A two-year field experiment was conducted during the rice growing seasons (May–October) of 2019 and 2020 at the Liaoning Irrigation Experiment Center Station (42°08′59″N, 120°30′44″E; 47 m altitude) in Shenyang, Liaoning Province. The site experiences a temperate continental monsoon climate with a mean annual temperature of 7.5°C and annual precipitation of 673 mm. Local meteorological data during the study period were recorded by an on-site automatic weather station (Figure 1). The soil is classified as clay loam, with the topsoil (0–15 cm) containing organic matter 22.3 g kg⁻¹, total nitrogen 0.78 g kg⁻¹, alkali-hydrolyzable nitrogen 75.4 mg kg⁻¹, Olsen-P 18.4 mg kg⁻¹, exchangeable potassium 81.3 mg kg⁻¹, pH 7.40, and bulk density 1.50 g cm⁻³. A mid-late season rice cultivar (Oryza sativa L. cv. Shennong 9765), characterized by high yield, superior grain quality, and strong disease resistance, was used in this study.

The experiment followed a randomized complete block design with two irrigation treatments—continuous flooded irrigation (I_CF) and alternate wetting and drying irrigation (I_AWD)—each replicated three times. Six concrete lysimeters (2.5 × 2 × 1.5 m) were used, each equipped with an individual water gauge and meter to precisely control irrigation. Waterproof sealing prevented lateral water and nutrient exchange, and a movable rainout shelter excluded natural precipitation. In I_CF plots, a 1–5 cm flooding depth was maintained from transplanting until two weeks before harvest. In I_AWD plots, standing water (1–3 cm) was maintained for the first two weeks after transplanting, after which fields were allowed to dry until the soil water potential at 0–15 cm depth reached –15 kPa, as monitored by installed tensiometers. Floodwater was then reapplied, and this cycle was repeated until two weeks pre-harvest. Soil moisture is monitored by recording water table and tensiometers at 8 a.m. and 2 p.m. every day to strictly control water in accordance with the irrigation regime. The experimental setup and agronomic management were consistent with a long-term study initiated in 2014 assessing I_AWD effects on rice yield, water productivity, and grain quality. This long-term alternate wetting and drying irrigation experiment initially primarily explored the effects of alternate wetting and drying irrigation on rice growth, water use efficiency, and rice yield. Subsequently, it investigated the impacts on rice photosynthetic characteristics, dry matter accumulation, and rice quality. Recently, it has mainly focused on its environmental effects, including leaching losses, greenhouse gas emissions and the impacts on the cycling of carbon, nitrogen and other nutrients (Zheng et al., 2018a, 2018b; Liu et al., 2022).

Seedings were manually transplanted on 24 May 2019 and 28 May 2020 to each completely cleaned, plowed, and flatted plot. Inter and inter-row spacing of seedings were 16 cm and 30 cm with four seedings per hill, respectively. A total of 210 kg ha⁻¹ urea (46% N) was applied as N fertilizer (43% in the basal fertilizer period, 43% in the tiller fertilizer period, and 14% in the panicle fertilizer period), a total of 60 kg ha⁻¹ P₂O₅ (12%) was applied as P fertilizer (100% in the basal fertilizer period) and a total of 37.5 kg ha⁻¹ K₂O (50%) was applied as K fertilizer (50% in the basal fertilizer period, 50% in the tiller fertilizer period). Diseases and insects were controlled with chemicals, and weeds were manually cleaned by farmers.

Prior to analysis, the Shapiro-Wilk normality test was used to examine the normal distribution characteristics of the data for each indicator. One-way analysis of variances (ANOVA) was conducted to assess whether there are significant differences in the average N₂O concentration within soil profiles at depths of 0–10, 10–20, 20–30, 30–40, and 40–50 cm between continuously flooded irrigation (I_CF) and alternate wetting and drying irrigation (I_AWD) during the basal fertilizer period, tiller fertilizer period, and panicle fertilizer period in 2019 and 2020. Significant differences between treatments were subsequently identified using Tukey’s HSD test at the 5% probability level. Linear regression analysis was performed to explore the relationship between soil surface N₂O flux and N₂O concentrations at depths of 0–10, 10–20, 20–30, 30–40, and 40–50 cm. Partial least-squares path model (PLS-PM) was conducted to infer the direct and indirect effects of N₂O accumulation in different soil profiles (depths of 0–10, 10–20, 20–30, 30–40, and 40–50 cm) on soil surface N₂O emission under different irrigation regimes (I_CF and I_AWD). R² donates the proportion of variance explained. Values adjacent to arrows represent standardized path coefficients. The “+” “−” indicate significant positive and negative effects, and absolute value of path coefficients are proportional to the strength of the effects. The above statistical analyses were conducted and visualized using the R software (version 4.4.3) with the package of “stats” for ANOVA, the package of “lm” for linear regression analysis, and the package of “plspm” for PLS-PM.

Our measurements revealed systematic patterns in how I_AWD irrigation reshapes the distribution and dynamics of N₂O within the soil profile (Figure 3). Temporally, N₂O concentrations peaked consistently 5–7 days after fertilization during both the tillering fertilizer period (TF) and panicle fertilizer period (PF) (Figure 3). In addition, the temporal dynamics of soil N₂O was directly orchestrated by the cyclic wetting and drying phases (Figure 3). During TF, During TF, N₂O peaks post-fertilization were comparable between I_AWD and I_CF. The key divergence emerged during the extended mid-season drainage (water control period), resulting in significantly higher background N₂O concentrations than in I_CF. During PF, the synergy between nitrogen application and the intense drying cycles of I_AWD was striking. I_AWD amplified post-fertilization N₂O peaks by 10–150% across the soil profile compared to I_CF (Figure 3).

Spatially, N₂O concentrations exhibited a pronounced vertical gradient, decreasing progressively with soil depth (Figure 3). The upper 0–20 cm layer served as the dominant reservoir, maintaining concentrations of 0.71–7.90 ppm, significantly higher than the 0.35–4.47 ppm observed in the 20–50 cm layer. Crucially, I_AWD amplified N₂O concentrations across the entire soil profile but with a depth-dependent magnitude (Figure 3 and Table 1). The most pronounced enhancements occurred in the top 30 cm during PF, where I_AWD significantly increased N₂O by 60.00–78.97% relative to I_CF. This effect was most intense in the 0–10 cm layer (up to 78.97% increase), gradually attenuating with depth to a 14–28% increase at 40–50 cm (Table 1). A similar, though less pronounced, pattern was observed during TF. These results identified the upper soil profile (0–20 cm) as the epicenter of I_AWD-induced N₂O production and accumulation. Crucially, the enhancing effect of I_AWD on N₂O concentration was not only depth-dependent but also modulated by the wetting/drying cycle intensity and duration, particularly during TF and PF.

Surface N₂O flux dynamics were fundamentally governed by the interplay between fertilization timing and the I_AWD cycles (Figure 4). While both irrigation regimes exhibited peak fluxes 5–7 days after fertilizer application—a period coinciding with maximum nitrogen availability—the magnitude and temporal pattern of these emissions were critically modulated by the distinct wetting and drying phases of I_AWD (Figure 4). I_AWD consistently generated higher peak fluxes (142.60–283.83 μg m⁻² h⁻¹) than I_CF (127.87–246.39 μg m⁻² h⁻¹), particularly after fertilization during the intense drying cycles of the panicle stage: a larger N₂O reservoir had been pre-formed during the preceding dry period (Figures 3, 4).

This elevated flux activity under I_AWD translated directly into significantly increased cumulative N₂O emissions (Figure 5). The effect was pronounced during both TF and PF—key windows for water and nutrient control. Across two consecutive growing seasons (2019–2020), I_AWD significantly enhanced cumulative N₂O emissions by 23.02–40.13% during TF and 21.17–24.72% during PF (Figure 5). Consequently, the total seasonal N₂O burden under I_AWD was significantly elevated by 20.59% to 28.72% compared to I_CF (Figure 5).

Linear regression analysis was performed to build the relationships between the N₂O concentrations in different soil profiles and soil surface N₂O emissions in I_CF and I_AWD paddy fields (Figure 6). Regardless of irrigation regimes, soil profile N₂O concentrations significantly (P < 0.01 in 0–30 cm soil profile and P < 0.05 in 30–50 cm soil profile) positively correlated with soil surface N₂O fluxes. Partial Least Square-Structural Equation Modelling (PLS-PM) was conducted to further quantify the contribution of soil profile N₂O accumulation to N₂O emissions in I_CF and I_AWD paddy fields (Figure 7). Soil surface N₂O emissions are both affected by direct and indirect effects induced by N₂O concentrations in different soil profiles. For the I_CF paddy fields, soil profile N₂O concentrations totally explained 73.9% of soil surface N₂O emissions. Soil surface N₂O emissions were significantly (P < 0.01) directly affected by N₂O concentration in the 10–20 cm soil profile (0.986) and indirectly affected by N₂O concentration in the 20–30 cm soil profile (0.975) (Figures 7A, B). For the I_AWD paddy fields, soil profile N₂O concentrations totally explained 45.56% of soil surface N₂O emissions. Soil surface N₂O emissions were significantly (P < 0.05) directly affected by N₂O concentration in the 10–20 cm soil profile (0.942) and indirectly affected by N₂O concentration in the 20–30 cm soil profile (0.632) (Figures 7C, D).

Our findings robustly identify the 0–20 cm soil layer as the persistent and dominant reservoir for N₂O accumulation in paddy fields, a phenomenon that transcends the influence of both fertilization events and irrigation regimes (Figure 3). This consistent vertical stratification challenges the conventional framing of N₂O dynamics as a primarily surface-led process and establishes a generalized principle for subsurface N₂O storage in submerged soils. The convergence of this hotspot across management practices suggests that its formation is governed by fundamental soil-biogeochemical constraints rather than water status alone (McCoy et al., 2023; Li et al., 2025). Specifically, the upper soil profile constitutes a critical zone where the abundance of fresh organic substrates from root exudates and senesced biomass, coupled with elevated microbial biomass and activity, creates a prime niche for N₂O generation via nitrification and denitrification (Kim et al., 2022; Song et al., 2024). Furthermore, the physical-hydraulic properties of this layer, including its porosity and gas diffusivity, likely favor the entrapment and dissolved-phase storage of N₂O following its production, rather than immediate emission (Bentzon-Tarp et al., 2023; Gao et al., 2014; Li et al., 2025). While previous studies in aerobic uplands have noted similar subsurface peaks, its confirmation in the unique, redox-fluctuating environment of paddy fields underscores a potentially universal hierarchy in soil profile gas dynamics (Li et al., 2021). A key limitation of this snapshot is its temporal scope; future investigations should employ higher-frequency monitoring to resolve how the stability of this subsurface reservoir responds to rapid redox oscillations. Moreover, integrating isotopic tracing and microbial functional gene analysis across these defined depth intervals will be crucial to definitively partition the relative contributions of nitrification versus denitrification to the observed accumulation (Qin et al., 2023; Yankelzon et al., 2024). Ultimately, explicitly acknowledging the 0–20 cm layer as a pre-eminent N₂O reservoir redefines the target zone for mitigation, shifting the focus from managing surface emissions to controlling subsurface production and storage.

Crucially, our results demonstrate that I_AWD induces a depth-dependent amplification of N₂O within the soil profile, establishing the 0–30 cm soil layer as the epicenter of this enhanced production (Figure 3). This finding reveals that I_AWD’s impact is not uniform but systematically diminishes with depth, a nuance previously obscured by a primary focus on surface flux measurements. The intensified response in the top 30 cm, particularly during the panicle fertilizer stage, points to a critical interaction between management-induced redox dynamics and depth-stratified biogeochemical drivers (Figure 3) (Pittelkow et al., 2013; Liu et al., 2022). We posit that the frequent oxygen fluctuations under I_AWD preferentially stimulate nitrifier and denitrifier activity in the upper horizons, where the confluence of labile carbon from root exudates and fresh fertilizer nitrogen is greatest (Chen et al., 2014; Tang et al., 2025). Furthermore, the physical-hydraulic properties of the soil matrix create a “bio-geochemical filter”: while the saturated conditions during flooding phases may limit gas diffusion and trap N₂O in solution (Islam et al., 2022), the subsequent drying phases enhance gaseous release from upper layers, thereby “resetting” the system for another cycle of production upon rewetting (Verhoeven et al., 2018; Grohs et al., 2024). This cyclic entrapment and pulsed production mechanism is inherently more pronounced in the biologically active topsoil than in the consistently anoxic subsoil, explaining the observed depth gradient. While previous studies have documented the overall stimulatory effect of I_AWD on N₂O, our profile data move beyond this established fact to delineate the precise vertical zonation of its impact (Chen et al., 2022; Liu et al., 2022; Sha et al., 2022). A limitation of this mechanistic interpretation is the challenge in disentangling the in-situ production from the potential vertical transport of N₂O. Future research employing isotopic techniques or in-situ sensors for O₂ and N₂O at a finer spatial-temporal resolution is crucial to partition the contribution of localized production versus physical transport in shaping this depth-dependent response. Ultimately, confirming that I_AWD’s primary influence is confined to the agriculturally manageable plough layer offers a clear target for developing stratified mitigation strategies, such as deep fertilizer placement or the use of subsoiling to physically disrupt the dominant production zone.

Our partial least-squares path model (PLS-PM) pinpoints the 10–20 cm soil layer as the dominant biogeochemical hotspot, exerting the strongest direct control on surface N₂O fluxes in I_AWD paddy fields (Figure 7). This finding moves beyond simply identifying where N₂O accumulates and reveals which specific subsurface zone mechanistically governs its atmospheric release. While previous research has established strong correlations between surface emissions and N₂O concentrations in the upper soil profile (e.g., 0–15 cm) (Li et al., 2021), our path analysis provides causal-weight evidence that the 10–20 cm layer is the pivotal engine room (Figure 7). We posit that this specific depth represents a critical interface where optimal conditions for N₂O production converge with efficient gas transport pathways. During I_AWD drying phases, the 10–20 cm layer transitions from anoxic to sub-oxic states, creating a thermodynamically favorable niche for both nitrification and denitrification, fueled by the diffusion of ammonium from deeper fertilizer bands and labile carbon from the rhizosphere (Kravchenko et al., 2017; Marushchak et al., 2021). Crucially, unlike the often-saturated and diffusion-limited layers below 20 cm, this horizon remains sufficiently connected to the atmosphere during drainage, allowing the produced N₂O to escape rather than be further reduced to N₂ (Wang et al., 2023). This creates a “Goldilocks Zone” for N₂O emission—deep enough to accumulate substantial substrate from the plough layer but shallow enough to facilitate efficient egress (Kravchenko et al., 2017). The stronger explanatory power of our model under I_CF compared to I_AWD suggests that the stability of the saturated environment creates a more predictable, diffusion-dominated system. In contrast, the hydraulic perturbations of I_AWD introduce greater variability in gas transport pathways, albeit with the 10–20 cm layer remaining the unequivocal primary controller (Liu et al., 2022). A key limitation is that our study resolves the “what” and “where” but not the precise microbial kinetics at this interface. Future research integrating in-situ, depth-resolved metatranscriptomics with gas diffusivity measurements is essential to unravel the relative contributions of nitrifier denitrification versus fungal denitrification to the observed dominance of this layer. Confirming the 10–20 cm layer as the command center for emissions provides a powerful new target for mitigation, suggesting that strategies like deep-point placement of enhanced-efficiency fertilizers or the sub-surface application of biochar could directly disrupt this critical nexus, potentially decoupling water savings from climate impact.

The consistent identification of the 0–20 cm layer as the primary N₂O reservoir and the 10–20 cm layer as the dominant source of N₂O emissions in our paddy system is consistent with findings from diverse agroecosystems outside Asia, indicating a widespread principle governed by soil-gas dynamics rather than specific pedoclimatic conditions. In temperate grasslands (Barneze et al., 2024), the lack of a direct correlation between subsurface N₂O concentrations and surface fluxes underscores the critical, and often overlooked, role of in-situ consumption during gaseous diffusion—a process that can decouple production from emission. This aligns with the classical concepts of N₂O entrapment, dissolution, and reduction during transport through the soil matrix, as detailed by Clough et al. (2005). Crucially, our observation that I_AWD most strongly amplifies N₂O in the top 30 cm, coupled with the PLS-PM identification of the 10–20 cm layer as the cardinal controller of surface flux, provides a clear spatial target for these processes. It suggests that under I_AWD, the “active layer” for N₂O production is shallow, but the “effective layer” governing its ultimate release is slightly deeper, where production, temporary entrapment, and partial reduction interact. This dynamic is analogous to mechanisms observed in cold climates, where spring-thaw N₂O bursts are driven more by rapid production in surface layers than by the release of deeply trapped gases (Wagner-Riddle et al., 2008). Similarly, in an Italian rice paddy system, Verhoeven et al. (2018) also found that the N₂O emissions were associated with enhanced nitrification in the upper soil layers during plant establishment. Therefore, our lysimeter-based evidence from Asian paddy fields demonstrates that the hierarchy of subsurface N₂O processes—wherein specific depth zones disproportionately control emissions—could be a transferable concept across contrasting environments. However, considering the current paucity of well-established relationships between soil profile gas concentrations and surface emissions on a global scale, future research ought to be carried out under a broader spectrum of climatic, soil, and crop conditions.