The research took place in the Heilongjiang Mohe Forest Ecosystem Research Station (122°06′–122°27′E, 53°17′–53°30′N) of the Daxing’an Mountains, Northeast China. This area has a typical cold-temperate continental monsoon climate. The annual average temperature is −4.9°C and the annual average is 430–500 mm (Gao et al., 2019). The frost-free period spans approximately 80–90 days from June to August and snow cover is present for more than half a year (Duan et al., 2020). The maximum snow depth was 68.6 cm in the meteorological station (Lin et al., 2018). The primary soil type is brown coniferous forest soil, which is classified as Podzol according to the FAO soil classification.

This area belongs to the northern taiga, with cold temperate deciduous tree species larch (Larix gmelinii L.) as the dominant species. Meanwhile, the Mongolian pine (Pinus sylvestris var. mongolica L.) forest, birch (Betula platyphylla L.) forest, and aspen (Populus davidiana L.) forest are also distributed in this area. The understory species are mainly dominated by Ledum palustre L., Rhododendron dauricum L., Vaccinium vitis-idaea L., and Sphagnum sp. Here, due to different site conditions, the larch forests can be characterized by the same tree layer but different understory vegetations, such as the Sphagnum-Bryum-Ledum palustre-Larix gmelinii forest (SLL), Rhododendron dauricum-Larix gmelinii forest (RL), and Ledum palustre-Larix gmelinii forest (LL). Thus, three typical larch forests, including RL, LL, and SLL, were selected in this study. These three types of larch forest were naturally regenerated from burned land, with a slope < 5°. The basic characteristics of the three larch forests are shown in Table 1. In each larch forest, three 20 m × 30 m plots were placed randomly and a total of nine sample plots were established. In each plot, three 1 m × 1 m subplots (a total of nine independent subplots in each plot) free of trees were randomly selected for GHG and soil sampling.

In the present study, soil GHG fluxes were measured by the static opaque chamber technique. In September 2018, a total of 27 permanent open square-framed base collars (50 cm × 50 cm) with a groove in the upper edge made of polypropylene were inserted into the mineral soil to approximately 5 cm depth. The base collars were set exactly in the middle in each subplot, respectively, with one collar in each subplot. During the study period, the aboveground biomass of shrubs and grasses in each subplot were removed by clipping when setting the base collars to minimize autotrophic respiration. The experiment was conducted during the snow-free period from May to October in 2019. During sampling, an opaque chamber (50 cm × 50 cm × 50 cm) was carefully placed on the base collar to avoid inducing air pressure pumping. Insulation material was fixed outside of the opaque chamber to avoid temperature changes during flux measurements. GHG fluxes were measured twice per month, but we only measured once in June, August, and October due to the weather conditions (rainfall and snowfall). Thus, nine events were recorded in total. Sampling took place between 9:00 and 11:00 a.m., when the soil GHG fluxes are recognized to best represent the daily mean (Alves et al., 2012; Gao et al., 2022).

Gas samples were collected with a 50-ml plastic syringe equipped with a three-way stopcock through the rubber septum in the lid at 0, 15, 30, and 45 min (a total of four samples during each event) after chamber closure. Gas samples were injected into a 100 ml pre-evacuated gas sampling bag (Delin Gas Packing Co., Dalian, China) and transported to the lab for gas analysis. Simultaneously, we measured the air temperature inside the chamber when the gas sample was collected. Gas samples were stored at air temperature and analyzed within 1 week. Gas concentrations in all samples were determined with a gas chromatograph (TRACE 1300 GC, Thermo Fisher Scientific, United States). Fluxes were determined based on linear regression analysis of the change in gas concentration in the chambers with time over a 45-min period for each chamber. The fluxes for further analysis were selected with strong linear relationships (r² > 0.75; Doroski et al., 2019).

Greenhouse gas flux rates for each chamber were calculated using the following formula:

Where F is the flux of the respective greenhouse gas (mg m⁻² h⁻¹ for CO₂, μg m⁻² h⁻¹ for CH₄ and N₂O), dc/dt is the slope of the linear regression for the gas concentration gradient over time (μmol mol⁻¹ h⁻¹); M is the molecular mass of gas (g mol⁻¹); V₀ is the gas molar volume (m³ mol⁻¹); P and P₀ are the atmospheric pressure (Pa) and atmospheric pressure under standard conditions (Pa), respectively; T and T₀ are the mean value of the air temperature inside the chamber during sampling (K) and absolute air temperature (K), respectively; and H is the effective height of chamber (m). Average gas flux and standard deviation were calculated from three replicates for each plot. Positive fluxes represent net soil GHG emissions and negative fluxes represent net soil GHG uptake.

Greenhouse gas global warming potential (GWP) can be used to assess and compare the potential climate impact of the emissions of different GHGs (IPCC, 2007). The comprehensive GWP of soil CO₂, CH₄, and N₂O in each larch forest was calculated as follow:

Where the C CO 2 , C CH 4 , and C N 2 O were the cumulative fluxes of soil CO₂, CH₄, and N₂O, respectively. When the GWP was estimated, the CO₂ is recognized as the reference gas. Thus, the CH₄ and N₂O are converted into “CO₂-equivalents.” Based on 100-year time horizon, the GWP of CH₄ and N₂O are 27.2 and 273 times of CO₂ (IPCC, 2021).

Soil samples were collected from the three types of larch forest at all 27 subplots when collecting the gas sample each time. For each sample subplot, one soil sample was collected at the depth of 0–10 cm using a stainless-steel corer in the vicinity of each chamber. Before soil cores were taken, the litter layer was carefully moved aside and only mineral soil was sampled. Soil samples from the same plot were mixed to form a composite sample. Thus, there were three composite samples per larch forest at each sampling time. The composite samples were stored in plastic bags and all stones, roots, and debris were removed carefully by hand. Then, all the samples were immediately transported in an insulated box to the laboratory for further study. Meanwhile, we also measured the soil temperature at 10 cm soil depth (ST) using a portable thermometer at each subplot (Delta TRAK, CA, United States).

Then the soil samples were separated into two parts after passing through a 2-mm screen. One was stored at 4°C for soil microbial biomass and soil inorganic nitrogen analysis; another part was air-dried at room temperature (20°C) for other analyses of soil properties. In this study, soil water content (SWC) was determined gravimetrically by oven drying the whole soil at 105°C for 24 h. Soil pH was determined using a pH meter equipped with a calibrated combined-glass electrode with the soil-to-water mass ratio of 1:2.5. Soil nitrate-nitrogen (NO₃⁻−N) and ammonium-nitrogen (NH₄⁺ − N) were measured by extraction with 1 mol L⁻¹ KCl and the extraction suspensions analyzed using a flow injection auto-analyzer (Seal Analytical AA3, Norderstedt, Germany). Soil microbial biomass carbon (MBC) and nitrogen (MBN) were measured by the fumigation extraction method (Vance et al., 1987). Meanwhile, in the middle of July, three soil samples were collected in each plot individually. Soil samples at 0–10 cm depth from the same plot were mixed and stored in sterile centrifuge tubes. Then these soil samples were transferred immediately to the laboratory in an insulated box (Esky) on dry ice for DNA extraction.

Microbial DNA from samples was extracted using a PowerSoil® DNA Isolation Kit (MoBio Inc., Carlsbad, United States). DNA concentration and purity were monitored on 1% agarose gel. According to the concentration, DNA was diluted to 1 ng/μl using sterile water. The V3–V4 region of bacterial 16S rRNA genes was amplified using the specific primer (primer: 338F/806R: 5′-ACTCCTACGGGAGGCAGCAG-3′/5′-GGACTACHVGGGTWTCTAAT-3′) with the barcode. All PCR reactions were carried out in 30 μl reactions with 15 μl of Phusion®High-Fidelity PCR Master Mix (New England Biolabs); 0.2 μM of forward and reverse primers, and about 10 ng template DNA. Thermal cycling consisted of initial denaturation at 95°C for 5 min, followed by 27 cycles of denaturation at 94°C for 30 s, and at 72°C for 45 s. Finally 72°C for 10 min. The PCR products were purified with GeneJET Gel Extraction Kit (Thermo Fisher Scientific, United States) sequenced on the Illumina Hiseq 2500 platform (Magigene Co., Ltd., Guangzhou, China). Sequence analysis was performed using the QIIME 1.6.0 pipeline software. Sequences were assigned to operational taxonomic units (OTUs) with 97% similarity using Uparse software. The original sequence retrieved in this study has been deposited with the National Center for Biotechnology Information (NCBI) under the accession number PRJNA797560.

One-way ANOVA followed by the LSD test was performed to test the differences in soil microenvironmental variables, properties, and soil GHG fluxes among three types of larch forest at p < 0.05 level. Two-way ANOVA was used to examine the effects of forest type and seasons, and their interactions on soil GHG fluxes and soil variables. Principal coordinates analysis (PCoA) based on Bray–Curtis was used to visualize soil bacterial community similarity among different types of larch forest. Pearson correlation analysis was used to explore the relationship between soil GHG fluxes and measured factors. Variation partitioning analysis (VPA) was used to identify the contribution of soil properties (soil NH₄⁺-N, NO₃⁻-N, MBC, and MBN), soil microenvironmental variables (soil temperature and soil water content), and soil bacterial community to soil GHG fluxes. The VPA was conducted in R software (Version 4.0.3) using the vegan package. Based on the VPA analysis, structural equation modeling (SEM) was used to explore whether the major pathways of soil factors and bacterial community affected soil GHG fluxes among three types of larch forest. The best fit model was assessed by fit indices, including non-significant paths (p > 0.05) and χ² test (0 ≤ χ²/df ≤ 2), and standardized root mean square residual (SRMR < 0.05). The bacterial community was represented using the scores of the first dimensions in the PCA plot in SEM. The SEM analysis was conducted using AMOS 20.0 software (SPSS Inc., Chicago, IL, United States). The statistical analyses were performed in SPSS (19.0 for Windows, SPSS Institute, Inc., Chicago, IL, United States) and R software (Version 4.0.3). The figures were drawn with OriginPro 2016 (OriginLab Corp., Northampton, MA, United States).

Mean soil temperature was highest in LL (6.89°C), then followed by RL (5.36°C), and lowest in SLL (4.17°C; p < 0.01). The highest mean soil water content was found in SLL (53.18%) and LL (47.35%) which were significantly higher than in RL (40.71%; p < 0.01). However, no significant differences in soil water content were found between LL and SLL (p > 0.05). Significant differences in soil properties were found among the three types of larch forest (Figure 1; p < 0.05). The soil NH₄⁺-N content was highest in SLL (44.29 mg kg⁻¹), significantly higher than in LL (37.27 mg kg⁻¹) and RL (34.07 mg kg⁻¹; p < 0.05), whereas there were no significant differences between the last two forest types (p > 0.05). Soil NO₃⁻-N content in SLL was significantly higher than that in RL (p < 0.05), while no significant difference was found between LL and SLL or RL (p > 0.05). Both soil MBC and MBN contents ranked in the order of LL > RL > SLL (p < 0.01), whereas no significant difference in soil MBN content was found between LL and RL (p > 0.05).

Soil GHG fluxes were significantly different among three types of larch forest (Figure 2). The soil CO₂ fluxes in LL (347.12 mg m⁻² h⁻¹) and RL (333.54 mg m⁻² h⁻¹) were significantly higher than that in SLL (295.58 mg m⁻² h⁻¹; p < 0.05). However, no significant differences in soil CO₂ fluxes were found between LL and RL (p > 0.05). The CH₄ flux in SLL was-21.07 μg m⁻² h⁻¹ which was significantly higher than in LL (−35.85 μg m⁻² h⁻¹) and RL (−35.21 μg m⁻² h⁻¹; p < 0.01) and no significant difference was found between the last two types of larch forest (p > 0.05). Similarly, the soil N₂O fluxes in LL and RL were 20.71 and 20.73 μg m⁻² h⁻¹, respectively, significantly higher than that in SLL (17.65 μg m⁻² h⁻¹; p < 0.05), whereas there were no significant differences between RL and LL (p > 0.05).

During the study period, there were significant differences in the cumulative soil GHG fluxes and their GWP among the three types of larch forest, which were consistent with the differences in average soil GHG flux (Table 2; Figure 2). The comprehensive GWP of soil GHG fluxes in LL (13.59 Mg CO₂-Eq ha⁻¹) and RL (12.89 Mg CO₂-Eq ha⁻¹) were significantly higher than in SLL (11.42 Mg CO₂-Eq ha⁻¹; p < 0.05), while no significant difference was found between LL and RL (p > 0.05). The contribution of CO₂, CH₄, and N₂O to GWP was significantly different. The comprehensive GWP among three types of larch forest was mainly determined by the cumulative soil CO₂ fluxes. The GWP of cumulative soil CH₄ was negative among the three types of larch forest.

Soil bacterial community composition was clearly clustered based on the types of larch forest (Figure 3). Differences in understory species composition caused some changes in bacterial community diversity. PC1 and PC2, the first two principal components, explained 65.91 and 17.21% of the total bacterial variability, respectively. The alpha diversity indices described the bacterial community richness (the Chao 1 index, based on abundance) and diversity (the Shannon and Simpson indices; Table 3). The Chao 1 index showed no significant differences among three types of larch forest (p > 0.05). However, there was a significantly higher Shannon index in LL than that in RL and SLL (p < 0.05). The Simpson indices in LL and SLL were lower than in RL (p < 0.05).

Further, there were similar bacterial community structures at phylum level, but their relative abundance was different among three types of larch forest (Figure 4A). The dominant bacterial phyla were Proteobacteria, Actinobacteria, Acidobacteria, and Chloroflexi among all types of larch forest (Figure 4A). However, there were significant differences in their abundance (Figure 4B). The relative abundance of Proteobacteria in RL and LL reached 25.35–38.48 and 24.97–30.17%, significantly higher than that in SLL (16.25–22.81%; p < 0.05). Meanwhile, the relative abundance of Actinobacteria in RL (18.51–28.73%) was significantly higher than that in LL and SLL (p < 0.05) whereas no significant difference was found between the late two forest types (p > 0.05). However, the relative abundance of Acidobacteria in SLL (22.13–26.72%) and LL (20.63–27.67%) was significant higher than in RL (p < 0.05). Moreover, the highest Chloroflexi abundance was found in SLL (29.34–33.15%), but no significant differences were found between LL and RL (p > 0.05). There were no significant differences in Verrucomicrobia abundance among three types of larch forest (p > 0.05). However, the highest Gemmatimonadetes abundance was found in RL, and the lowest Bacteroidetes abundance in SLL (Figure 4B; p < 0.01).

Variations in soil GHG fluxes over the three types of larch forest were driven by soil environmental factors, soil properties, soil bacterial community, and their interactions (Figure 5). Variation partitioning analysis showed the interactions of soil environmental factors, soil properties, and soil bacterial composition contributed 47.18, 66.50, and 52.51% of variations in soil CO₂, CH₄, and N₂O fluxes among three types of larch forest (Figure 5). Meanwhile, the relative importance of soil factors differed in regulating different soil GHG fluxes. Soil environmental factors, soil properties, and soil bacterial community played important roles in regulating soil CO₂, CH₄, and N₂O fluxes alone, respectively.

Structural equation models (SEM) further indicated the key pathways through which understory changes regulated soil GHG flux variations among three types of larch forest (Figure 6). SEM explained 87, 98, and 80% of the variation in soil CO₂, CH₄, and N₂O fluxes, respectively. Soil bacterial community and soil temperature were the main drivers in regulating directly CO₂, CH₄, and N₂O flux variations among three types of larch forest. Meanwhile, soil inorganic nitrogen content (NH₄⁺-N and NO₃⁻-N) played a vital role in regulating soil GHG fluxes. Specifically, soil NO₃⁻-N content affect soil CO₂ fluxes indirectly by regulating soil bacterial community (Figure 6A). Soil NH₄⁺-N also affected soil CH₄ fluxes via its effect on soil bacterial community (Figure 6B). Both soil NH₄⁺-N and NO₃⁻-N content had significant indirect effects on soil N₂O fluxes variations via their effects on soil bacterial community (Figure 6C).

As we expected, differences in understory species significantly altered soil CO₂ fluxes among three types of larch forest. This point is consistent with previous studies (Zhang et al., 2014; Liu et al., 2019), suggesting that changes in understory shrubs biomass correlate to the spatial variation of soil respiration and understory replacement enhances soil GHG emission. Variations in soil CO₂ fluxes with understory species were directly linked to soil bacterial composition and soil temperature (Figure 6A). Proteobacteria, Bacteroidetes, Acidobacteria, Chloroflexi, and Actinobacteria at the phylum level are believed to play major roles in the degradation of soil organic matter and carbon cycling (Xiao et al., 2017; Jia et al., 2018; Zhang et al., 2022). Proteobacteria and Bacteroidetes are believed to have positive correlations with soil organic matter mineralization (Fierer et al., 2007). Thus, the significantly higher abundance of Proteobacteria and Bacteroidetes led to higher soil CO₂ emission in LL and RL than that in SLL in our study (Figure 4B). Acidobacteria and Actinobacteria have ability to degrade recalcitrant organic carbon (Trivedi et al., 2013). Although there were significant differences in both Acidobacteria and Actinobacteria abundance (Figure 4B), no significant difference was found in total Acidobacteria + Actinobacteria abundance among three types of larch forest. This means that differences in soil CO₂ fluxes among the three types of larch forest may not come from the degradation of recalcitrant organic matter. Meanwhile, Chloroflexi has been intensified to perform a unique C fixation metabolism (Shih et al., 2017; He et al., 2021). Thus, the higher Chloroflexi abundance in SLL indicated the higher C fixation rather than CO₂ emission in SLL than that in LL and RL (Figure 4B), which is another reason resulting in the lower soil CO₂ emission in SLL. Partly agreeing with our hypothesis two, there were no significant differences in soil CO₂ fluxes between RL and LL, which can also be explained by their similar bacterial abundance. Apart from soil bacterial communities, soil fungal communities also have potential effects on CO₂ fluxes (Lee et al., 2022). We could not determine fungal community composition in our study, but our measurements of microbial biomass C and N include also this biota. The higher MBC content did not cause higher soil CO₂ fluxes, which may be due to the differences in soil fungal communities (Liu et al., 2021). Meanwhile, compared to bacteria, soil fungal communities are sensitive to litter composition (Habtewold et al., 2020). Thus, the analysis of fungal communities among three types of larch forest remains an important area for future research.

Moreover, soil temperature and NO₃⁻−N content also had significant effects on soil CO₂ fluxes directly and indirectly, respectively. High soil temperature can accelerate the turnover rates of soil bacterial communities (Wu et al., 2015). Thus, higher soil temperature in LL and RL can lead to higher soil CO₂ fluxes than SLL. There was higher soil temperature in LL than in RL, but the higher soil water content in LL may offset the effects of soil temperature on soil CO₂ emission and ultimately induce similar soil CO₂ fluxes between them (Wu et al., 2019). It has been well suggested that soil NO₃⁻−N content is closely related to soil respiration (Chen et al., 2021) because NO₃⁻−N can inhibit the activity of the enzymes in the soil (Li et al., 2015). Meanwhile, boreal forests are nitrogen limited ecosystems, and the availability of this nutrient is essential to organic matter decomposition and to soil bacterial metabolism (Nordborg et al., 2007; Sponseller et al., 2016). Thus, the higher soil NO₃⁻−N content is another potential reason explaining the lower soil CO₂ fluxes in SLL than that in RL.

The forest soils exhibited net CH₄ uptake with the mean CH₄ fluxes ranging from −35.85 to −21.07 μg m⁻² h⁻¹ among three types of larch forest. This result is consistent with previous studies which suggest the larch forest soils show a net sink of atmospheric CH₄ in this study area (Wu et al., 2019; Duan et al., 2022). However, the capacity of soil net CH₄ uptake showed significant differences in RL, LL, and SLL. Particularly, the soil CH₄ fluxes in SLL were significantly higher than that in RL and LL, whereas no significant differences were found between LL and RL (Figure 2), which partly agrees with our hypothesis 2. In the present study, differences in understory species affected soil CH₄ fluxes by regulating directly soil bacterial community and soil temperature among three types of larch forest (Figure 6B). Soil CH₄ fluxes are the balance of production and consumption mainly mediated by methanogenesis and methanotrophic microorganisms, respectively (Oertel et al., 2016). Methanotrophic bacteria are taxonomically affiliated with the Proteobacteria and Verrucomicrobia phyla while methanogens belong to the Euryarchaeota phylum (Gagliano et al., 2014; Liu X. et al., 2017; Andressa, 2019). In the present study, there were higher relative abundances of Proteobacteria in RL and LL than that in SLL (Figure 4B) and significantly negative correlations between Proteobacteria and soil CH₄ fluxes were also found (Supplementary Table S1). The higher Proteobacteria can lead to more CH₄ oxidized in soil and thereby lead to lower CH₄ fluxes in LL and RL (Figure 4B). Soil CH₄ is produced under strictly anaerobic conditions. Although there was no information about Euryarchaeota in our study, the significantly higher soil water content (Figure 1) and bulk density (Xiao et al., 2020a) in SLL can lead to better anaerobic conditions. Thus, we can conclude that there will be more CH₄ production in SLL than in LL and RL. This is another reason resulting in higher soil CH₄ fluxes in SLL.

In addition to bacterial composition, soil temperature can also regulate the activity of methanotrophic microorganisms to control CH₄ fluxes (Song et al., 2021). However, the effect of soil temperature on soil CH₄ fluxes is also modified by soil water content (Oertel et al., 2016). Coinciding with previous studies (Chen et al., 2021; Wang et al., 2022), our study showed a significantly positive relationship between soil water content and soil CH₄ fluxes (Supplementary Table S1). High soil water content in SLL can not only enhance anaerobic soil conditions which prefer soil CH₄ production but also inhibit gas diffusion and ultimately lead to higher soil CH₄ fluxes in SLL, compared with RL and LL. Meanwhile, soil temperature in LL was significantly higher than in RL, but the higher soil water content in LL led to similar soil CH₄ fluxes between RL and LL. Moreover, soil NH₄⁺ − N content is another vital factor affecting soil CH₄ microbial processes in permafrost (Song et al., 2021). High NH₄⁺ − N content can not only contribute to soil CH₄ production by enhancing the abundance of methanogens but also can inhibit CH₄ oxidation by competing for methane monooxygenase (Bédard and Knowles, 1989; Kuzyakov and Xu, 2013). Our study also documented the significant positive correlation between soil CH₄ fluxes and soil NH₄⁺ − N content, which is in line with previous studies (Song et al., 2021). Therefore, higher soil CH₄ fluxes were found in SLL where there was higher soil NH₄⁺ − N content. Further, it is worth noting that Sphagnum is a key species associated with methane oxidation (Kox et al., 2020). Thus, the Sphagnum in the SLL may lead to higher CH₄ oxidation. However, the lowest soil CH₄ oxidation in SLL may be due to the lower soil temperature and aboveground plant removal. The lower soil temperature can inhibit Sphagnum metabolism. It must also be remembered that the aboveground biomass of the vegetation was cut inside each collar. This treatment was necessary to avoid the effects of interference from vegetation on soil GHG fluxes, but it may have induced lower soil CH₄ oxidation in SLL.

N₂O is produced mainly by incomplete nitrification and denitrification processes mediated by microbial metabolism (Hu et al., 2015; Oertel et al., 2016). Our study also showed that soil NH₄⁺ − N and NO₃⁻−N content, as the important substrates of nitrification and denitrification, both can significantly regulate soil N₂O fluxes by affecting soil bacterial community (Figure 6C). Proteobacteria at the phylum level are recognized as an important factor in affecting soil N₂O production from nitrification processes by regulating ammonium oxidation and nitrite accumulation (Yang et al., 2017). Ammonia oxidation is the rate-limiting step for the whole nitrification process, which is critical for N₂O production from nitrification (Hu et al., 2015; Xue et al., 2021). Thus, the higher Proteobacteria can lead to higher soil N₂O fluxes in RL and LL than that in SLL in the present study (Figure 4B). Meanwhile, nitrite accumulation can also induce higher N₂O production by both nitrite reduction (NO₂⁻ to NO) and further NO reduction (NO to N₂O) in nitrifier denitrification pathways (Wrage et al., 2001; Supplementary Figure S2A), which is another reason resulting in higher N₂O production in RL and LL.

Apart from nitrification, heterotrophic denitrification is also a vital pathway of soil N₂O production which is conducted by denitrification microbes under oxygen-limited conditions (Oertel et al., 2016). However, given the soil net CH₄ uptake and low soil NO₃⁻−N content and nitrification rates (Xiao et al., 2022) among all types of larch forest, nitrite accumulation may be mainly from the ammonia oxidation pathway (Supplementary Figure S2A). This means that soil N₂O production in our study may be mainly from nitrification-related pathways (ammonia oxidation and nitrifier denitrification). In both pathways, N₂O production from ammonia oxidation usually occurs in high ammonia contents, low nitrite concentrations, and high nitrification rates (Wunderlin et al., 2012). In our study, there were high NH₄⁺ − N contents, but the soil nitrification rates are low among the three types of larch forest (Xiao et al., 2022). Meanwhile, the higher abundance of ammonia-oxidizing bacteria (Nitrosomonadaceae) than nitrite oxidation bacteria (Nitrospire and Nitrobacter) also showed lower nitrification rates and higher nitrite accumulation (Supplementary Figure S2B). On the contrary, nitrifier denitrification can be a major pathway of N₂O production under low pH conditions (Wrage et al., 2001). Thus, variations in soil N₂O fluxes among three types of larch forest may be mainly from the nitrifier denitrification process in our study. Although there was the same pathway from NO₂⁻ to N₂O in both heterotrophic denitrification and nitrifier denitrification processes (Supplementary Figure S2B), the whole nitrifier denitrification processes are solely regulated by ammonia-oxidizing bacteria which may not need a strictly anaerobic process (Beaumont et al., 2004; Shaw et al., 2006). However, previous studies also speculate that denitrification is the dominant process of soil N₂O production in our study area due to the high relationship between NO₃⁻−N content and soil N₂O fluxes (Gao et al., 2019; Wu et al., 2019). Thus, further studies are needed to confirm the soil N₂O production mechanism from the gene level.