Refining the Factors Affecting N₂O Emissions from Upland Soils with and without Nitrogen Fertilizer Application at a Global Scale

Nitrous oxide (N₂O) is the third most important anthropogenic greenhouse gas, with an atmospheric lifetime of one hundred and twenty-three years and a global warming potential 300 times that of carbon dioxide (CO₂) over a 100-year timescale (Wang et al., 2021). N₂O is also a major ozone-depleting substance in the stratosphere (Ravishankara et al., 2009). Global atmospheric N₂O concentration has increased by more than 23% since the pre-industrial era, from 270.1 parts per billion (ppb) in 1750 to 335.9 ppb in 2022 (Lan et al., 2022). As expected, further increases in atmospheric N₂O concentration are projected for the future (Cavigelli et al., 2012; Thompson et al., 2019). This is why the so-called laughing gas has attracted considerable interest from scientists. Currently, approximately two-thirds of N₂O emissions originate from nitrogen (N) fertilizer application soils and natural soils (Tian et al., 2020). According to Wang et al. (2020), annual global N₂O emission from cropland was estimated at 0.82 ± 0.34 Tg N yr^–1 during 1961−2014, mostly from the upland farming systems, including pasture, upland, and lowland-upland rotation. In addition, studies have shown that upland soil, which supports more than 38% of the global population and is one of the most important cropland systems in the world, significantly contributes to N₂O emissions (Li et al., 2023). Therefore, it is essential to understand the mechanism and influencing factors of N₂O emissions from upland soils to provide a theoretical basis for the development of emission reduction measures.

The microbial-mediated processes contributing to N₂O emissions in soils are identified as nitrification, denitrification, chemo-denitrification, and nitrifier denitrification (Wrage et al., 2001; Shcherbak et al., 2014; Fowler et al., 2015). Nitrification is a process that oxidizes ammonium (NH₄⁺) to nitrate (NO₃⁻) by soil nitrifying microorganisms, with the release of N₂O as a by-product (Firestone and Davidson, 1989). Denitrification is a NO₃⁻ reduction process, in which NO₃⁻ is sequentially converted to nitrite (NO₂⁻), nitric oxide (NO), N₂O, and eventually molecular nitrogen (N₂) by soil denitrifiers under anaerobic conditions, where NO₂⁻, NO, and N₂O are obligate intermediates (Firestone and Davidson, 1989). Chemo-denitrification is a process that reduces NO₂⁻ to NO, then to N₂O, and finally to N₂ under abiotic conditions (Pilegaard, 2013). Nitrifier-denitrification is the pathway of nitrification in which ammonia (NH₃) is oxidized to NO₂⁻ followed by the reduction of NO₂⁻ to NO, N₂O, and N₂ (Wrage et al., 2001). Denitrification and nitrification are the main pathways of N₂O emission from agricultural soils (Smith, 2017; Byrne et al., 2020).

These processes of N₂O production and subsequent N₂O emission rate are influenced by many factors, which include soil physicochemical properties, climatic factors, and agricultural management practices (Butterbach-Bahl et al., 2013; Charles et al., 2017; Xia et al., 2018; Wu et al., 2021; Shen et al., 2022). For soil physicochemical properties such as soil texture, finer textured soils can form more capillary pores which in turn tend to create anaerobic conditions that promote denitrification reactions (Cui et al., 2023). For climatic factors, such as air temperature, higher temperatures enhance denitrification reactions and thus promote N₂O emissions (Huang et al., 2022). For agricultural management practices such as the application amount of N fertilizer, additional N fertilizer increase soil NO₃⁻ and NH₄⁺ concentrations, which increases N₂O emissions (Yu et al., 2023). Currently, many studies have also explored the main influencing factors of N₂O emission. For example, Abdalla et al. (2011) concluded that the main influencing factors on N₂O emissions from agricultural soils were soil N content, soil moisture, soil temperature, precipitation, N fertilizer application, and other management practices. Bouwman et al. (2002) identified N application rate per fertilizer type, soil organic C content, and soil drainage as the primary influencing factors, according to N₂O observations published before 2002 on a global scale. Lu et al. (2022) showed that human land use change and agricultural management were the main factors in agricultural soils and N deposition and climate warming were dominant drivers for N₂O emission in natural soils across the United States. Although many studies have been conducted to describe how individual impact factors work, the key influencing factors and hierarchical relationships between the impact factors of N₂O emissions from the upland farming systems globally remain highly uncertain. The strong variability of N₂O emissions over space and time, the limited available measurement data, and the synergistic or antagonistic effects of combined drivers are the main limiting factors for exploring key influencing factors and hierarchical relationships globally (Tian et al., 2018).

In this study, we used the direct field measurement datasets of N₂O emissions around the whole world, which were established by the Consultative Group on International Agricultural Research (CGIAR), to determine the key factors and hierarchical relationships of factors influencing N₂O emissions from upland soils. The specific issues addressed in this study were: i) recognizing the key controlling factors and hierarchical relationships of influencing factors of N₂O emission from upland soils in the N fertilized and non-N fertilized groups on a global scale; ii) identifying the key influencing factors on N₂O emissions from upland soils in the major climate zones; iii) assessing the reduction potential of N₂O emissions by the improved agricultural practices.

The remainder of this paper is organized as follows. Section 2 presents the data sources and analysis methodology used in this study. This includes a description of the data composition and an introduction to the correlation, ANOVA, structural equation modeling and variance partitioning analysis methods used to analyse the data. Section 3 presents the results, including the correlation of the influence factors with N₂O emissions, the presentation of the hierarchy of influence factors using structural equation modelling in the N and non-N fertilized groups, and the identification of the key influence factors for N₂O emissions in different climatic zones. Section 4 analyses the mechanism of influence factors on N₂O emissions, explores the possible reasons for the key controlling factors and hierarchy of influencing factors of N₂O emissions in this dataset, as well as a discussion of agricultural measures to reduce N₂O emissions. Finally, section 5 provides a summary and conclusion.

We compiled the datasets of the direct measurements of N₂O emission from uplands supported by CGIAR programs (CRPs) on Climate Change, Agriculture, and Food Security (CCAFS) (https://samples.ccafs.cgiar.org/n2o-dashboard/). The datasets of the direct field measurements of N₂O emissions were established by collecting the scientific studies published worldwide between 1980 and 2016. The datasets comprised 341 scientific studies, with observations at 251 study sites, which is representative because of the large volume of data. The data for this paper were extracted by the following two criteria: i) including N₂O emission observations from at least one unfertilized and one fertilized group, and ii) reporting the application amount of N fertilizer and the accumulative N₂O emission over the trial period. The original datasets comprised 2918 direct observations of N₂O emissions worldwide. In this study, 2736 direct observations with 685 and 2051 observations of the non-N fertilized group and the N fertilized group, respectively, were selected, excluding those observations with a typical upland field, e.g., a paddy field [Fig. S1 in the electronic supplementary material (ESM)].

Each observation recorded the cumulative N₂O emissions over the trial period and the corresponding auxiliary information. The auxiliary information, consisting of those factors influencing the N₂O emission, was divided into four categories: climate (CLM), soil physical properties (SLP), soil chemical properties (SLC), and agricultural practices [(including N fertilizer rate (NRA) and type (FER) and the other agricultural practices (PRA)] to conduct the analysis. The CLM category included the mean annual precipitation (MAP), as recorded in the datasets, and the mean annual air temperature (MAT) which was newly obtained from the corresponding scientific study by the authors. The SLP category was comprised of sand content (SAN), silt content (SIL), clay content (CLA), bulk density (BD), and texture (TEX). Among these, SAN, SIL, BD, and CLA originated from the dataset record. TEX was classified into 13 texture categories, including 12 texture categories according to the soil texture triangle divisions of the U.S. Department of Agriculture [based on the proportion of sand (0.05 to 2 mm), silt (0.002 to 0.05 mm), and clay (< 0.002 mm) content], and organic soils. The SLC category included pH, soil organic carbon (SOC), and total soil nitrogen (TSN). The FER category included controlled release fertilizer (CRF), nitrification inhibitor (NI), urease inhibitor (UI), and organic matter (OM). The PRA category included crop type (CRT), the fertilization application method (FAM), biochar application (BIO) (including vinasse burn and biochar application), tillage (TIL), and irrigation (IRR). The CRT was divided into 9 species of crops planted in upland fields, namely cereals, fallowlands, fibers, grasslands, legumes, oilseeds, perennials, sugars, and vegetables. The FAM included deep application, incorporation, surface banding, and surface broadcast. Information about the SLC, FER, and PRA categories were all recorded in the datasets. The N₂O emission intensity (NEI, g N ha^–1 d^–1) for each observation was calculated using the recorded cumulative N₂O emissions and the experiment duration which was newly obtained from the corresponding scientific study.

To provide a quantitative basis for the estimation of N₂O emission from upland soil, we calculated the mean, standard deviation (SD), minimum, median, maximum, skewness, and kurtosis of N₂O emission intensity and its influencing factors. Skewness is a statistic that measures the asymmetry of the probability distribution of a random variable, and kurtosis is a statistic that examines the steepness or smoothness of the data distribution. Skewness and kurtosis are used to test whether a sample conforms to a normal distribution. When the absolute value of skewness is less than 3 and the absolute value of kurtosis is less than 10, the sample is largely acceptable as normally distributed.

If the sample does not satisfy the normal distribution, the box-cox transformation, a generalized power transformation, was performed to improve the normality and variance chi-square of the data. The box-cox transformation was performed according as follows:

where y (λ) is the new variable obtained by the box-cox transformation, y is the original continuous dependent variable, and λ is the transformation parameter. The above transformation requires the values of the original variable to be positive. If the value was negative, a constant should be added to all the original data to make them positive, thereby allowing the box-cox transformation to be performed.

The box-cox transformation parameters for the variables of the non-N fertilized and N fertilized groups were obtained using the “MASS” package in the R language, as shown in Table S1 in the ESM.

Correlation analyses were conducted to understand the degree to which the influencing factors and N₂O emissions were correlated. The ANOVA analysis was also performed on the influencing factors in the N fertilized and non-N fertilized groups to understand whether there were differences in the influencing factors between the fertilized and non-fertilized groups as well as the extent to which the impact factors affected N₂O emissions. The correlation analysis was performed using the “corrplot” packages in R language, and the ANOVA analysis was performed by the function “aov”.

Because structural equation modeling (SEM) can show the structure and relationships among the influencing factors on N₂O emissions, we established the SEMs to explore the hierarchical relationship between factors influencing N₂O emission intensity based on the correlation and the confirmatory factor analysis (Text S1, Fig. S2, Tables S2, S3 in the ESM). SEM was performed using the “corrplot”, “lavaan” and “semplot” packages in R language. All environmental factors were initially included in the SEMs; however, the models did not qualify. The SEMs were tested by reducing the number of variables one by one. Six model fit indicators, including the cardinality of freedom ratio (X²/df), the comparative fit index (CFI), the Bentler-Bonett normed fit index (NFI), the goodness of fit index (GFI), root-mean-square error of approximation (RMSEA) and the standardized root-mean-square residual (SRMR) were used to determine whether the constructed structural equation model was feasible and satisfactory. If X²/df is less than 5, CFI, NFI, and GFI values are greater than 0.9, RMSEA is less than 0.1 and SRMR is less than 0.05, it indicates that the SEM fits well.

To further spatially analyze the influencing factor of soil N₂O emission intensity, three climatic zones were selected with data volumes greater than 40 in the non-N fertilized group and greater than 100 in the N fertilized group, respectively, for variance partitioning analysis. These clearly documented zones included the subtropical monsoon zone (61 and 202 observations for the non-N fertilized and N fertilized groups, respectively), the temperate continental zone (57 and 203 observations for the non-N fertilized and N fertilized groups, respectively), and the temperate monsoon zone (47 and 148 observations for the non-N fertilized and N fertilized groups, respectively). The influencing factors in explaining N₂O emission were categorized into four groups, namely CLM, SLP, SLC, and PRA, and then analyzed using redundancy analysis (RDA)-based variance partitioning (Lazcano et al., 2021). The RDA analysis utilized backward selection to eliminate non-significant variables from each set of explanatory factors, resulting in the identification of significant influence factors. These significant influence factors were subsequently employed in the variance partitioning analysis to determine the extent to which they accounted for N₂O emissions. The entire method was implemented using the “vegan” package in R language.

The descriptive statistical analysis result of the influencing factors and N₂O emissions intensity for the non-N fertilized and N fertilized groups is shown in Table 1. There was no significant difference in the influencing factors of N₂O emission, while there were significant differences in N₂O emission intensities between non-N fertilized and N fertilized groups. Specifically, the MAP, MAT, SOC, TSN, pH, BD, SAN, SIL, and CLA indicated no significant difference between non-N fertilized and N fertilized groups. The mean daily N₂O emission from non-N fertilizer application soils was 5.34 g N ha^–1 d^–1, significantly lower than that from N fertilizer application soils (17.83 g N ha^–1 d^–1) (p< 0.01). In the non-N and N fertilized groups, large variations were observed in N₂O emission intensities. The maximum and minimum values of N₂O emission intensities from the non-N fertilizer application soils were 135 g N ha^–1 d^–1 and −6.33 g N ha^–1 d^–1, respectively. Likewise, the corresponding maximum and minimum values from N fertilizer application soils were 630.65 g N ha^–1 d^–1, and −7.67 g N ha^–1 d^–1, respectively. The large differences in N₂O emission intensity could be attributed to two reasons. One was the difference in the study regions, and the other was the difference in the measurement periods such as the growing season and slack season, in addition to the length of the observation period. The large skewness and kurtosis values of the MAP, SOC, TSN, BD, NRA, and NEI indicated that the data did not satisfy a normal distribution, and box-cox transformations were necessary. After performing the box-cox transformation, we found that the influencing factors and NEI showed a normal distribution. The results of the descriptive statistical analysis of the box-cox transformed data are shown in Table S4.

The influencing factors (IF) of MAP, MAT, SOC, TSN, BD, SAN, SIL, CLA, TEX, NRA, and NEI are referred to the mean annual precipitation (mm), mean annual temperature (°C), soil organic carbon (%), total soil nitrogen (%), bulk density (g cm⁻³), sand content (%), silt content (%), clay content (%), texture, nitrogen rate (kg N ha⁻¹) and nitrogen emission intensity (g N ha⁻¹ d⁻¹), respectively. Skewness is a statistic that measures the asymmetry of the probability distribution of a random variable and kurtosis is a statistic that examines the steepness or smoothness of the data distribution. SD, Min., and Max. are referred to as the standard deviation, minimum value, and maximum value, respectively. Between the same influencing factors in the N fertilizer and non-N fertilizer groups, displaying different lower case letters (e.g. a vs. b) indicates a significant difference (p< 0.05), while displaying the same lower case letters indicates a non-significant difference.

Significant correlations were observed between NEI and various factors in non-N and N fertilized groups (p< 0.01) (Fig. 1). Specifically, NEI exhibited a positive correlation with MAP, MAT, TEX, CLA, SOC, and TSN, and negatively correlated with SAN, BD, and pH. Moreover, irrigation significantly affected N₂O emission intensity in the N fertilized group but had no impact on the non-N fertilized group. Additional positive correlations were found between NEI and NRA, organic matter, and biochar application in the N fertilized group. Conversely, the use of nitrification inhibitors and urease inhibitors displayed negative correlations. Notably, no significant correlation was observed between NEI and the controlled-release fertilizer.

Figure 1.  Pearson’s correlations analysis of N₂O emissions and bcMAP (box-cox transformed mean annual precipitation), bcMAT (box-cox transformed mean annual temperature), bcSAN (box-cox transformed sand content), bcSIL (box-cox transformed silt content), bcCLA (box-cox transformed clay content), bcBD (box-cox transformed bulk density), bcSOC (box-cox transformed soil organic carbon), bcTSN (box-cox transformed total soil nitrogen), pH, bcTEX (box-cox transformed texture), bcNRA (box-cox transformed nitrogen rate), CRF (controlled release fertilizer), NI (nitrification inhibitor), UI (urease inhibitor), OM (organic matter), BIO (biochar), TIL (tillage), IRR (irrigation) and bcNEI (box-cox transformed nitrogen emission intensity) for the (A) non-N fertilized and (B) N fertilized datasets. The symbols *, **, and *** followed by the figures indicate significant effects at p< 0.05, p< 0.01, and p< 0.001, respectively.

Further, the specific effects of the categorical influencing factors on N₂O emissions obtained using ANOVA analysis are as follows. In the non-N fertilizer application group, there were no significant differences in N₂O emission among the different crops (Fig. 2A), but in the N fertilizer application group, the mean daily N₂O emissions intensity from the soils with different crops had significant differences, which were ranked from the highest to the lowest in the following order: vegetables > sugars > fibers > fallowlands > perennials > grasslands > oilseeds > cereals > legumes (Fig. 2B). The NEI from soils with legumes (6.09 g N ha^–1 d^–1) were similar to cereals (14.4 g N ha^–1 d^–1) and oilseeds (8.87 g N ha^–1 d^–1) and significantly lower than those from soils planted with other crops.

Figure 2.  N₂O emission intensity from upland fields with different growing crop types in the (A) non-N fertilized and (B) N fertilized datasets. The scattered dots represent the raw data. The red dots represent the means of the groups. Different lower-case letters above bars such as a, b and c indicate significant differences in the N₂O emissions among the crop groups in the Tukey test (p< 0.05). Error bars are the standard deviations.

The mean daily N₂O emissions from different soil textures (including 12 texture categories and organic soil) both showed significant differences in the non-N and N fertilized datasets (Fig. 3), but the differences were more significant in the N fertilized group than in the non-N fertilized group. The N₂O emission intensity of the organic soil was the highest and the sand clay soil had the lowest N₂O emission intensity in both non-N and N fertilized groups. There was no significant difference in N₂O emission intensity with or without tillage (Figs. S3A and S3B). In the non-N fertilized group, there was no significant difference in the N₂O emissions with or without irrigation (Fig. S3C). However, irrigation can significantly increase the intensity of N₂O emissions in the N fertilized group (Fig. S3D).

Figure 3.  N₂O emission intensity from upland fields with different soil textures in the (A) non-N fertilized datasets and (B) N fertilized datasets. San, Los, Sal, Loa, Silo, Sil, Sacl, Cll, Sicl, Sanc, Sic, Cla, and Orga represent sandy, loamy sandy, sandy loam, loam, silty loam, silty, sandy clay loam, clay loam, silty clay loam, sandy clay, silty clay, clay, and organic soils, respectively. The scattered dots represent the raw data. The red dots represent the means of the groups. Different lower-case letters above bars such as a, b, c, d and e indicate significant differences in the N₂O emissions among soil texture classes in the Tukey test (p< 0.05). Error bars are the standard deviations.

The fertilizer application method also had a significant effect on N₂O emission intensity for the N fertilizer application group (Fig. 4). The N₂O emission intensity from the surface banding treatment soil was significantly higher than those of the incorporation and surface broadcast (p< 0.05). The incorporation treatments provided the lowest N₂O emission intensity (p< 0.05).

Figure 4.  N₂O emission intensity from upland fields with different fertilizer application methods in the N fertilized datasets. The scattered dots represent the raw data. The red dots represent the means of the groups. Different lower-case letters above bars such as a, b and c indicate significant differences in the N₂O emissions among the fertilizer application methods in the Tukey test (p< 0.05). Error bars are the standard deviations.

The N₂O emission intensity differed among different chemical fertilizers, but the differences were not significant (Fig. S4). The application of biochar and organic matter in upland fields significantly increased the N₂O emission intensity (p< 0.05) (Fig. 5). While the applications of nitrification inhibitor and urease inhibitor could significantly reduce the N₂O emission intensity (p< 0.05) (Fig. 5). The average soil N₂O emission intensity with a nitrification inhibitor application was 7.4 g N ha^–1 d^–1 (n = 129), which was 60% lower than that of the soil without nitrification inhibitor application (18.4 g N ha^–1 d^–1, n = 1922). The average soil N₂O emission intensity with urease inhibitor application was 6.0 g N ha^–1 d^–1 (n = 40), which was 66.9% lower than that of the soil without urease inhibitor application (18.1 g N ha^–1 d^–1, n = 2011). There were no significant differences in the N₂O emission intensities between with and without the controlled-release fertilizer applications (Fig. 5).

Figure 5.  N₂O emission intensity from upland fields with (A) controlled-release, (B) urease inhibitor, (C) nitrification inhibitor, (D) organic matter, and (E) biochar application in the N fertilized datasets. The scattered dots represent the raw data. The red dots represent the means of the groups. Different lower-case letters such as a and b indicate significant differences in the N₂O emissions among the agriculture practices in the Tukey test (p< 0.05). Error bars are the standard deviations.

The results of SEMs indicated that CLM was the key influencing factor of global N₂O emissions in the non-N fertilized group and agricultural practices were the key influencing factors in the N fertilized group. The most optimized SEM was depicted in Fig. 6A for the non-N fertilized group, while two additional SEMs were constructed, as represented in Fig. S5, accompanied by the presentation of the model fit indices in Table S5. The most optimized SEM included only two groups of influencing factors, namely CLM and SLP in the non-N fertilized group. The CLM group, including MAT and MAP, played the most important role in explaining the variations of soil N₂O emission intensity. CLM can markedly accelerate N₂O emissions with a total standard coefficient of 0.26 (p< 0.001), including a direct standard coefficient of 0.23 and an indirect standard coefficient of 0.03 for the effect of SLP on N₂O emissions. Moreover, the two factors of sand and clay content in the SLP group directly affected the N₂O emission intensity with a standard coefficient of −0.17 (p< 0.01).

Figure 6.  Structural equation modeling of the factors influencing the intensity of N₂O emission (NEI) in the (A) non-N and (B) N fertilized datasets. The concerned factors were divided into five groups: climate (CLM), soil physical properties (SLP), soil chemical properties (SLC), fertilization (FER), and other agricultural practices (PRA). The CLM group includes the box-cox transformed mean annual temperature (bcMAT) and the box-cox transformed mean annual precipitation (bcMAP). The SLP group includes the box-cox transformed sand content (bcSAN), silt content (bcSIL), clay content (bcCLA), and texture (TEX). The SLC group includes the box-cox transformed soil organic carbon (bcSOC) and total soil nitrogen (bcTSN). The FER group includes the box-cox transformed nitrogen fertilizer rate (bcNRA), controlled release fertilizer (CRF), nitrification inhibitor (NI), urease inhibitor (UI), and organic matter (OM). The PRA group includes biochar application (BIO) and irrigation (IRR). The symbols *, **, and *** indicate significance at p< 0.05, p< 0.01, and p< 0.001, respectively.

For the N fertilized group, the most optimized SEM was depicted in Fig. 6B, as well as two additional SEMs, represented in Fig. S6. The model fitting indices of three SEMs are shown in Table S5. The most optimized SEM included five influencing groups of CLM, SLP, SLC, FER, and PRA. The agricultural practices, including FER and PRA, had the most important impacts on N₂O emission in the N fertilized group with the standard coefficients of 0.60 (p< 0.001) and 0.64 (p< 0.01), respectively. The SLC was followed with a standard coefficient of 0.29 (p< 0.001). Finally, the SLP also had a significant effect on N₂O emission with a standard coefficient of 0.11 (p< 0.001). The CLM indirectly influenced soil N₂O emissions by affecting the SLP with the standard coefficient of −0.09 (p< 0.001) in the most optimized SEM.

For non-N fertilized group, the mean soil N₂O emission intensities were 6.0 g N ha^–1 d^–1 with an SD of 6.6 (n = 61), 7.0 g N ha^–1 d^–1 with an SD of 12.5 (n = 57), and 4.1 g N ha^–1 d^–1 with an SD of 5.3 (n = 47) in the subtropical monsoon, the temperate continental and the temperate monsoon zones, respectively, and displayed no significant difference among three climatic zones. Regarding the N fertilized group, large variations of the mean soil N₂O emission intensities among different climatic zones were found, which provided 18.6 g N ha^–1 d^–1 with an SD of 23.5 (n = 202), 28.5 g N ha^–1 d^–1 with an SD of 38.3 (n = 203) and13.4 g N ha^–1 d^–1 with an SD of 35.9 (n = 148) in the subtropical monsoon, the temperate continental, and the temperate monsoon zones, respectively (Table 2). The temperate continental climate zone provided a significantly larger soil N₂O emission intensity than other climate zones in the N fertilized group.

The influencing factors (IF) of MAP, MAT, SOC, TSN, BD, SAN, SIL, CLA, NRA, and NEI refer to the mean annual precipitation (mm), mean annual temperature (°C), soil organic carbon (%), total soil nitrogen (%), bulk density (g cm⁻³), sand content (%), silt content (%), clay content (%), nitrogen rate (kg N ha⁻¹), and nitrogen emission intensity (g N ha⁻¹ d⁻¹), respectively. Different lower case letters (e.g. a vs. b) indicate a significant difference ( p< 0.05) across climatic zones, while displaying the same lower case letters indicates a non-significant difference.

There were differences in the key factors affecting N₂O emissions in the three climatic zones (Fig. 7). In the subtropical monsoon zone, the important influencing factors collectively contributed 45% and 51% of the variation in N₂O emission intensity in non-N and N fertilized groups, respectively. The SLP was the most important explanatory variable, explaining 23% and 26% of the variation independently in subtropical monsoon zones in non-N and N fertilized groups, respectively.

Figure 7.  Variance partition analysis of the contribution of climate (CLM), soil physical properties (SLP), soil chemical properties (SLC), and other agricultural practices (PRA, including N fertilizer rate) to N₂O emission intensity for the subtropical monsoon, temperate continental and temperate monsoon climatic zones in the (A, C, E) non-N fertilized groups and (B, D, F) N fertilized groups, respectively. All components contributing < 1% were omitted.

In the temperate continental zone, these influences explained 31% and 42% of the variation in N₂O emission intensity for non-N fertilized and N fertilized groups, respectively. Among them, CLI significantly affected soil N₂O emission intensity in the non-N fertilized group, independently explaining 22% of the variation in N₂O emissions, while 28% of the variation in N₂O emissions in the N fertilized group was independently explained by SLP.

In the temperate monsoon zone, these influences explained 47% and 62% of the variation in N₂O emission intensity for non-N fertilized and N fertilized groups, respectively. CLM also was the most important influencing factor, and can independently explain 23% of the variation in N₂O emissions in the non-N fertilized group. Climatic and soil physical factors exhibited a robust interaction, jointly contributing to 13% of the variation in N₂O emissions. In the N fertilized group, the PRA was the important influencing factor and explained 17% of the variation in N₂O emissions. The combined effects of the four influencing groups synergistically and collectively accounted for 19% of the variation in N₂O emissions. In summary, the SLP was the key influencing factor in the subtropical monsoon zone, CLM and SLP were the key influencing factors in non-N and N fertilized groups, respectively, in the temperate continental zone, and in the temperate monsoon zone, CLM and PRA were the key influencing factors in non-N and N fertilized groups, respectively.

The influencing factors, including MAP, MAT, TEX, BD, pH, SOC, and TSN, were all significantly correlated with N₂O emission intensity in non-N fertilized and N fertilized groups (p< 0.05), which were similar to previous studies (Attard et al., 2011; Meurer et al., 2016; Oertel et al., 2016; Smith, 2017; Mehnaz et al., 2018; Lai et al., 2019; Shcherbak and Robertson, 2019; Yang et al., 2022). Among them, MAP, MAT, TEX, SOC, and TSN factors contributed to N₂O emissions, mainly by creating an environment conducive to N₂O production and providing reaction substrates. For example, MAP created an anaerobic environment to promote denitrification (Gu and Riley, 2010). BD and pH were negatively correlated with N₂O emission, and as they increased, they will be detrimental to N₂O production. For example, acidic soils (e.g., pH = 5.0) emit more than three times more N₂O than the alkaline soils (pH = 8.0) (Wang et al., 2018). Significantly acidic soils (pH< 6.0) emit a higher proportion of N₂O than N₂; however, when pH = 6.0, soils emit approximately equal amounts of N₂O as N₂ (Rochester, 2003).

Regarding agricultural management factors, except for N fertilizer application, N₂O emissions are not correlated with either tillage or irrigation in the non-N fertilized group. However, in the N fertilized group, N₂O emissions are not correlated with tillage but are significantly correlated with irrigation. This result is different from other studies. Previous studies have shown that tillage and irrigation influenced N₂O emission by altering soil structure, moisture, and other soil properties (Deng et al., 2018; Pareja-Sánchez et al., 2020). The reasons for this difference were the large spatial variability of the global data and deficiencies in some recording data on tillage and irrigation, leading to the effect of tillage and irrigation on N₂O emissions being less significant, especially in the non-N fertilized groups.

In the N fertilized group, the cultivation of legumes, with the lowest rate of N fertilizer (99.7 kg N ha⁻¹), provided the fewest N₂O emissions, which was consistent with previous studies (Jensen et al., 2012; Schwenke et al., 2015). The use of legumes has been proposed as a mitigation strategy for two reasons. On the one hand, legumes can fix N from the atmosphere with N-fixing rhizobia as an alternative to the mineral N fertilizer, which yield high greenhouse gas emissions in their manufacturing process (Brentrup and Pallière, 2008); on the other hand, they provide a N source that was better synchronized with the plant's needs and therefore a higher utilization efficiency. Other cultivated plants had significant differences in N₂O emissions, mainly because 1) the rates of N fertilizer on different plants were applied and 2) different plants take up different amounts of soil C/N and thus reduce different amounts of soil SOC and TSN, which in turn affect N₂O emissions, especially in fertilized soils (Abalos et al., 2014).

Significantly higher N₂O emissions came from the silty clay and organic soils in the non-N fertilized group and from the organic soil in the N fertilized group, and the smallest N₂O emissions were from the sandy clay soils in both groups. Compared to sandy clay soils, silty clay soils have a finer texture, which tends to create conditions favorable to denitrification, and therefore have a higher N₂O emission intensity (Meurer et al., 2016; Wang et al., 2021). Organic soils have higher N₂O emissions because of the higher organic matter content of the soil, which provides more carbon sources and increases microbial activity, thereby promoting nitrogen oxidation and nitrification and increasing N₂O emission. (Liu et al., 2023). The results of Teepe et al. (2000) also demonstrated that N₂O emissions from the organic soils were higher than mineral soils in similar situations. Annual emissions of N₂O from cultivated soils, where birch was planted for 22 years and cut, were about twice as high as those from adjacent forests (Maljanen et al., 2003).

In the group without N fertilizer, the SEM with CLM and SLP as latent variables had a better fit than other SEM at a global scale. The result show that CLM was the determining factor that affected the intensity of N₂O emissions in the non-N fertilized group. The climatic factors, with standard coefficients of 0.26, had a greater effect on the intensity of N₂O emissions (total standard coefficients = 0.43) than other influencing factors. The result is in line with the result of Griffis et al. (2017), which considered climatic factors as the most critical influence factor. Other studies also showed that climatic factors (air temperature and precipitation) affected microbial activity as well as soil porosity, soil moisture, which subsequently influenced N₂O emissions (Lai et al., 2019; Zhang et al., 2019; Wang et al., 2020). In addition, SLP is another important influence factor of N₂O emission, which has two main manifestations. On the one hand, it affects N₂O emission, such as through its potential influence on soil texture; on the other hand, climate factors act upon it, thus affecting N₂O emission. Studies showed that the effect of soil physical properties on N₂O emission partly originated from the long-term influence of climate (MAP and MAT) on pedogenic processes (Wang et al., 2018).

In the group with N fertilizer, the best SEM contained five potential variables as PRA, FER, SLC, CLM, and SLP at a global scale. The agricultural practices including PRA and FER had greater effects on N₂O emission than other influencing factors. This result occurred because FER could either promote N₂O emissions or inhibit N₂O emissions, such as the application of N fertilizer or inhibitors (Soares et al., 2023). PRA also was an important influencing factor of N₂O emission, e.g., straw incorporation can facilitate N₂O emission by affecting soil N effectiveness (Bhattacharyya et al., 2012; Li et al., 2020).

In three climatic zones at a global scale, the key factors affecting N₂O emission intensity differed under the variance partitioning analysis in non-N and N fertilized groups. In the non-N fertilized group, SLP (including SIL and SAN) made the largest contribution to N₂O emission intensity in the subtropical monsoon zone (23%), but CLI (including MAP and MAT) made a great contribution to N₂O emission intensity in the temperate continental (22%) and monsoon (23%) zones. The different key influence factors in the three climatic zones might result from the following two reasons. First, in the subtropical monsoon zone, climatic factors have less influence on N₂O emissions due to adequate temperature and precipitation with less variance than in other climatic zones. In the temperate continental and monsoon zones, inadequate and greater variations in temperature and precipitation, as well as the characteristics of four distinct seasons, become sensitive factors that affect N₂O emissions. This is consistent with the findings of Roelandt et al. (2005) in that spring temperature and summer precipitation explain 35% of the variance in annual N₂O emissions in temperate climates. Secondly, SLP has a greater impact on N₂O emissions in the subtropical monsoon zone because of a larger variance. There is an average SD of 21% in sand content in the subtropical monsoon zone than in temperate continental and monsoon zones with SDs of 17.3% and 17.7% in the collected datasets, which can affect porosity and soil moisture, further influencing denitrification reaction conditions (Meurer et al., 2016). In addition to CIL and SLP, SLC is also an important influencing factor of N₂O emissions which affects microbial activity, thus influencing the N₂O emission intensity (Kim et al., 2019).

In the N fertilized group, SLP had the greatest impact on soil N₂O emissions in the subtropical monsoon (26%) and temperate continental (28%) zones, while PRA had the greatest impact in the temperate monsoon zone (17%). The difference in the key factors affecting N₂O emission in the three climate zones also indicated that the factors affecting N₂O emission have spatial variability, which cannot be generalized (Aliyu et al., 2018). In this study, there are three main reasons for the difference in key influencing factors of N₂O emission. First, the SAN was smaller and CLA was larger and had larger variation in the subtropical monsoon and temperate continental zones than temperate monsoon zone, therefore SLP was more sensitive to N₂O emissions in the subtropical monsoon and temperate continental zones. The mechanisms by which SAN and CLA affect N₂O emissions have been described above. Second, in the temperate monsoon zone, PRA accounts for tillage and irrigation in addition to the rates of fertilizer application, whereas in the subtropical monsoon and temperate continental zones, tillage and irrigation information is rarely counted. Additionally, in the temperate continental zone, NI also had an important influence on N₂O emission intensity, so the contribution of PRA to N₂O emission was larger than that in the subtropical monsoon zone. As agriculture practices such as NI application were less applied in the selected datasets, the impact of agricultural practices on N₂O emission was underestimated. The agricultural practices affected the substrate for nitrification and denitrification reactions, soil property, and microbial activity, thus influencing N₂O emissions (Kim et al., 2021). Third, the variation of N fertilizer application in the temperate monsoon zone was larger than in other climatic zones in the collected datasets, which also was one reason of the contribution of PRA to N₂O emission was larger than in other climatic zones. In the future, additional field N₂O emission measurements with various agricultural practices, especially tillage, irrigation, and NI, for each climatic zone should be observed and collected to evaluate their effects on N₂O emissions.

The mean N₂O emission intensities of the non-N fertilized group (5.3 g N ha^–1 d^–1) were lower than that of the N fertilized group (17.8 g N ha^–1 d^–1), which showed that the application rate of N fertilizer significantly increased the N₂O emission by 70%. Studies have shown that the application of N fertilizer provided more inorganic N (e.g., NH₄⁺ and NO₃⁻) as a substrate for nitrification and denitrification processes, which were responsible for most of the N₂O emission from soils (Kim et al., 2021). Grave et al. (2018) also showed that the cumulative N₂O emissions from urea and slurry applied to tilled soils increased by 33% and 46% compared to the control treatment. Although the non-N fertilized group emits less N₂O than the N fertilized group, as the world's population grows, the use of N fertilizers during crop growth is essential to ensure high crop yields (Xia and Yan, 2023a, b). Higher fertilizer application increases N₂O emissions and other N losses, while lower fertilizer application leads to low crop yields, so there is a need to find better agricultural practices to pursue high yields and low N losses. Ren et al. (2022) showed that a 15–19% reduction in N fertilizer application (application rate of about 210–220 kg ha⁻¹) in wheat, maize, and rice in China resulted in an increase in yields by 10%–19%, a reduction in N surplus by 40% without actually changing farmers’ operational practices, which would ensure yields and also reduce environmental N pollution. Because of geographical differences, the different optimal amount of N fertilizer needs to be explored for different areas.

Except for the rate of N fertilizer application, other agriculture practices also affected N₂O emissions (Bell et al., 2016). Organic matter fertilizer applications significantly increased N₂O emissions (p< 0.05), which was because of the increased content of soil organic carbon (Lazcano et al., 2021; Wang et al., 2021). The biochar application increases N₂O emissions because it may increase the nitrification pathway, and increase the abundance of fungal-nirK genes (Wu et al., 2021). Although N₂O emission intensity differed among different types of chemical fertilizer applications, the differences were not significant (Fig. S4). This may be related to the various amounts of the different N fertilizer types. The controlled-released application of fertilizers did not significantly reduce N₂O emissions because of offsetting factors. Although N₂O emissions were initially reduced by the application controlled-released fertilizers, the continuous release of mineralized N into the soil can expand N₂O emissions (Lewis, 2010; Braun and Bremer, 2018). Nitrification and urease inhibitors in this study effectively reduced N₂O emissions by 59.8% and 66.9% (Fig. 5). Nitrification inhibitors generally work to inhibit soil enzymes, which were responsible for nitrification and denitrification and, ultimately, reduced N₂O emissions (Ruser and Schulz, 2015). Urease inhibitors slow down the urea hydrolysis process by blocking the active site of urease, thus limiting the amount of N lost to the atmosphere (Byrne et al., 2020). Studies have shown that the application of nitrification inhibitors and urease inhibitors not only reduces N₂O emissions, but also significantly increases grain yields by 10.0% and 7.1%, respectively, and reduces other reactive N loss (ammonia, NH₃, N leaching, and runoff) (Xia et al., 2017). These results also show that the type of N fertilizer application has significant effects on N₂O emissions from upland soils, which coincides with the results of Wang et al. (2021). Currently, N₂O emissions have been increasing, and as a significant greenhouse gas, it is associated with a number of environmental problems, such as climate warming and sea level rise. So it is necessary to reduce agricultural N₂O emissions while meeting the growing demand for food and other agricultural products. The results of this study provide further evidence that proper agricultural practices such as optimal fertilizer application and nitrification inhibitors application can reduce N₂O emissions.

To achieve the ambitious goal of mitigating greenhouse gas emissions, it is critical to understand and quantify soil nitrous oxide (N₂O) emissions, as well as the primary influencing factors of N₂O emissions on a global scale. Here, we performed descriptive statistics, correlation, ANOVA analysis, and structural equation analysis using N₂O emission data from the global datasets to evaluate the hierarchy of influencing factors of N₂O emissions on a global scale for upland systems with N and non-N fertilizer applications. We found that N₂O emission intensity was significantly correlated with the mean annual precipitation and temperature, soil pH, soil texture, soil bulk density, soil organic carbon, total soil nitrogen, and pH in both fertilized and non-N fertilized groups. For the group without N fertilization, the climate-related factors played a large part in determining the intensity of N₂O emissions at the global scale. For the group with N fertilization, the fertilizer rate and other agricultural practices had the greatest influences on the intensity of N₂O emissions. The soil properties and climate-related factors were the key influencing factors in non-N fertilized groups and the agricultural practices and soil properties were the key influencing factors in N fertilized groups in the three climatic zones of concern. The mean N₂O emission intensity of the non-N fertilized group was significantly lower than that of N fertilizer (5.34 g N ha^–1 d^–1 versus 17.83 g N ha^–1 d^–1, p< 0.05). In the N fertilization group, we found that the biochar and organic fertilizer application significantly increased N₂O emissions but the nitrification and urease inhibitors effectively reduced N₂O emission intensity by 60%−66.9% (p< 0.05). The cultivated plants provided significant differences in N₂O emissions, noting that legume cultivation significantly reduced N₂O emissions (p< 0.05). In summary, both fertilized and natural soil release large amounts of N₂O, and our results suggest that proper agricultural management practices, such as proper N fertilizer application rates and types, can reduce N₂O emissions and thus slow down the rate of global warming. In addition, this study further clarifies the hierarchical relationship between the factors affecting N₂O emissions and provides a theoretical basis and direction for controlling N₂O emissions.

Acknowledgements. This study was financially supported by the National Natural Science Foundation of China (Grant No. 42161144002), the National Key Research and Development Programs of China (Grant No. 2022YFE0209200-03), the Suzhou Agricultural Science, Technology and Innovation Programs of Suzhou Agricultural Department (Grant No. SNG2022011), and the special fund of State Environmental Protection Key Laboratory of Formation and Prevention of the Urban Air Pollution Complex (SEPAir-2022080590).

Electronic supplementary material: Supplementary material is available in the online version of this article at https://doi.org//10.1007/s00376-024-3234-7.