Soil samples (0–15 cm) were collected from agricultural fields in two locations: Shang Zhuang (SZ), Beijing (39°48′N, 116°28′E) and Luan Cheng (LC), Hebei (37°53′ N, 114°41′E), North China Plain, in October 2020. The annual average temperature is 12.5°C, and the annual precipitation is 500–700 mm with high variation among different years. The cropping system in this region is winter wheat-summer maize rotation. The fertilizer application rates were 280 and 600 kg N ha⁻¹ year⁻¹ in SZ and LC soils, respectively. Collected soils were air-dried and sieved to 2 mm. Visible roots and leaves were removed with tweezers and the soil was immediately stored at 4°C until the beginning of laboratory experiment. The soils are both classified as silt loam, with 36.1% sand, 56.4% silt, and 7.5% clay for the SZ soil and 29.2% sand, 64.1% silt, and 6.7% clay for the LC soil. For the SZ soil, pH was 7.89, bulk density was 1.02 g cm⁻³, soil organic carbon was 10.93 g kg⁻¹, total N was 1.13 g kg⁻¹, NH₄⁺-N was 3.07 mg kg⁻¹, and NO₃⁻-N was 22.5 mg kg⁻¹. For the LC soil, pH was 7.92, bulk density was 1.00 g cm⁻³, soil organic carbon was 19.82 g kg⁻¹, total N was 2.11 g kg⁻¹, NH₄⁺-N was 2.08 mg kg⁻¹, and NO₃⁻-N was 30.49 mg kg⁻¹.

Soils (20 g oven-dry equivalent) were placed into 120 ml incubation flasks and distilled water was added to the soils to below the target moisture contents [i.e., 40, 60, 70, 80, 90, 95, 100, and 120% water-filled pore space (WFPS)]. The microcosms were then pre-incubated at 25°C for 7 days to initiate microbial activity. For each moisture content treatment, ¹⁵NH₄Cl (10.08 atom%) + KNO₃ or K¹⁵NO₃ (10.16 atom%) + NH₄Cl were applied at a rate of 50 mg NH₄⁺-N kg⁻¹ and 50 mg NO₃⁻-N kg⁻¹ after pre-incubation. To assure uniform distribution, 2 ml of ¹⁵N solution was applied in water solution and sprayed onto the soils to obtain the target moisture content. The experimental design and treatment application were set up as completely randomized blocks and incubated in dark for 48 h at 25°C after ¹⁵N application.

Each treatment was replicated three times for gas analyses, with gas samples collected at 12, 24, and 48 h. Before sampling, the flasks were flushed with ambient air using a multiport vacuum manifold, and the N₂O concentration in the headspace was then measured. Thereafter, the flasks were immediately sealed for 12 h and N₂O concentration was measured again. The difference between the two N₂O concentrations was used to calculate the N₂O production rate. The concentrations of N₂O and CO₂ were determined using gas chromatography (Agilent 7,890, Santa Clara, CA, United States) and the ¹⁵N signature of N₂O was determined using a Thermo Finnigan MAT-253 spectrometer (Thermo Fisher Scientific, Waltham, MA, United States). Another group of flasks, also replicated three times, were used for soil sampling at 0.5, 12, 24, and 48 h after N application. Soils were extracted with 1 M KCl (20 g soil to 100 ml KCl solution), shaken for 1 h, and filtered. The concentrations of NH₄⁺-N and NO₃⁻-N in the extracts were measured using a continuous-flow analyzer (Skalar Analytical, Breda, Netherlands). Isotope analysis of NH₄⁺-N and NO₃⁻-N were performed on aliquots of the extracts using a diffusion technique (Brooks et al., 1989) and the ¹⁵N isotopic signature was measured by isotope ratio mass spectrometry (IRMS 20–22, Sercon, Crewe, United Kingdom).

Nitrous oxide and CO₂ fluxes (F, μg N kg⁻¹ h⁻¹ or mg C kg⁻¹ h⁻¹) were determined from the concentrations at each sampling time, using the background N₂O and CO₂ concentrations in the ambient air as the initial time point, which were calculated as follows:

where ρ is the density of gas under standard conditions (kg m⁻³), ΔC is the variation in gas concentration during the flask-covering period (the units of N₂O and CO₂ are ppbv and ppmv, respectively), and V is the effective volume of a given flask (m³), T is the incubation temperature (°C), Δt is the incubation time (h), and W is the weight of soil (oven-dried basis, kg).

The contributions of denitrification, Cd, and nitrification, Cn to the production of N₂O were calculated using the following equation (Stevens et al., 1997):

where a_N2O is the ¹⁵N atom% enrichment of the N₂O produced by both processes, and a_NO3 and a_NH4 are the ¹⁵N atom% enrichment of soil NO₃⁻ and NH₄⁺ at the time of gas sampling.

Rates of N₂O production from nitrification (N₂O_n) and denitrification (N₂O_d) were calculated as follows:

where N₂O_T is the total N₂O production rate from the soils, N₂O_T = N₂O_n + N₂O_d.

Since the concentrations and abundances of NH₄⁺ at 48 h could not be reliably determined in most treatments, the average Cd, Cn, N2On and N2Od over the first 24 h incubation were used to analyze the rates of N₂O production from nitrification and denitrification.

Data on the N₂O production rates of nitrification and denitrification were collected from published peer-reviewed journal articles. The following criteria were used for data collection: (1) incubation experiments used agricultural soils solely; (2) soil moisture metric was expressed as WFPS. Meanwhile, soil characteristics and incubation conditions, including pH, BD, clay content, SOC content, concentrations of TN, NH₄⁺, and NO₃⁻, incubation temperature and WFPS, were collected. GetData Graph Digitizer 2.26 was used when data were only graphically shown. The autotrophic nitrification and heterotrophic nitrification were summed and treated as nitrification during the data analysis if they were reported as individual pathways in the literature. In total, 80 groups of data from 17 studies were obtained (Supplementary Table S1).

All statistical analyses were evaluated by one-way analysis of variance (ANOVA) for comparisons among multiple factors and t-test for contrasts between two factors, followed by the least significant difference test at P<0.05. The relationships between the contributions of nitrification and denitrification to N₂O production or their rates and the controlling factors were examined by correlation and regression analysis. All statistical analyses were carried out in SPSS v25.0 software for Windows (SPSS Inc., Chicago, United States).

The concentration of soil NH₄⁺ decreased over the incubation course in all moisture treatments (Figures 1A,B). For both SZ and LC soils, the declining rates of NH₄⁺ over the first 24 h were nearly twice larger in the treatments of WFPS ≤80% than in the treatments of WFPS ≥90%. After the first 24 h, the declining rate slowed down clearly when WFPS ≤80%, especially for the LC soil (Figure 1B), while it nearly kept constant under WFPS ≥90%. Among all the WFPS treatments, the largest consumption rate of NH₄⁺ occurred at 60% WFPS for both SZ and LC soils.

The concentration of soil NO₃⁻ increased as NH₄⁺ was nitrified (Figures 1C,D). In correspondence to the changes in NH₄⁺ concentration, NO₃⁻ concentration increased faster when WFPS ≤80% than when WFPS ≥90%, especially for the LC soil during the first 24 h. The initial NO₃⁻ concentration exhibited large variances for different moisture contents, since nitrification increased NO₃⁻ concentration under low moisture content while denitrification reduced NO₃⁻ concentration under high moisture during the pre-incubation period. As the initial NO₃⁻ concentration markedly reduced as WFPS increased, the NO₃⁻ concentration varied largely at the end of incubation especially for the LC soil, changing from 170.3 to 75.0 mg N kg⁻¹ as WFPS increased from 60 to 120%.

The N₂O production rate changed substantially with moisture content and time (Figure 2). At the beginning of incubation, high N₂O production rates (> 5 μg N kg⁻¹ h⁻¹) occurred under 80% ≤ WFPS ≤100% in the SZ soil and under 70% ≤ WFPS ≤100% in the LC soil, whereas the rates remained low under the lower or higher moisture conditions. As the incubation proceeded, the N₂O production rate first increased and then decreased under the intermediate moisture conditions (e.g., WFPS = 70, 90, and 95%) in the SZ soil, but consistently reduced under all moisture conditions in the LC soil. Finally, the N₂O production rates declined to below 5 μg N kg⁻¹ h⁻¹ under all moisture contents for both soils at the end of incubation. By contrast, CO₂ production rates were higher at WFPS ≥90% than at WFPS <90% for both soils, except for 95% WFPS in the LC soil (P<0.05; Supplementary Figure S1).

The ¹⁵N enrichment of N₂O remained between the ¹⁵N enrichments of NH₄⁺ and NO₃⁻ during the first 24 h, illustrating that N₂O was derived from both nitrification and denitrification (Supplementary Figure S2). The average contribution of denitrification to N₂O production, Cd, increased with moisture content in the SZ and LC soils up to 100 and 95% WFPS, respectively, after which Cd declined significantly (Figure 3). In both soils, nitrification was the main pathway producing N₂O under low moisture conditions while denitrification dominated N₂O production under high moisture conditions, with the threshold occurred at 70 and 60% WFPS for the SZ and LC soils, respectively. Denitrification contributed more than 65% of the total N₂O production when WFPS ≥70%, and this percentage promoted as the incubation proceeded (Supplementary Table S2).

Nitrous oxide production rates derived from nitrification (N₂O_n), denitrification (N₂O_d) and the combined processes (N₂O_T) responded to moisture change in a pattern similar to Gaussian function in both SZ and LC soils (Figure 4). As moisture increased, the N₂O_n increased slowly, reaching peaks around 2.5 μg N kg⁻¹ h⁻¹ in both SZ and LC soils, while the N₂O_d increased steeply, reaching peaks of 10.1 and 12.5 μg N kg⁻¹ h⁻¹ in the SZ and LC soils, respectively. Correspondingly, the optimal WFPS with respect to the peak rates were the same for the nitrification and denitrification processes (90% WFPS) in the SZ soil, but diverged for the two pathways (80 and 95% WFPS, respectively) in the LC soil. The N₂O production rates remained below 3 μg N kg⁻¹ h⁻¹ under either low or flooded moisture condition.

By synthesizing literature data across global agricultural soils, moisture (WFPS) and incubation temperature (T) were found to be the most significant factors controlling the contributions of nitrification and denitrification to N₂O production (Table 1), with WFPS exerting a stronger correlation (R = 0.45) than T (R = 0.37; Figure 5; Supplementary Figure S3). Compared with the literature data (R = 0.36), the measured data in this study exhibited a stronger positive correlation between Cd and WFPS (R = 0.73; Figure 5). Furthermore, a stronger correlation between Cd and WFPS occurred in alkaline soils than in acidic soils (Supplementary Figure S4A). Similarly, compared with carbon-rich soils with SOC ≥ 4%, mineral soils with SOC < 4% showed a stronger correlation (Supplementary Figure S4B).

Based on the literature synthesis, the N₂O_n and N₂O_d generally first increased and then decreased as WFPS increased (Figure 6). The relationships between the N₂O production rates of nitrification and denitrification and WFPS were fitted by Gaussian function. Compared with nitrification (Figure 6A), denitrification (Figure 6B) showed a smaller standard deviation, 5% vs. 14%, and a higher maximum rate, 106 vs. 12 μg N kg⁻¹ h⁻¹, though both of their peak rates occurred at around 85% WFPS. The correlations between N₂O production rates and various soil properties were also analyzed (Supplementary Table S3). The results indicated that NH₄⁺ and NO₃⁻ concentrations were the most powerful drivers to explain the changes in N₂O_n and N₂O_d. Both N₂O_n and N₂O_d increased positively with the increases in NH₄⁺ (P<0.05; Supplementary Figures S5A,C) and NO₃⁻ concentrations (P<0.01; Supplementary Figures S5B,D), though the variances of rates were large as the concentrations were high.

Both laboratory incubation and literature synthesis showed that nitrification and denitrification dominated N₂O production under low and high moisture conditions, respectively. Under high moisture conditions as soil oxygen availability was constrained, denitrification outcompeted nitrification as the main source of N₂O production (Smith, 2017; Song et al., 2019; Chang et al., 2022), which was aligned with other experiments (Pihlatie et al., 2004; Friedl et al., 2021). The dominant pathway of N₂O production switched between 60 and 70% WFPS (Figure 5), depending on soil properties and climatic conditions. For instance, the thresholds for SZ and LC soil were 70 and 60% WFPS (Figure 3), respectively. This is because the SOC content was higher in the LC soil (19.82 g kg⁻¹) than in the SZ soil (10.93 g kg⁻¹), stimulating N₂O production by promoting denitrification process (Ruser et al., 2006; Chantigny et al., 2013). Besides, the N₂O production rate in the LC soil (1.02 μg N kg⁻¹ h⁻¹) was almost 10 times that in SZ soil (0.1 μg N kg⁻¹ h⁻¹) under 60% WFPS, further indicating the dominating effects of denitrification in the N₂O production in the LC soil. The literature synthesis also confirmed that large SOC content increased the contribution of denitrification to N₂O production under relatively low soil moisture content (Supplementary Figure S4B). Besides SOC, other factors such as BD, NH₄⁺ and NO₃⁻ concentrations, and especially incubation temperature, also modulated the contributions of nitrification and denitrification to N₂O production (Table 1), which might explain why the contribution proportions between nitrification and denitrification varied significantly among different soils even though the soil moisture status were similar (Figure 5).

Accurately determining the contributions of nitrification and denitrification to N₂O production is crucial to evaluate N₂O emissions from agricultural soils (Zhu et al., 2013). Currently, different approaches were used to quantify these contributions, including ¹⁵N site preference (Thilakarathna and Hernandez-Ramirez, 2021), acetylene inhibition (Pihlatie et al., 2004), and ¹⁵N tracing techniques (Friedl et al., 2021). The applications of these approaches often caused large discrepancies in quantifying Cd and Cn under different moisture conditions (Butterbach-Bahl et al., 2013), and likely resulted in different contribution proportions even though the experimental setup and the operating conditions were the same (Zhu et al., 2013). Therefore, a careful comparison among different approaches and developing a guideline or protocol for using these approaches merit further investigations. Although certain factors such as pH value and N concentrations exerted insignificant impacts on the contribution of different pathways to N₂O production (Table 1), their integrative impacts remain unclear (Hu et al., 2015). In addition, factors such as moisture and temperature, often changed synchronously in fields (Song et al., 2018), and studying their integrative impacts will significantly improve our understanding of N₂O emission dynamics and facilitate N₂O abatement (Mathieu et al., 2006).

Both laboratory study and literature synthesis validated the hypothesis that the rates of N₂O production from both nitrification and denitrification first increased and then decreased as soil moisture increased (Figures 4, 6). The relationships between N₂O production rate and moisture content followed the classic hole-in-pipe model (Davidson et al., 2000), though the rates changed with soil properties (Figure 4). For instance, the LC soil produced generally larger N₂O_d than the SZ soil, since it contained more NO₃⁻ and SOC, which stimulated N₂O production from denitrification under high moisture content (Smith, 2017). By comparison, the two soils exhibited approximate N₂O_n due to the similar NH₄⁺ concentrations. The literature synthesis further confirmed that NO₃⁻ and NH₄⁺ were the two most important factors to determine N₂O production rates (Supplementary Table S3). Interestingly, NO₃⁻ concentration was the most powerful driver to explain the changes in N₂O derived from nitrification, although its explaining power was close to that of NH₄⁺ concentration. This result might be caused by the large N₂O production rates from nitrification under high NO₃⁻ concentrations and large soil moisture contents (Supplementary Figure S5) and warrant further investigations. However, the rates of N₂O_d and N₂O_n depended on not only the above factors but also moisture content, and their interactions control N₂O emission from soils (Zhu et al., 2013). Therefore, higher substrate concentration unnecessarily resulted in larger N₂O emissions, as being observed in many laboratory and field experiments (Senbayram et al., 2012; Liu et al., 2018).

In contrast to the first increased and then decreased N₂O production rates in response to increase in soil moisture content from the laboratory incubation in this study, the studies in the collected literatures presented divergent consequences among different experiments (Supplementary Table S1). Among the 17 collected studies, as moisture increased, only five reported a decline in N₂O production rate for nitrification and no study found a decline for denitrification. The underrepresented decline in the rates can be mainly attributed to the insufficient gradients and inadequate levels of moisture content applied in these studies, which commonly used soil moisture containing less than four levels and below 90% WFPS (Supplementary Table S1). Such sparse moisture levels likely did not capture the inflection point of N₂O production rate (Barton et al., 2015), while the low moisture condition might not be adequate to capture the turning point (Bateman and Baggs, 2005; Liu et al., 2016). Therefore, N₂O emission under relatively high moisture conditions with sufficient moisture treatments deserves further investigations. The interactions of soil moisture with other factors such as SOC content (Qin et al., 2017), nutrient availability (Senbayram et al., 2012), and pH value (Zhang et al., 2015) together determine the relationship between N₂O emission rates and moisture contents (Zhu et al., 2020).

Both laboratory study and literature synthesis illustrated that N₂O emissions declined as moisture content exceeded certain threshold. Current models using linear or exponential relationships between N₂O production rate and moisture content could significantly overestimate N₂O emissions from agricultural systems (Yue et al., 2019; Wang et al., 2021), especially as the intensive irrigation and extreme rainfall are projected to increase under climate change scenarios (Smith et al., 2017). Therefore, comprehensive relationships that can capture the first increased and then decreased N₂O production rates in response to elevated soil moisture content are required. However, the large variances in N₂O production rates of both nitrification and denitrification among different studies induce great challenges to develop such a relationship. One potential breakthrough can be to quantify this relationship for different types of soils by incorporating intense moisture treatments similar to this study. Meanwhile, additional experiments are required to quantify the impacts of other key factors, such as temperature, NO₃⁻ and NH₄⁺ concentrations and their interactions, on the relationship. Once sufficient data measured using the same experimental protocol are collected, it will be possible to derive quantitative relationships between N₂O production rate and moisture content across different soils by using a general function, such as Gaussian function, with parameters depending on key edaphic and climatic drivers (Yan et al., 2018).