALS AND METHODS
To assure high quality, we followed a checklist for quality criteria specifically compiled for a meta-analysis in soil and agricultural sciences (Fohrafellner et al., 2023).

| Studies collection
The research question was structured according to the PICO framework (population, intervention, comparator and outcome): Population: European arable land.Intervention: Organic matter inputs (green manure, crop residues, livestock manure, slurry, digestate, biochar, compost), alone or in combination with mineral N fertiliser.
Outcome: Cumulative N 2 O emissions per unit land area for a period.
We found the articles by searching for the keywords "soil*" AND ("agr*" OR "farm*" OR "field") AND ("Europe" OR Name of European country) AND ("crop residue*" OR "cover crop*" OR "green manure" OR "livestock manure" OR "slurry" OR "compost" OR "biochar" OR "digestate") AND ("N 2 O" OR" nitrous oxide"), using Web of Science, Scopus, Agricola (USDA National Agricultural Library), ScienceDirect, the AGRIS International System for Agricultural System and Technology and Google Scholar.The outcome from the first search was a large number of studies that were later refined by selecting European research organisations with the assumption that only European research organisations had the ability to conduct field experiments in Europe.At least one (co-) author needed to be affiliated with these institutes.
The screening was conducted in two stages: 1.The title of each study was examined for relevance.If at this stage it did not indicate the presence of exclusion criteria (Table 2), the abstract was screened.2. All studies that passed the abstract screening were checked for suitability in the form of a full text screening.

| Inclusion criteria
To be included in the database, a study had to meet the inclusion criteria listed in Table 2.About 200 screened articles contained exclusion criteria, amongst which the most common were laboratory studies and zero N fertilisation.

| Data extraction
The data extraction method is crucial for dealing with the non-independence of the observations that can lead to underestimations of the standard error of the mean effect and therefore liberal evaluations of the statistical significance of effects (Nakagawa et al., 2017).
To avoid problems with the non-independence of the effect sizes, only one pair comparison corresponding to the longest period of N 2 O measurements was extracted from an article.If an article reported results for several OM inputs, a treatment was randomly selected, taking care that the number of studies for each OM type was comparable.If an article reported results from different experimental sites with different pedoclimatic characteristics or from the same site but with different soil characteristics, those sites were considered as independent studies and were included in the database.If several articles referred to the same experimental site with the same pedological characteristics, the article with the longest experimental duration was selected.However, when different articles reported the results from the same site such as El Encín (Spain), Berge (Germany), Cascina Baroncina (Italy), but experiments were conducted in different decades or/and with different OM inputs, we included them in the database as independent studies.
The data were extracted from tables and digitised from figures using the ImageJ 1.37 program (Schneider et al., 2012).Standard errors (SE) were converted to standard deviations (SD) where necessary (SD ¼ SE Ã ffiffiffi n p , where n is the number of replicates).When no measure of variability was provided, we extracted the SD from the ANOVA table using the EX-TRACT tool (Acutis et al., 2021(Acutis et al., , 2022)).This tool allows the estimation of the experimental error (i.e. standard deviation and standard error of treatments mean) associated with the statistical analysis results of published articles (i.e.estimated from the LSD, P(F) values or even from the assignment of letters indicating differences amongst means based on the results of a multiple comparison test).

| Database creation
We collected 46 articles published between 1993 and 2022 in peer-reviewed scientific journals, as well as a project report and a PhD thesis (Table 3; Appendix in Data S1).The database consists of 53 field studies, located in 46 sites with mostly loamy soil textures, across 15 European countries covering all European climate zones, from Alpine North to Mediterranean South (Table 3, Figure 1).A total of 13 studies was conducted in Spain, 11 in Germany, 4 in Italy, 3 in Denmark, England, Finland, France and Norway, 2 in Scotland, Switzerland and the Netherlands, and 1 in Greece, Cyprus, Slovakia and Sweden.The entire database for the meta-analysis is available in Zenodo (https://zenodo.org/doi/10.5281/zenodo.10907111).
Two studies on the effect of crop residues (Essich et al., 2020;Nett et al., 2016) were not included in the database due to extremely large effect sizes, ln R = 1.8 (500%) and 2.05 (680%) respectively, which violated the normal distribution of effect sizes.
All calculated cumulative N 2 O emissions were estimated from chamber measurements in the fields during periods from 30 to 1070 days, throughout the growing season (31 studies), outside growing season (3 studies), or throughout the year (19 studies).The types of OM inputs included 10 studies of crop residues, 10 of slurry, 9 of green manure, 8 of biochar, 7 of digestate, 6 of compost and 3 of livestock manure.Total N supply due to OM inputs ranged between 20 and 418 kg ha À1 a À1 , and the C/N ratio ranged between 2.7 and 390.In 32 field studies, organic materials were added to soils in combination with mineral N fertiliser, resulting in total N amounts more than in the control (26 studies), equal to the control (5 studies), or less (1 study).In the other studies, OM inputs were applied solely at the N amounts either more than in the control (nine studies), equal to the control (seven studies), or less (four studies) or an unknown amount (one study).The annual average precipitation and average annual temperature measured on the experimental sites that are included in this meta-analysis ranged between 250 and 1300 mm, and between 4.5 and 19.6 C respectively.The studies included in this meta-analysis cultivated diverse arable crops in the field experiments, mainly cereals (maize, spring wheat, winter wheat and spring barley).In 47 of these studies, arable crops were cultivated in monoculture or in crop rotations in a conventional farming system on the treatment plot, and in only six studies, an organic farming system was used on the treatment plots (control plots were always conventional).
The soil management of the treatments included conventional tillage at a soil depth of 20-30 cm in 42 studies, minimum tillage in 5 studies and no-tillage in 1 study, whilst 5 studies did not report soil tillage management.No cover crops were used in 41 studies, non-legume cover crops in 9 studies and legumes in 3 studies.Fields were not irrigated in 37 studies, whilst they were irrigated in 16 studies.

| Explanatory variables (moderators)
To explain the variation in the N 2 O changes due to OM inputs, we included 19 explanatory variables derived from the database and grouped them into five categories, namely, study characteristics, OM input characteristics, climate, soil and agronomic management (Table 4).Tillage practices were not included in the list of explanatory variables, as the fields were conventionally tilled in most studies, and there was a lack of studies with minimum tillage or no tillage.A Spearman rank order correlation (r s ) was run between soil characteristics and climate to assess their intercorrelation.

| Meta-analysis
Meta-analysis was conducted using Meta Win 2.0 statistical software (Rosenberg et al., 2000) and IBM SPSS Statistics 29.
F I G U R E 1 The location of 46 experimental sites used for meta-analysis and related environmental zones of Europe (Metzger et al., 2005) A quantitative meta-analysis involves calculating an effect size (i.e. the magnitude of the treatment effect) that can be averaged across independent studies.As two experimental groups were compared, the response ratio (R) was computed for the response variables as an index of the effect size: where X OM and X C represent the means for cumulative N 2 O emissions (kg N 2 O-N ha À1 ) in treatments (OM input solely or in combination with mineral N fertiliser) and in controls (mineral N fertiliser) respectively, averaged for experimental replicates.
As the distribution of R is skewed, performing statistical analyses in the metric of the natural logarithm of R is usually preferred due to its much more normal distribution in small samples than that of R (Hedges et al., 1999) A normal distribution for ln R was tested by Shapiro-Wilk test in SigmaPlot15.
We calculated the variance of ln(R) (Hedges et al., 1999): where SD OM and SD C are the corresponding standard deviations and n is the sample size (number of replicates).We assumed that studies did not share the same effect sizes, and we therefore used a random effects model to combine estimates across the studies.The application of this kind of model accounts for experimental method differences between studies (that are considered only a random sample of possible effect sizes) which may introduce variability ("heterogeneity", τ 2 ) amongst the true effects.
We calculated the weighted mean of the log response ratio for all studies as: where ln R i is the log response ratio for study i, n is the number of studies and w i is the weight for study i, defined as (Borenstein et al., 2009): where V i is the variance of the study i and τ 2 denotes the amount of residual heterogeneity (between-study variance).Because the variance of the effect sizes is a function of the sample size (Equation 3), studies with a larger sample size had lower variances and received heavier weights.
The τ 2 parameter is considered the variance of the true effect size.As it is impossible to compute it from the entire population of the effect size, τ 2 is an estimation of the observed effect by using DerSimonian and Laird method (Borenstein et al., 2009): where , where w i is the study weight, Y i is the study effect size, M is the summary effect and n is the number of studies.
A random effects model served to combine estimates across the studies, assuming that the studies in each subgroup did not share the same effect size.Because meta-analytic data often have small sample sizes and may violate basic distributional assumptions (such as normality), resampling techniques can be important to accurately determine the significance of meta-analytic metrics (Rosenberg et al., 2000).We used a bootstrap statistical method (Efron & Tibshirani, 1986) to generate bias-corrected 95% confidence intervals (CIs) around the log response ratios from 4999 iterations.To test whether ln R differed between the groups of categorical explanatory variables, we used the χ 2 test to examine the between-group heterogeneity (Q B ).To study the effect of continuous explanatory variables, we ran weighted meta-regressions, with ln R as the dependent variable, and the continuous variables as independent ones.We also used the χ 2 test to examine the model heterogeneity (Q M ), which describes the amount of heterogeneity explained by the regression models.The significant level of Q M indicates that an independent variable (a moderator) explains a significant amount of variability in effect sizes (ln R).
To identify the outliers, we used the backward search algorithm specifically developed for meta-analysis (Mavridis et al., 2017).Backward search algorithms start with the full data set and remove sequentially outlying observations until all outliers have been removed.This method can be useful when there are a few outlying studies (Mavridis et al., 2017).
Results were back transformed, except for metaregression, and reported in the text and figures as percentage changes from the controls: The OM input effects on the N 2 O emissions were considered to be significantly different from the controls if the 95% CIs did not overlap with zero.

| Sensitivity analysis
Funnel plot asymmetry, which may indicate publication bias in meta-analysis, was examined by plotting ln R against its SE (Sterne & Egger, 2001).Moreover, Egger's regressionbased test was conducted, enabling the detection of funnel plot asymmetry.A statistically non-significant p-value of Egger's test indicates no publication bias.
To estimate the magnitude of the file-drawer problem a fail-safe number (Nfs) was calculated.A fail-safe number is the number of non-significant, unpublished or missing studies that need to be added to a meta-analysis to change its results from significant to non-significant.Specifically, we used Rosenthal's method that estimates how many missing studies we would need to retrieve and incorporate in the analysis before the p-value became non-significant (Borenstein et al., 2009).
Trim-and-fill analysis was performed to allow one to enter values for "missing" studies to generate a symmetric funnel plot from which a new mean effect size can be estimated (Duval & Tweedie, 2000).
European climate zones from the Alpine North to the Mediterranean South.These studies were published mainly in peer-reviewed scientific journals between 1993 and 2022.The impacts of pedoclimatic characteristics, agricultural management practices and the nature and quality of OM inputs on N 2 O emissions were also studied.

| Overall effect
The effect sizes for 53 studies examining the effect of OM inputs on cumulative N 2 O emissions were normally distributed (W-statistic = 0.984, p = 0.697; Figure 2).The forest plot indicates the large variability of effect sizes, ranging from À75% (ln R = À1.35) to +200% (ln R = 1.10) (Figure 3).The overall effect across all studies had a slight tendency to emissions reduction by 10% (ln R = À0.12)compared to the control, that is mineral N fertiliser (median of 1.8 kg N 2 O-N ha À1 ).Since the 95% CI (À20%; 0%) overlapped with zero (the control), this indicates that the overall effect of OM inputs on N 2 O emissions was not statistically significant.

| C/N ratio
There was a statistically significant positive relationship between the C/N ratio of OM inputs and N 2 O emission reduction (Figure 4).According to the meta-regression, OM with C/N ratio < 20 had a risk for increased N 2 O emissions, whilst OM with C/N ratio of 20, 30, 50 and 100 reduced N 2 O by 1%, 4%, 11% and 25% respectively, for example.OM inputs with the C/N ratio of 300, such as biochar, may have a potential to reduce emissions by up to 65% (ln R = À1.01).
For the further moderator analyses, based on the similarity of effect sizes, the OM inputs were merged into two larger groups, namely OM 1 (green manure, crop residues, livestock manure, slurry and digestate) and OM 2 (biochar and compost).This allowed us to obtain enough studies per group.The response of N 2 O emissions was statistically different between the groups (Q B = 10.5, df = 1, 52; p = 0.001; Figure 5b): OM 1 had no effect on N 2 O emissions compared to mineral N fertiliser (0%, 95% CI: À14% to 15%, n = 39), whilst OM 2 reduced N 2 O emissions by 29% (95% CI: À40% to À18%, n = 14).

| Pedoclimatic factors
The meta-regressions indicated no statistically significant relationships between climatic characteristics and N 2 O emission change due to the inputs of green manure, crop residues, livestock manure, slurry and digestate (Figure 6a,b).In addition, soil characteristics such as the content of sand, clay or SOC, soil pH and soil C/N ratio was unrelated to N 2 O emission changes (Table S1 in Data S1).This suggests that the impact of these OM inputs on N 2 O emissions is similar for all European pedoclimatic zones.
In contrast, the annual average temperature and annual precipitation was correlated with the efficiency of biochar and compost to mitigate N 2 O emissions (Figure 6c,d).A smaller efficiency was observed under warmer or drier climatic conditions such as in the Mediterranean South than that in a temperate or boreal climate.For example, increasing the annual average temperature Effect size (ln R) F I G U R E 3 Forest plot showing effect sizes for 53 independent studies examining the effect of organic matter inputs on N 2 O emissions compared to mineral N fertiliser (control).Black squares are summary effect estimates for each study with lower and upper 95% CIs.The square size corresponds to study weight, the white square indicates the weighted average with 95% CIs across all studies and the dashed vertical line indicates the control (mineral N fertiliser).
from 5 C (boreal) to 15 C (Mediterranean South) reduced the efficiency of biochar and compost to mitigate N 2 O from 55% to 23%.With a further temperature increase to 20 C, the efficiency of biochar and compost to mitigate N 2 O emissions dropped to zero.Moreover, increasing soil pH and sand content was related to a decline of the mitigation effect of biochar and compost (Figure 7).According to the meta-regression, in soil with high sand content (70%, e.g.sandy loam), the efficiency of biochar and compost was low, and it dropped twice in alkaline soils (À7%) compared to neutral soils (À15%).In contrast, in soil with low sand content (20%, e.g.silt loam), the efficiency was as high as À43% in neutral soils, and it dropped slightly to À37% in alkaline soils.
Several confounding factors such as the intercorrelation between soil pH and both annual average temperature (r s = 0.781, p < 0.001, n = 14) and annual precipitation (r s = À0.673,p < 0.001, n = 14) should be interpreted with care.For example, studies in the Mediterranean South had alkaline soils, with pH ranging between 7.4 and 8.4, whilst studies in temperate and boreal zones had acidic soil, with pH ranging between 5.7 and 6.5.This does not allow us to draw a clear conclusion on which factor, soil pH or climate, is the most important in terms of driving N 2 O emissions reduction by biochar and compost.

| Management practices and study characteristics
There were no statistically significant relationships between total N supply (mineral N + OM) and effect sizes (Table S2 in Data S1).Agronomic management practices (farming systems, cropping systems, the presence of cover crops or irrigation) did not modify the effect sizes (Table S2 in Data S1).This indicates that N 2 O emission changes due to OM inputs were independent of the management practices studied here.Finally, the effect sizes were not related to the duration of experiments, the measurement period (covering a period from 1 month to 3 years) or to the season of N 2 O measurement (Table S2 in Data S1).Thers et al., 2020).For the backtransformation of ln R, see Equation ( 7).For abbreviations for environmental zones, see Figure 1.n, number of independent studies; Q M , model heterogeneity.

| Sensitivity analysis
The funnel plot for N 2 O emissions studies showed no asymmetry (Figure S1 in Data S1).In addition, no publication bias was detected by the Egger's regression-based test (p = 0.896; Table S3 in Data S1).The trim-and-fill analysis of publication bias for meta-analysis was also implemented.The imputed five "missing" studies shifted ln R slightly from À0.12 (À11%) to À0.17 (À16%) (Figure S1 in Data S1).However, the adjusted estimate is close to the original.For studies on biochar and compost, a fail-safe number is 77, indicating that the results are robust.A consistent number of unpublished or missing studies would need to be added (that is 5.5 times more than in the present meta-analysis) to change the results from significant to non-significant.

| DISCUSSION
We hypothesised that soil management strategies involving the application of OM inputs would mitigate soil N 2 O emissions.Indeed, the results of meta-analysis showed that the overall effect of OM inputs had a slight tendency to reduce N 2 O emissions by 10% (n = 53).One of the important factors related to the N 2 O emission reduction was the C/N ratio of OM inputs as indicated by metaregression (Figure 4).However, it should be noted that amongst all OM types, biochar with high C/N ratio was the major contributor to this strong relationship.The threshold of emissions reduction was set at the C/N ratio of 20-30, which is commonly considered a threshold for net N immobilisation (Mooshammer et al., 2014).Similarly, the meta-analysis by Cayuela et al. (2014) showed that biochar with a C/N ratio higher than 30 decreased N 2 O emissions but not for biochar characterised by ratios lower than 30 C/N.Meta-analyses on the effect of crop residues amendments also demonstrated that the C/N ratio was one of the main factors determining the variability of N 2 O emission response (Chen et al., 2013;Xia et al., 2018).
In our meta-analysis, we clearly demonstrated a statistically significant reduction of N 2 O emissions by 33% due to biochar and by 25% due to compost (Figure 5a).The result of biochar effect was in accordance with the global meta-analysis by Liu et al. (2018) and by Borchard et al. (2019), who estimated emissions cutting by 32% and 38% respectively.The meta-analysis by Cayuela et al. (2014) demonstrated an even larger reduction effect (À54%), which can probably be attributed to the experiment type, as the experiments were conducted in laboratory and greenhouse conditions but not in the field.The results of our study contrasted with the meta-analyses by Verhoeven et al. (2017) and Zhou et al. (2017), who showed minor effects of biochar (for upland À11.5%) and compost (+2.8%).The inconsistency between the outcomes of different meta-analyses may have been driven by the locations considered, and the variability in pedoclimatic conditions of the experiments may therefore have been included in both meta-analyses.
Our study demonstrates that the effect of biochar and compost depended on climate, with less emissions reduction in warmer or drier climatic conditions such as in the Mediterranean South than in temperate climates (Figure 6c,d).However, due to intercorrelation, it is difficult to confidently conclude which factor, soil pH or climate, was the most important moderator in terms of driving N 2 O emission reduction by biochar and compost.In line with our results, S anchez-García et al. (2014) showed that the dominant soil microbial community, usually characterised by pedoclimatic conditions, strongly influences the dominant N 2 O pathway (denitrification vs nitrification), and that biochar can both increase and decrease N 2 O emissions in different climates or soils.
Indeed, our meta-analysis confirmed that the mitigation effect of biochar and compost declined with an increasing soil pH or sand content (Figure 7) and became zero in soils with a sand content ranging from 80% to 95%, depending on soil pH.N 2 O emissions are more likely to be higher in acidic soils due to the suppression in the production of the N 2 O reductase enzyme (NosZ) at lower pH (B.Liu et al., 2014).Sandy soils will normally have a lower incidence of waterlogging compared to clay soils and thus lower anaerobic sites where denitrification can take place.The mitigation effect of biochar and compost may therefore be less pronounced in these conditions.
Several mechanisms may contribute to the N 2 O emission reduction after compost application, as observed in our study.The typically lower availability of mineral N compounds in compost (Omirou et al., 2020;Zhou et al., 2017) may have limited the denitrification process (Zhou et al., 2017).Furthermore, the addition of carbon associated with compost application may stimulate microbial respiration resulting in anaerobic microsites, where complete denitrification to N 2 can take place (Omirou et al., 2020), reducing the emissions of N 2 O. Dalal et al. (2010) suggested that the reduction of N 2 O emissions due to compost may be explained by its chemical properties.The typically high C/N ratios favour the immobilisation of mineral N, whilst high lignin concentrations and high lignin/N ratios slow down organic N decomposition.Both processes result in reduced mineral N levels in the soil, limiting denitrification.
Although high biochar C/N ratios are correlated with lower N 2 O emissions, the immobilisation of N in microbial biomass is not the main reason biochar reduces N 2 O. Biochar with high C/N ratio is typically produced at higher temperatures (>500 C), whereby N is driven off, and the remaining solid product is enriched in C, with higher alkalinity, increased porosity and surface area (Mukherjee et al., 2011).Several controlled experiments have concluded that biochar reduced N 2 O through the entrapment and sorption of N 2 O in biochar pore space (Cornelissen et al., 2013), slowing of the N 2 O diffusion to the surface in water saturated biochar pores (Harter et al., 2016), greater alkalinity (supporting complete denitrification to N 2 ) (Weldon et al., 2019), abiotic reduction of N 2 O to N 2 on redox active biochar surfaces (Quin et al., 2015) and immobilisation of N, which limits N substrate access to denitrifying bacteria (Singh et al., 2010;Spokas et al., 2009).In summary, whilst the C/N ratio for both compost and biochar was correlated with reduced N 2 O emissions, the mechanisms for reduction due to compost were mostly due to biotic factors, whilst the reasons for reduction due to biochar were mostly abiotic.
It should be noted that the extensive use of biochar in European agriculture is scant, and N 2 O reductions from biochar observed in research experiments (also in the studies included in this meta-analysis) have mostly been achieved when high dose rates have been used (>10 t ha À1 ).It is unrealistic to expect these high dose rates in broadacre agriculture due to the current high market price for biochar (€800 t À1 , Garcia et al., 2022).Furthermore, the N 2 O suppression effect from biochar has been shown to abate with time as it ages and loses its alkalinity (O'Toole, 2021; Thers et al., 2020).However, more long-term field experiments are needed to assess this effect in a European meta-analysis.
The effect of other OM types (green manure, crop residues, livestock manure, slurry and digestate) on N 2 O emissions ranged from À18% to +15% compared to mineral N fertiliser, but not statistically significantly (Figure 5a).Inputs of crop residues tended to decrease N 2 O emissions per unit area, but not significantly.This contradicts the general statements of previous metaanalyses, where crop residues in upland soils reported stimulation of N 2 O emissions, but with large variability in the effects (Table 1).However, this is mainly because the N 2 O stimulation was significant for vegetables and legumes, but not for cereals (Abalos et al., 2022;Chen et al., 2013).This pattern is also confirmed by the results of other incorporated plant residues such as green manures (Basche et al., 2014), where the stimulation by legumes was many times higher than by non-legume plants (+490% and +7% respectively, Table 1), arguably due to the N content of the residues, that is low C/N ratio.
In the European studies included in our database, there were mainly non-legume crops, whilst the scarce data on legumes were reflected in the knowledge gap on the effect of legume residue incorporation/removal on N 2 O emissions.Compared to other OM types, most cereal crop residues (except crops for silage) are already drier and more depleted in N content compared to green residues, as they are incorporated after the grain harvest.Another factor is the biochemical composition of the crop residues (Abalos et al., 2022), as the more lignin or cellulose content in the incorporated aboveground biomass at the time of application may tend to reduce cumulative N 2 O emissions after incorporation.The yield-scaled N 2 O emissions have been proposed as a suitable metric for the evaluation of N 2 O mitigation by considering the crop productivity (Van Groenigen et al., 2010).Van Groenigen et al. (2010) observed a negative correlation between the nutrient use efficiency and yield-scaled N 2 O emissions on the basis of a global meta-analysis but due to shortage of data, did not quantify the effect of organic amendments.Original field experiments showed inconsistent results for the yield-scaled N 2 O emissions due to crop residues: either increased N 2 O emissions due to non-legume cover crops (Taghizadeh-Toosi et al., 2022) and their mixture with legumes (Kim et al., 2017), or decreased N 2 O emissions by the incorporation of legume residues (Sanz-Cobena et al., 2014), whilst no changes by incorporation of cereal residues (Rahman et al., 2024;Sanz-Cobena et al., 2014).
Our results for livestock manure and slurry were consistent with the global meta-analyses by Han et al. (2017), Wei et al. (2020) and Fan et al. (2023), who demonstrated statistically non-significant effects on N 2 O emissions (À15% to +12%, Table 1).In their meta-analysis, Zhou et al. (2017) observed some differences in effect sizes between manure types (p = 0.087), with the largest stimulation of N 2 O emissions due to the application of poultry manure (+45.4%),whilst there was a statistically non-significant emission reduction (À21.4%)due to farmyard manure compared to mineral N fertiliser.The variation in the extent of emissions from different types of manure demonstrates the effects of manure properties such as moisture content, total N and available N content on emission generation (Bell et al., 2016).In our metaanalysis, however, we were unable to address the effect of manure types due to the limited number of studies (n = 3 for livestock manure and n = 10 for slurry).
This meta-analysis revealed that neither soil characteristics nor climate impacted N 2 O emission changes due to inputs of green manure, crop residues, livestock manure, slurry and digestate, suggesting that the results were valid for all European environmental zones.In contrast, several global meta-analyses demonstrated the importance of pedoclimatic characteristics in the regulation of N 2 O emissions due to manure (Han et al., 2017;Zhang et al., 2022;Zhou et al., 2017), cover crops (Muhammad et al., 2019) and straw return (Xia et al., 2018).For example, the significant stimulatory effects on N 2 O emissions due to manure were observed for warm temperate climates, acid soils (pH < 6.5) and the soil texture classes of sandy loam and clay loam, as shown in the global meta-analysis by Zhou et al. (2017).Zhang et al. (2022) concluded the greater power of key abiotic factors (annual precipitation, soil pH and soil C/N ratio) in explaining N-induced changes in soil N 2 O emission, since there was no clear relationship between changes in soil N 2 O emission and shifts in ammonia oxidizer and denitrifier abundances, which are the main producers of N 2 O.
Here, we stress the importance of the input strategy for organic fertilisers (green manure, livestock manure, slurry and digestate), as their addition to soils in combination with mineral N fertiliser increased N 2 O emissions by 30%, whilst their inputs alone showed a decline trend by 16%.This can be attributed by the lower availability or the slower release of mineral N compounds from OM compared to the application from mineral N fertiliser, as was also suggested by Zhou et al. (2017) for N 2 O emission reductions after compost application.In the metaanalysis by Wei et al. (2020), it was shown that the organic substitution of mineral fertiliser non-significantly decreased N 2 O emissions by about 13%, whilst at the high fertilisation rate, N 2 O emissions were significantly decreased by about 37%.Similarly, the meta-analysis by Kong et al. (2023), who mainly summarised Chinese studies, demonstrated that the partial replacement of mineral N fertiliser by biogas slurry (a liquid with a high moisture content and a low C/N ratio) reduced N 2 O emissions by 16%.

| CONCLUSION
The major challenge of conducting this meta-analysis was the ambitious goal of synthesising the effects of seven OM inputs, which had a different nature and quality (i.e.plant-or animal-based, raw or pre-treated), on N 2 O emissions in the European field experiments.
Amongst the seven types of OM inputs studied, only biochar and compost application to soils reduced N 2 O emissions statistically significantly compared to mineral N fertiliser.Other OM inputs (green manure, livestock manure, slurry and digestate) added solely to soils tended to reduce N 2 O emissions.In contrast, the addition of OM inputs to soils in combination with mineral N fertiliser entailed a risk of increasing N 2 O emissions.
Although the total number of independent experiments included in the database exceeded 50, which is considered as a large number of studies for a metaanalysis (Hedges et al., 1999), the availability of field experiments for each OM type was relatively small.This impeded a detailed study on the role of pedoclimatic factors and management practices in relation to a specific OM type.There is therefore a need for more European studies that monitor N 2 O emissions in the field.
Figure 2 Upland The distribution of effect sizes for 53 studies examining the effect of organic matter inputs on cumulative N 2 O emissions.The dashed line indicates the control (mineral N fertiliser).The result for the normality test (Shapiro-Wilk) is shown.
0.0036x + 0.0665 Q M = 11.2, p = 0.0008, n = 36 F I G U R E 4 The relationship between the C/N ratio of organic matter inputs and N 2 O emission change compared to mineral N fertiliser (control).The symbol size represents the study weight.The dashed line indicates the control (mineral N fertiliser) and the crosses indicate the outliers (ID38, ID39, ID40-1).For the backtransformation of ln R, see Equation (7).n, number of independent studies; Q M , model heterogeneity.N 2 O emission change compared to mineral N fertiliser ( change compared to mineral N fertiliser (ln R) ( o C) Annual precipitation (mm)y = 0.0520x -1.0424 Q M = 10.1, p=0.002, n=13 y = -0.0006x+ 0.0967 Q M =7.7, p=0.005, n=14 Q M = 2.0, p=0.154, n=39 Q M = 0.57, p=0.451, n=39 F I G U R E 6 (aand b) Scatter plots between N 2 O emission change due to green manure, crop residues, livestock manure, slurry and digestate and annual average temperature and annual precipitation.(c and d) Weighted meta-regressions (solid lines) between N 2 O emission change due to biochar and compost and annual average temperature and annual precipitation.The symbol size represents the study weight.The dashed line indicates the control (mineral N fertiliser), whilst the cross indicates the outlier (ID45 1.374 + 0.093*pH + 0.008*sand Q M = 22.6, p < 0.001, n = 14 F I G U R E 7 Weighted meta-regressions between N 2 O emission change due to biochar and compost and the combination of soil pH and sand content.The symbol size represents the study weight.The dashed line indicates the control (mineral N fertiliser).For the back-transformation of ln R, see Equation (7).n, number of independent studies; Q M , model heterogeneity.
Inclusion and exclusion criteria for the literature screening process.
T A B L E 2 T A B L E 3 Studies included in the meta-analysis.
Categorical and continuous explanatory variables (moderators) included in the meta-analysis.
T A B L E 4Abbreviation: OM, organic matter.
emission change due to organic matter (OM) types (a) individually and (b) clustered to the larger groups (OM 1 ; OM 2 ), and separated by input strategy (OM alone, OM applied alone; OM + N, OM applied in combination with mineral N fertiliser).The filled circles indicate OM 1 (green manure, crop residues, livestock manure, slurry and digestate), and the open circles indicate OM 2 (biochar, compost).The square corresponds to the overall effect.The dashed vertical line indicates the control (mineral N fertiliser).The numbers in parentheses indicate the number of independent studies.The N 2 O emission changes were considered significantly different from the controls if the 95% CIs did not overlap with zero.Q B is the between-group heterogeneity test.