In this study, 25 soil samples were collected in April 2021 according to different land use types in Linzhi, Tibet (Grassland G1–G4, Cropland C1–C8, Forest F1–F8, Wetland W1– W3, and Bare land BL1–BL2, Fig. S1 in the Supplement). The sampling site is located in the southeastern part of the Tibetan Plateau (26°52^′ N–30°40^′ N, 92°09^′ E–98°47^′ E), where a variety of climatic types coexist, ranging from tropical to frigid. The altitude range was 150–7782 m, with an average elevation of ∼ 3100 m. The annual average temperature, precipitation and sunshine duration are 8.7 °C, 650 mm and 2022.2 h, respectively, and the frost-free period was 180 d (Linzhi Municipal People's Government, 2018). At each sampling site, a 10 m × 10 m sample square was delineated, and three sub-samples of topsoil (0–5 cm) were collected along the diagonal, subsequently homogenized to form a composite sample (total weight ∼ 1.5 kg). Subsequently, using a sterile sampling spoon, 1–2 g of soils were collected from the composite sample and placed into 2 mL sterilized centrifuge tubes for subsequent microbiological analysis. The soil samples were transported back to the laboratory within 48 h via SF Express's standard cold-chain service (maintained at a constant temperature of 4 °C throughout the entire process). Upon arrival at the laboratory, soil samples for microbiological analyses were stored at -80 °C, while soil samples for determining Nr gas emissions and physicochemical properties were stored at -20 °C.

Soil samples were passed through a 2 mm analytical sieve and used for the determination of physicochemical properties such as soil water content, pH, particle size and inorganic nitrogen. Soil water content (SWC) was calculated by drying at 105 °C for 24 h and using the change in soil mass before and after drying. Soil pH was determined by a pH meter (FE28, Mettler Toledo, Zurich, Switzerland) in a soil-water suspension (1:2.5, w/v). Particle size fractions were determined by a laser diffraction particle size analyzer (LA-960A, HORIBA, Kyoto, Japan) (Song et al., 2020). Soil inorganic nitrogen (NO3-, NO2-, NH4+) content was determined by mixing soil with deionized water (DIW) at a ratio of 1:5 (w/v) (Song et al., 2024), shaking and centrifuging, and then the extracted supernatant was analyzed by a continuous flow analyzer (Skalar San++ System, Skalar, Breda, Netherlands) (GB/T 42487-2023). At the same time, another identical sample was incubated in a thermostatic incubator at 25 °C for 7 d, and the soil inorganic N content was determined by the method described above, thus calculating the net mineralization rate and net nitrification rate (Lv et al., 2023). For the determination of soil dissolved organic carbon (DOC), 5 g of fresh soil was mixed with 25 mL of 0.5 molL-1 K2SO4 solution, shaken for 1 h at room temperature, centrifuged at 4000 rmin-1 for 5 min through a centrifuge (H1850, Xiangyi, Xiangtan, China). Then the supernatant was extracted by filtration with 0.45 µm aqueous filter membrane, and was analyzed by a total organic carbon analyzer (TOC-L CPN, Shimadzu, Kyoto, Japan). Air-dried soil samples passing a 0.15 mm analytical sieve were measured with an elemental carbon and nitrogen analyzer (Elementary Vario MAX CN, Frankfurt, Germany) to determine the total carbon (TC) and nitrogen (TN) content.

Dynamic chamber systems were typically used to study the release of Nr (HONO, NO, NO2, and NH3) from soils (Behrendt et al., 2014). Here, soil Nr gas fluxes were measured using a MUlti-gaS dynamIC measurement system (MUSIC, Fig. S2 in the Supplement). This system simulates soil Nr gas emissions during the wetting-drying cycle and the results are in good agreement with field fluxes and have been applied in several studies (Song et al., 2023; Wu et al., 2020, 2022a, 2019). Specifically, 40 g of fresh soil samples were added to petri dishes (inner diameter = 94 mm) and moistened with purified water until the maximum soil water holding capacity (WHC) was reached, which was measured by filtration (Whatman filter paper no. 42) (Behrendt et al., 2014). Subsequently, the petri dish containing soil was placed in a Teflon chamber (∼ 10 L) in the dark and at a constant temperature (25 ± 0.5 °C) and continuously flushed with dry purified air (free of HONO, NO, NO2, NH3, O3, CxHy, H2O, etc.) at a constant flow rate of 6 Lmin-1 until the soil in the petri dish was completely dry. The system was set up with three Teflon chambers, and one of the chambers without soil samples was set up as the reference chamber for calculating the flux. Each chamber was measured for 2 min and then switched to the next chamber via a multipoint switching valve, resulting in a temporal resolution of 6 min per chamber. During the drying process, HONO and NO2, NO, and NH3 concentrations in the chamber headspace were measured continuously by incoherent broadband cavity-enhanced absorption spectroscopy (IBBCEAS, AIOFM-CEAS-TY19-1), chemiluminescent NOx analyzer (Thermo 42i-TL, Waltham, MA, USA), and NH3 analyzer (Picarro G2103, USA), with instrumental time resolutions of about 10, 10, and 1 s, respectively. The lower limits of detection (LOD) for instrumental measurements of HONO, NO2, NO, and NH3 were approximately 150, 200, 100, and 100 ppt, respectively (Song et al., 2023). In addition to the concentration-based instrument LODs, we have now calculated the minimum detectable fluxes for each target gas. Specifically, the minimum detectable fluxes were approximately 0.8 ngNm-2s-1 for NO, 1.6 ngNm-2s-1 for NO2, 0.8 ngNm-2s-1 for NH3, and 1.2 ngNm-2s-1 for HONO. To avoid uncertainties caused by switching chambers, only the respective mean value for the 30 s before chamber switching was calculated for each measurement. Furthermore, an infrared CO2/H2O analyzer (LI-COR 840A, Lincoln, NE, USA) was used to record soil desiccation and calculate the soil water content (SWC) of the samples.

The fluxes of soil Nr gases (HONO, NO, NO2, and NH3) were calculated as follows: 1FN=QA×1Vm×(xout(t)-xin) where FN denotes the flux of Nr gases (ngNm-2s-1); Q denotes the flow rate of the gases (m3s-1); A denotes the surface area of the soil (m2); Vm denotes the molar volume of air (m3mol-1); t represents the 30 s before chamber switching; and χout and χin denote the headspace mixing ratios at the outlet and inlet of the chamber (ppb). The soil sample preservation protocol used in this study (storage at -20 °C followed by thawing) may have amplified the potential for soil Nr emissions; therefore, we adopted the optimal flux (i.e., the highest value) under specific stimulating conditions, which represents the maximum potential emissions at the maximum WHC.

Meanwhile, based on the above flux data, the integrated soil Nr gases (HONO, NO, NO2, and NH3) emission during a single wet and dry alternation were calculated as follows: 2EN=∫t=0t=max⁡F(t)×dt106 where EN denotes the integrated soil Nr gas emission (mgm-2) and F(t) denotes the soil Nr gas emission flux at moment t. The fluxes were assumed constant over the switch intervals.

Three typical soil types, forest (F3), grassland (G5) and cropland (C2), were selected for the fumigation experiment to simulate the effects of atmospheric NOx and O3 concentrations on soil Nr gas emissions. Four fumigation treatments were installed, i.e., 10 ppb NO, 50 ppb NO, 10 ppb NOx + 30 ppb O3, and 50 ppb NOx + 100 ppb O3. The first two treatments (solely NO) were designed to isolate the direct effect of NO gas on soil nitrogen cycling, independent of its rapid oxidation process, the other two treatments (NOx + O3) were designed to simulate the synergistic effects after photochemical aging, where NO is rapidly converted to NO2 in the presence of O3, thereby assessing the combined impacts of NO, NO2, and strong oxidants on the soil system. Although NO rapidly reacts with O3, and O3 is concurrently consumed by reactions with soil organic matter, the physicochemical kinetic conditions governing gas conversion and consumption are consistent across all treatment groups under identical flow rates and fumigation regimes. Thus, the relative ranking of actual exposure intensities among the different treatment groups is still considered valid. Three replicates were provided for each treatment, with a total of 36 samples. Specifically, ∼ 100 g of fresh soil samples were weighed and placed into 500 mL brown incubation flasks (Bottom diameter: 7.5 cm; Bottle height: 17.5 cm; Soil thickness: 3–5 cm). The NO standard gas (5.2 ppm, Air Liquide, Shanghai) was subsequently diluted to a pre-set NO/NOx+O3 mixing ratio concentration by the Sonimix ozone gas master standard (SONIMIX 4001, LNI, Italy) and piped into an incubation bottle containing soil (flow rate ∼ 3 Lmin-1) (Fig. S3 in the Supplement). The lid of the incubation bottle was equipped with two connecting pipes and loaded with valves that enable controlled on/off switching. The second pipe of the incubation bottle was connected to the NOx analyzer to allow real-time observation of changes in NOx concentration in the bottle. When the fumigant gas in the bottle reaches the pre-set concentration, promptly close the valve of the pipeline connected to the instrument. The whole process will take about 5 min. Then, the soil was fumigated for another 10–20 s before closing the valve of the fumigant gas input pipe. The tube junction was sealed with a sealing film and then placed in a thermostatic incubator (25 ± 0.5 °C). Based on the results from preliminary tests, we selected a fumigation period of 5 d (Table S1 in the Supplement), during which the fumigant gas input was repeated every 24 h to ensure a stable concentration of fumigant gas in the incubation bottles. After fumigation incubation was completed, ∼ 2 g of soil samples were collected within 3 min into 2 mL sterilized centrifuge tubes, and then placed in a refrigerator at -80 °C for storage for subsequent soil microbial analyses. In addition, ∼ 40 g of soil samples were removed from the incubation bottles for the determination of Nr flux. The remaining samples were used for the determination of soil pH, inorganic nitrogen content, and DOC content.

DNA was extracted from 25 natural soil samples using a DNA extraction kit (MOBIO Laboratories, Carlsbad, CA, USA). The concentration and quality of DNA were determined by NanoDrop2000 (Thermo Fisher Scientific, Waltham, Massachusetts USA) and 1 % agarose gel electrophoresis. The primer pairs 338F (5^′-ACTCCTACGGGAGGCAGCAG-3^′) and 806R (5^′-GGACTACHVGGGTWTCTAAT-3^′), 524F10extF (5^′-TGYCAGCCGCCGCGGTAA-3^′) and Arch958RmodR (5^′-YCCGGCGTTGAVTCCAATT-3^′), ITS1F (5^′-CTTGGTCATTTAGAGGAAGTAA-3^′) and ITS2R (5^′-GCTGCGTTCTTCATCGATGC-3^′) for bacterial V3-V4 region, archaeal V4-V5 region, and fungal ITS1-ITS2 region by PCR amplification (GeneAmp^®9700 PCR Thermal Cycler, ABI, CA, USA), respectively. The volume of the PCR reaction system was 20 µL, see Table S2 in the Supplement for specific reagents and volumes. The PCR amplification cycling conditions were: pre-denaturation at 95 °C for 3 min, followed by denaturation at 95 °C for 30 s, annealing at 55 °C for 30 s, extension at 72 °C for 45 s for a total of 35 cycles, and finally extension at 72 °C for 10 min until 10 °C. In 3 replicates per sample, PCR products from the same sample were mixed and detected by 2 % agarose gel electrophoresis and purified using the AxyPrep DNA Gel Extraction Kit (Axygen Biosciences, Union City, CA, USA) according to the manufacturer's instructions, and Quantus™ fluorometer (Promega, USA) was used to quantify. Library construction of purified PCR products was performed using the NEXTFLEX^® Rapid DNA-Seq Kit. Finally, the library was subjected to paired-end sequencing on the Illumina MiSeq PE300 platform (Meiji Biomedical Technology Co., Ltd., Shanghai, China). Raw sequencing data were uploaded to the Scientific Data Centre of Beijing Institute of Genomics, Chinese Academy of Sciences (https://ngdc.cncb.ac.cn/gsub/submit/gsa/subCRA019428/finishedOverview, last access: 22 August 2023) under the accession number CRA012324.

The raw sequencing sequence of the double-end was quality controlled using Fastp software (V0.19.6, https://github.com/OpenGene/fastp, last access: 14 July 2022) and spliced using FLASH software (V1.2.11, https://ccb.jhu.edu/software/FLASH/index.shtml, last access: 14 July 2022). Using UPARSE software (V11, http://www.drive5.com/uparse/, last access: 14 July 2022), the quality-controlled spliced sequences were clustered by operational taxonomic unit (OTU) and chimeras were excluded based on 97 % similarity. The RDP classifier (V2.13, https://sourceforge.net/projects/rdp-classifier/, last access: 14 July 2022) was used to compare the Silva 16S rRNA gene database (V138, https://www.arb-silva.de/, last access: 14 July 2022) as well as the Unite fungal ITS database (V8, https://unite.ut.ee/, last access: 14 July 2022) for OTU species taxonomic annotation with a confidence threshold of 70 %, and the community composition of each sample was counted at different species classification levels.

Natural soil DNA was extracted from 0.25 g of frozen soil using the MoBio PowerSoil DNA Kit (MOBIO Laboratories) and fumigated soil DNA was extracted from 0.25 g of frozen soil using the DNeasy^® PowerSoil^® Pro Kit (QIAGEN, USA). Quantitative real-time PCR (qPCR) (7300 Real-Time PCR System, Applied Biosystems, USA) was performed on natural soil DNA samples using SYBR fluorescent dye quantitative PCR to determine the copies of amoA, narG, nirK, nosZ and napA genes. The qPCR reaction mixture for natural soil consisted of 2 µL Template DNA, 0.8 µL forward and reverse primers, 10 µL 2×ChamQ SYBR Color qPCR Master Mix (Vazyme), 0.4 µL 50×ROX Reference Dye 1 and 6 µL ddH2O. Fumigated soil DNA samples were also used by SYBR fluorescent dye method for qPCR (7500 Real-Time PCR System, Applied Biosystems, USA) to determine the copies of amoA_AOA, amoA_AOB, comammox Clade A, nirK, nirS, nxrA, nxrB, napA, and nosZ genes. The qPCR reaction mixture for fumigated soil consisted of 1 µL Template DNA, 0.4 µL forward and reverse primers (0.8 µL for comammox Clade A and napA genes), 10 µL 2×TB Green Premix Ex Taq (Takara), 0.4 µL 50×ROX Reference Dye II and 7.8 µL ddH2O (7 µL for comammox Clade A and napA genes). Primer sets and qPCR conditions for each gene were listed in Table S3 in the Supplement. A qPCR standard curve based on threshold cycling (CT) values was constructed using a 10-fold dilution of the linearized plasmid. The gene copies of each soil sample were then determined by correlating the CT value of each sample with the known copies in the standard curve using a known linear logarithmic relationship. Plasmid copies conversion formula (copies µL-1) = plasmid concentration (ngµL-1) ×10-9×6.02×1023/(molecular mass×660). The amplification efficiencies of all qPCRs in this study were in the range of 80 %–110 % and R2 > 0.99 (Table S3).

We compiled a dataset of soil Nr fluxes in Tibet, including measured data in this study and published data in literature (Table S4 in the Supplement). Stepwise regression analysis was performed to establish quantitative relationships between soil Nr fluxes and key soil properties, including soil pH, inorganic nitrogen contents, and the abundance of nitrogen-cycling functional genes. Based on these relationships and the flux dataset, we estimated the annual soil emissions of HONO, NO, and NOx in Tibet using parameterized soil Nr emissions model (Wu et al., 2022b). This model takes into account factors such as soil temperature, precipitation, and nitrogen fertilizer application (on farmland). It should be clarified that the current annual emission model does not directly incorporate the effects observed in the fumigation experiments. The upscaling model relies on statistical relationships established between soil Nr fluxes and key in-situ soil properties. While these long-term in-situ properties inherently reflect historical deposition, capturing the short-term chemical feedback of ambient NOx/O3 into a regional-scale model is complex. Thus, these effects are qualitatively discussed in relation to the fumigation results (Figs. S5–S7 in the Supplement).

We quantified the contribution of soil HONO emissions to NOx-independent (external) atmospheric HONO sources by integrating the external source framework of Wang et al. (2023a) with the soil-atmosphere exchange model of Meusel et al. (2018), based on field data collected at the Nam Co Multisphere Observation and Research Station (30°46.44^′ N, 90°59.31^′ E, 4730 ma.s.l.) (Wang et al., 2023a). Here, community-weighted maximum HONO flux (Fcomm,max⁡) was calculated by weighting land-use-specific maximum HONO fluxes by their areal coverage (bare soil: 12 %; grassland: 74 %; cropland: 0.1 %; forest: 9 %; wetland: 5.6 %) (Yang and Huang, 2025): 3Fcomm,max⁡=∑i=1nFmax⁡,i×pi100 where Fmax⁡,i = maximum HONO flux for land-use type i (ngNm-2s-1); pi = areal percentage of land-use type i. Land-use-specific Fmax⁡,i values were: bare soil (21.6 ± 8.4); grassland (18.3 ± 6.2); cropland (42.1 ± 28.1); forest (15.7 ± 5.3); wetland (9.8 ± 3.7) ngNm-2s-1.

Soil HONO molecular flux normalized to boundary layer volume (Sground,cm-3) was calculated as: 4Sground,cm-3=Fcomm,max⁡×aBLH×106 where a = unit conversion factor; 10⁶ = scaling to cm3; BLH (boundary layer height) of 182 m was used derived from ERA5 reanalysis data for Nam Co station. A constant BLH is a simplification. The 182 m daily-average from ERA5 for Nam Co was adopted to focus on daytime oxidation capacity driven by HONO photolysis.

The volume mixing ratio production rate Smix (pptvh-1) representing soil HONO contributions to atmospheric external sources was derived as: 5Smix=Sground,cm-3×k where k = 3.11 × 10⁻⁴ pptvh-1 per molec.cm-3s-1, calculated using the ideal gas law, calibrated based on the air molecular number density (1.159 × 10¹⁹ cm-3) and Nam Co's in-situ meteorology (mean daytime temperature: 278.15 K; average atmospheric pressure: 4.45 × 10⁴ Pa) to account for high-altitude thermodynamic properties.

Kruskal–Wallis test, and one-way analysis of variance (ANOVA) and Duncan's multiple comparisons (SPSS Version 26.0., IBM Corp., Armonk, NY, USA) were used to distinguish the differences of soil physicochemical properties, functional genes, microbial community diversity and Nr fluxes for different soil use types and fumigated soils, respectively. Non-metric multidimensional scaling (NMDS), analysis of similarity (ANOSIM) and non-parametric multivariate analysis of variance using distance matrices (ADONIS) were used to reveal differences in the composition of archaeal, bacterial and fungal communities across soil use types (R V3.6.3, “vegan” package, https://www.r-project.org/, last access: 26 October 2022). Spearman correlation analysis resolved the relationship between physicochemical properties, functional genes, microbial communities and Nr fluxes in Tibetan soils, while Pearson correlation analysis was applied to resolve the relationships in fumigated soils. Both methods were performed using Origin 2023.

Tibetan region has remarkable differences in physicochemical properties among different soil types, such as pH, TN, TC, C/N ratio, NH4+-N, particle size, and moisture (p<0.05, Table S5 in the Supplement). In this study, we detected that pH was alkaline in bare soil (8.14 ± 0.18) and wetland (7.84 ± 0.19), acidic in grassland (5.10 ± 0.19) and cropland (6.36 ± 0.64), whereas in forest pH (4.76–8.16) had a wide range of distributions in the Tibetan region, generally in accordance with the results of other studies (Ma et al., 2020; Wang et al., 2023b; Yang et al., 2021). Soil clay content was low (≤ 1 %), with essentially consisting of silt and sand in all soils (Table S5). It has been reported that carbon and nitrogen storage is substantial in surface soils of the Tibetan Plateau, particularly in grasslands (Chen et al., 2022a). Similar results were also obtained in this study, with the TN and TC contents of grassland being 3–26 times more than other soil types (Table S5). Agricultural soils showed comparatively higher NO3--N content and N transformation rate (net mineralization and nitrification rate) than other soil types, indicating fast N turnover effects caused by fertilization. In addition, Tibetan soils had elevated NO3--N content compared with NO2--N and NH4+-N content, which was similar to the results of other studies (Gao et al., 2016; Lin et al., 2019; Yao et al., 2019).

Overall, 1 202 964, 1 079 854 and 1 589 093 high-quality bacterial, archaeal and fungal gene sequences were detected from 25 soil samples, respectively. The sequences were clustered with 97 % similarity to obtain 8355, 531 and 5806 OTUs (Table S6 in the Supplement). The Shannon–Wiener dilution curve was flattened, indicating that the sequencing volume was able to reflect most of the microbial diversity information in the soil samples (Fig. S4 in the Supplement) (Wang et al., 2012; Wu et al., 2022a). The number of shared species of bacteria, archaea, and fungi across all soil types was 878, 46, and 75, respectively. However, archaea (average ∼ 71) and fungi (average ∼ 778) had more unique species than shared species in different soil types (Table S7 in the Supplement). In general, microbial community alpha-diversity was assessed in terms of richness, evenness, diversity and coverage, while community beta-diversity reflected differences in species composition between populations via comparisons of species composition structure between samples (He et al., 2013). In this study, the Good's coverage indices of bacterial, archaeal and fungal communities all ranged from 0.97 to 0.99 (Table S8 in the Supplement), indicating that the results of the current sequencing were highly plausible. It was found that the diversity and community structure of soil archaea, bacteria and fungi were moderately heterogeneous among the different soil types (Kang et al., 2021). This study revealed that archaeal community richness and fungal community richness, evenness and diversity were statistically different between soil types, whereas bacterial richness, evenness and diversity did not differ significantly (Fig. 1a). Meanwhile, based on the results of NMDS ordination (Fig. 1b) and statistical analyses (Table S9 in the Supplement, p<0.05), it was shown that beta-diversity of bacteria, archaea, and fungi varied significantly across soil types, in particular between cropland soils and other soil types. The results demonstrated that soil habitat specificity shaped the diversity of soil microbial communities on the Tibetan Plateau.

Through the sequencing results, soil bacteria, archaea and fungi were affiliated to 50, 12 and 17 phyla, respectively (Table S6). Figure 2a showed that the soil archaeal community was dominated by Crenarchaeota (10 %–99 %), Thermoplasmatota (0 %–59 %), Euryarchaeota (0 % –62 %), and Halobacterota (0 %–47 %) in the soil archaeal community. The dominant phyla of the bacterial community were Actinobacteriota (6 %–65 %), Proteobacteria (13 %–46 %), Acidobacteriota (3 %–25 %), Chloroflexi (3 %–21 %), and Firmicutes (1 %–23 %). The dominant phyla of the fungal community were Ascomycota (1 %–90 %), Basidiomycota (0 %–69 %), Mortierellomycota (0 %–84 %), and Chytridiomycota (0 %–62 %). It was similar with other studies reporting the composition of soil microbial communities on the Tibetan Plateau (Cao et al., 2022, 2021; Yang et al., 2022; Zou et al., 2023). Crenarchaeota predominated in the Tibetan soil archaeal community, especially in cropland soils (average relative abundance ∼ 98 %), far above other soil types. Additional studies confirmed that Crenarchaeota extensively exist in marine, hot spring and soil environments, which constitute the most numerous and widespread ammonia-oxidising archaeal taxa in nature, with the ability to oxidize ammonia to nitrite for chemoenergetic autotrophic growth, thus being critical for nitrogen fertiliser utilization (Baker et al., 2020; Brochier-Armanet et al., 2008; Dang et al., 2008). Further clarification was provided to explain the low NH4+-N content and high net mineralization/nitrification rates in croplands in Tibet (Table S5). It was found that Actinobacteriota genomes harboured numerous functional genes related to plant residue degradation, and that Ascomycota fungi exhibited the ability to degrade multiple types of plant cellulose and hemicellulose, which increased nitrogen availability during microbial-driven decomposition of plant residues (Bao et al., 2021; Swarnalakshmi et al., 2016; Zhang et al., 2013). Thus, those in cropland soils with high abundance of Actinobacteriota (average relative abundance ∼ 36 %) and Ascomycota (average relative abundance ∼ 54 %) presumably provide additional nitrogen for themselves and other microorganisms. Meanwhile, the abundance of key functional genes for N cycling in grassland and cropland soils was higher than that in other soil types, with napA, narG, nirK, and nosZ gene abundance higher in grassland soils, and amoA gene abundance higher in cropland soils, whereas all of the above genes had lower abundance in bare soils (Fig. 2b). This indicated that grassland and cropland soils in the Tibetan Plateau region possessed high nitrification and denitrification potentials.

Under natural conditions, soil–atmosphere exchange of reactive nitrogen is typically bidirectional (Bao et al., 2022). Here, we quantify net soil emissions and eliminate the masking effect of atmospheric Nr background by using dry Nr-free air as the carrier gas. For example, the observed HONO flux represents the net apparent flux, which integrates direct soil biogenic emission and additional HONO formation via heterogeneous uptake and conversion of atmospheric NO2 on soil particle surfaces (Fan et al., 2025). This dual-source characteristic helps explain the co-occurrence of HONO release and NO2 consumption observed in Fig. 3. It is important to acknowledge the methodological limitations regarding the interpretation of our flux results. In this study, all soils were found to emit of HONO, NO, NO2 and NH3 during wetting and drying cycles (Fig. 3). The optimum fluxes ranged from 2.5 to 235.2 ngNm-2s-1 for HONO, 1.2 to 268.4 ngNm-2s-1 for NO, 7.6 to 43.7 ngNm-2s-1 for NO2, and 6.4 to 547.8 ngNm-2s-1 for NH3 (Fig. 3a). Integrating the single wetting and drying cycle, the release of HONO was 0.02–0.30 mgNm-2, NO was 0.01–0.51 mgNm-2, NH3 was 0.47–3.07 mgNm-2, and NO2 was < 0 (Fig. 3b). When compared to studies with similar conditions or methods, the emissions in this study were consistent with those reporting soil HONO, NO, or NO2 emissions ranging from 1 to 3000 ngNm-2s-1 (Mamtimin et al., 2016; Meusel et al., 2018; Oswald et al., 2013; Song et al., 2023; Su et al., 2011; Weber et al., 2015; Wu et al., 2022a, 2019). Optimal fluxes and integrated emissions of HONO, NO and NO2 showed high consistency and convergence across cropland soils. Cropland soils consistently remained the primary source of HONO, NO and NO2 emissions (42.1 ± 28.1, 49.2 ± 31.8 and 19.9 ± 3.6 ngNm-2s-1, respectively), whether in terms of instantaneous peaks or cumulative emissions. The findings suggest that the emission mechanisms in cropland ecosystems for HONO, NO, and NO2 are relatively continuous over time and highly stable (Fig. 3). Studies have shown that high Nr gas emissions from cropland soils depend strongly on fertiliser application factors (Luo et al., 2022). Nitrogen fertiliser application promoted strong HONO emissions, the flux of HONO from fertilised agricultural soils can exceed 1000 ngNm-2s-1 (Tang et al., 2019; Xue et al., 2021, 2019). In addition, it has been shown that fertilization or grazing increased soil HONO emissions, which were expressed as farmland ≥ grassland ≥ forest and peatland for different soil types (Bhattarai et al., 2021). It is worth noting that the substantial NO2 emissions observed in this study cannot be fully explained by classical microbial nitrification and denitrification mechanisms alone. Apart from biological pathways, direct NO2 emissions are likely driven by several abiotic or non-classical biotic mechanisms. Chemodenitrification is recognized as a significant source of NO2 in soils; specifically, under acidic conditions, the accumulated NO2- in soil undergoes chemical decomposition to produce NO and NO2 (Cai et al., 2012).

However, it should be noted that extrapolating the results of this study to actual field conditions has inherent limitations. Xue et al. (2024) showed that surface water exchange plays a critical role in mediating soil Nr emissions. Under the dry-air conditions employed in our experimental setup, the evaporation of surface water is likely maximized. This enhanced evaporation could potentially facilitate the escape of dissolved Nr gases from the soil matrix, thereby leading to an overestimation of the net fluxes compared to ambient conditions with higher relative humidity. Therefore, our findings provide theoretical maximum emission potentials, and caution is advised when directly applying the absolute flux magnitudes obtained here to field assessments without considering local in situ environmental data (e.g., soil moisture, temperature). Future studies integrating in situ field measurements or environmentally controlled microcosm systems are essential to delineate the dynamic impacts of wet-dry cycles on Nr emissions.

Spearman's correlation analysis revealed that the potential of soil Nr gas emission in Tibet was remarkably correlated with soil physicochemical properties, microbial communities and key functional genes of the nitrogen cycle (Fig. 4, p<0.05). Generally, soil HONO and NO emissions were significantly correlated with soil pH and nitrogen content. Su et al. (2011) found that HONO emissions increased with higher nitrogen content and acidity in the soil. In contrast, Oswald et al. (2013) measured soil HONO emissions under different pH conditions globally and found that the fluxes did not decrease with increasing pH. Furthermore, another study demonstrated that the pH at the surface of soil mineral particles, rather than the pH of the soil solution, influences soil HONO emissions (Donaldson et al., 2014). Meusel et al. (2018) reported that HONO and NO emissions were significantly correlated with soil NO3- and NO2-. Wu et al. (2022b) found that HONO and NO emissions from different types of soils in Shanghai had been correlated remarkably with the concentrations of NH4+, NO3- and NO2- concentrations. Nevertheless, in this study, soil HONO was significantly correlated with amoA gene abundance and archaeal evenness, while soil NO was significantly correlated with amoA, narG, nirK, and Actinobacteriota abundance (Fig. 4, p<0.05). Wu et al. (2022a) found that amoA_AOA abundance directly affected the wetland soil HONO emission, while NO emission were mainly controlled by amoA_AOA, nirK and nirS genes. In this instance, soil nitrification rate was dictated by AOA and negatively correlated with the C/N ratio, particularly when the C/N ratio exceeds 25 (Maljanen et al., 2013; Mushinski et al., 2019). At this threshold, soil HONO and NO emissions decreased rapidly. Additionally, the fluxes of HONO and NO were affected by AOB, which influenced amo gene abundance through soil pH, subsequently impacting soil HONO emissions (Mushinski et al., 2019; Scharko et al., 2015).

Actinobacteria belong to Gram-positive bacteria, which can degrade soil organic matter, provide nutrients to the soil, as well as maintain high activity and DNA repair ability even at low temperatures, acting as a dominant group of bacteria in low-temperature environments (Johnson et al., 2007). In our study, we also found Actinobacteriota (6 %–65 %) to be the dominant phylum of the bacterial community in Tibet, especially in cropland soils (Fig. 2a). High NO emissions from cropland soils showed a strong consistency with Actinobacteriota abundances. Therefore, it is suggested that Actinobacteriota may act as a significant microbial regulator for NO emissions in Tibetan cropland soils. Overall, soil reactive N emission potential in Tibet was explained by soil pH, carbon and nitrogen content, (de)nitrification functional genes, and key microflora.

Atmospheric Nr emissions and deposition rates have been highly elevated in China during recent years, and studies have shown that a shift towards approximately equal wet and dry nitrogen deposition occurred in 2011–2015, accompanied by increasing dry nitrogen deposition (Liu et al., 2020; Yu et al., 2019; Zhu et al., 2022). Total N emissions from terrestrial ecosystems to be 100 TgNyr-1, with an eventual 70 TgNyr-1 being deposited back into terrestrial ecosystems (Fowler et al., 2013). Total annual input of atmospheric nitrogen dry deposition in China was 7.78 kgNha-1, of which the annual deposition flux of NO2 was 0.67 kgNha-1 (Jia et al., 2016). In this study, we found that exposure to simulated atmospheric NOx/O3 concentrations altered soil Nr emissions, while HONO and NO emissions from different soil types responded differently to NOx/O3 fumigation with different concentration gradients (Figs. 5 and S5–S7). Specifically, the HONO and NO emissions from forest soils were enhanced after NO and NOx/O3 mixture fumigation, which resulted in peak emissions that were 2–8 times higher than the previous levels (Fig. S5). However, HONO and NO emissions from cropland and grassland soils were suppressed by fumigation, with HONO and NO peaks in cropland soils decreasing by an average of approximately 31 % and 31 % (Fig. S7), while HONO and NO peaks in grassland soils were decreased by an average of approximately 68 % and 84 % (Fig. S6 in the Supplement). Interestingly, while cropland and grassland soils typically serve as sources of HONO and NO under natural conditions, their fluxes were significantly suppressed under fumigation with elevated NO and NOx + O3. These results indicate that high concentrations of NO and NOx + O3 may lead to a decrease in the soil production rates of HONO and NO, thereby causing a reversal of the source-sink balance. Previous study showed that the net atmospheric Nr budgets for paddy fields, forests, and tea plantations in a sub-tropical catchment to be +3.7, -36.1, and +23.8 kgNha-1yr-1, respectively, suggesting that forests served as sinks of atmospheric Nr, whereas paddy fields and tea plantations served as sources of atmospheric Nr (Zhu et al., 2021). In addition, forests have a higher efficiency of atmospheric nitrogen capture than other land-use types (Du et al., 2019). Therefore, forest soils under NOx/O3 exposure might absorb more inputs from external nitrogen sources, accelerating the rate of nitrogen transformation under the dry-wet cycle, which led to increased emissions of HONO and NO.

Fumigation with different concentration gradients of NO and NOx/O3 also altered soil physicochemical and biological properties (Fu et al., 2020; Ning et al., 2015). Forest (pH ∼ 5.59 ± 0.16), cropland (pH ∼ 5.44 ± 0.18) and grassland (pH ∼ 5.10 ± 0.05) soils were acidic before fumigation in the study. It was found that soil pH tended to increase after fumigation, especially in forest soils, with an average increase of ∼ 0.53 (Fig. S8 in the Supplement). The increase in pH is primarily attributed to changes in the redox state, H+ consumption during the reduction of nitrogen intermediates, and shifts in microbial functional communities (Chen et al., 2023; Slessarev et al., 2016; Wu et al., 2026). Under strong oxidative fumigation conditions, the abundance of denitrification genes increased (Fig. S8), which likely led to greater H+ consumption, thereby raising the soil pH. Furthermore, the differential responses among soil types, rather than a uniform shift across all treatments, suggest that the pH increase was predominantly driven by the fumigation gases rather than merely by the duration of laboratory incubation. Fig. S9 in the Supplement results showed that soil pH was remarkably positively correlated with HONO and NO emissions (p<0.001), in agreement with the findings of (Scharko et al., 2015). Therefore, soil pH increased probably one reason for the increased emissions of HONO and NO from forested soils under fumigation. Moreover, it has also been shown that soils with pH near neutrality were capable of more HONO emissions (Oswald et al., 2013). While atmospheric nitrogen deposition, as a new nitrogen input to terrestrial ecosystems, possibly causes an increased soil carbon storage (Bala et al., 2013; Magnani et al., 2007), consistent with the results of this study on increased soil DOC content under fumigation. It was shown that atmospheric nitrogen deposition decreased the soil C/N ratio, facilitated the mineralization of organic matter in the soil, increased the accumulation of inorganic nitrogen in the soil (Fu et al., 2020; Lu et al., 2011). This phenomenon was seen in cropland soils, where increased NO3- and NH4+ content was found at elevated fumigation concentrations (Fig. S8). The results showed that under NO and NOx/O3 fumigation, cropland soils possibly preferred to undergo mineralization and nitrification processes, which resulted in higher nitrogen accumulation and lower gaseous losses, consistent with the results of the Nr emissions we measured (Fig. S7). Grassland soils under high concentration of NO fumigation increased pH and decreased NO2- content, which was not beneficial to soil HONO emissions in this condition, as well as raised NO3- content indicating that grassland soil tended to undergo the process of NO3- accumulation more than NO3- consumption under high concentration of fumigation (Figs. 5 and S8). Therefore, HONO and NO emissions from grassland soils under NO and NOx/O3 fumigation tended to decrease (Figs. 5 and S6). Functional microbial genes associated with nitrogen cycling in forest soils were sensitive to nitrogen addition (Levy-Booth et al., 2014). In this study, the abundance of nxrB, nirK and nirS genes was altered in forest soils under different concentration gradients of NO and NOx/O3, which increased significantly under high NOx/O3 fumigation, resulting in the decomposition of more NO3- and NO2-, thereby releasing more HONO and NO (Figs. S5 and S8). It should be noted that the effects observed in the fumigation experiments (Figs. S5–S7) are discussed qualitatively but are not incorporated into the annual emission parameterization. The annual model relies on long-term in-situ soil properties, which makes it challenging to extrapolate short-term chemical feedback to a regional scale. Therefore, this qualitative discussion highlights the complex interplay between atmospheric chemistry and soil biogeochemistry. Furthermore, exposure to elevated NOx/O3 also affected the diversity and composition of soil microorganisms, as well as the key flora involved in the nitrogen cycle, all of which were also related to soil Nr gas emissions (Berthrong et al., 2014; Fierer et al., 2012; Leff et al., 2015; Treseder, 2008; Wang et al., 2018; Zhang et al., 2018b). However, this study has not obtained the relevant data, which leads to the limitation of the study modeling the effects of atmospheric NOx/O3 exposure on soil Nr gas emission.

The annual emissions of soil Nr gases in Tibet were estimated as 7.0 ± 3.4 GgNyr-1 for HONO, 11.6 ± 7.8 GgNyr-1 for NO, and 20.3 ± 7.0 GgNyr-1 for NOx in 2021, respectively (Fig. 6 and Table S10 in the Supplement). These emissions were lower with the recently reported HONO value of 12.3 GgNyr-1 but higher than the NOx value of 8.9 GgNyr-1 for 2015 (Li et al., 2026). This discrepancy arises mainly from improved model inputs in this study, including in-situ soil moisture and microbial activity data, in contrast to statistical interpolation with sparse field observations in previous work.

By land-use type, bare soils are the largest contributor due to their vast coverage (43 % of land area), accounting for 43 % of HONO (3.0 ± 2.0 GgNyr-1), 46 % of NO (5.3 ± 4.2 GgNyr-1), and 67 % of NOx (13.6 ± 3.6 GgNyr-1) emissions. This aligns with studies highlighting high Nr emissions from desert soils and biological soil crusts (Bhattarai et al., 2021; Meusel et al., 2018). Grasslands contribute 29 % to HONO and 44 % to NO emissions. Our estimated grassland NO emission of 0.2 (0–1.7) kgNha-1yr-1 is comparable with previous plateau studies (Gao et al., 2016; Holst et al., 2007; Lin et al., 2019; Yao et al., 2019; Zhang et al., 2018a). Croplands, despite having the highest fluxes per unit area (Fig. 3), contribute only 8 % to HONO emissions (0.5 ± 0.1 GgNyr-1) due to their limited spatial extent (∼ 0.3 %). This contrasts with global patterns where croplands are often the dominant source (Wu et al., 2022b), highlighting the unique emission profile of the Tibetan Plateau where natural ecosystems prevail. To validate our estimates, the upscaling model integrated both measured and published field data from the Tibetan Plateau (Table S4). Our HONO and NOx estimates are consistent with recent regional bottom-up inventories (Li et al., 2026; Wu et al., 2022b) and provide useful constraints for GEOS-Chem simulations over the Third Pole (Wang et al., 2025a).

To quantify the contribution of soil HONO emissions to the atmospheric budget, we integrated our flux data with observations from the Nam Co Multisphere Observation and Research Station. Our optimized calculation framework, combining the external source definition and the soil-atmosphere exchange model (Meusel et al., 2018), reveals that soil HONO emissions on the Tibetan Plateau produce a volume mixing ratio of 10.5 pptvh-1. This accounts for approximately 10.5 % of the 100 pptvh-1 external daytime HONO source identified in the plateau's background atmosphere (Wang et al., 2023a). When combined with the estimated surface nitrate photolysis contribution of ∼ 35 pptvh-1 from the same study, these two pathways jointly explain 45.5 % of the missing HONO source. These findings highlight that surface processes, particularly soil emissions, play a crucial role in regulating atmospheric HONO and consequently influence the oxidative capacity of the plateau's atmosphere.

For NOx, our estimated soil emissions of 20.3 ± 7.0 GgNyr-1 dwarfs the weak anthropogenic emissions of less than 5 GgNyr-1 reported by Wang et al. (2023a). This establishes soils as the dominant natural NOx source on the plateau. The high HONO to NOx (HONO/NOx) emission ratio, we observed for bare soil (0.49) is consistent with the elevated daytime HONO/NOx concentration ratio of 0.31 ± 0.06 observed at Nam Co. This observed concentration ratio significantly exceeds the photo-stationary state ratio (0.02–0.05) for internal cycling, which confirms that soil emissions are a primary driver of this chemical disequilibrium. Under intense solar radiation on the Tibetan Plateau, HONO emitted from soils undergoes rapid photolysis to generate OH radicals, while NOx emissions from soils provide key precursors for tropospheric O3 production. Continuous warming and wetting of the Tibetan Plateau (Yan et al., 2026) likely further enhance soil Nr emissions, forming positive feedback that strengthens atmospheric oxidation capacity. This climate–chemistry feedback requires accurate representation in Earth system models to support reliable projections of air quality, ecosystem stability, and the lifetime of greenhouse gases over this climatically sensitive region.

Our study further identifies that soil Nr emissions uniquely modulate the regional atmospheric chemical balance by elevating the HONO/NOx ratio beyond the atmospheric internal cycling equilibrium. Our field-constrained soil NOx and HONO emission estimates provide key constraints to reduce the underestimation of O3 production in GEOS-Chem simulations (Wang et al., 2025a) and narrow the gap between modelled and observed atmospheric chemistry over the Third Pole. The distinctiveness of this effect stems from the plateau's extreme environment: intense solar radiation (jO1D up to 5.5 × 10⁻⁵ s-1, double the sea-level value) accelerates HONO photolysis to OH radicals, while the high background O3 concentrations (Yin et al., 2017) mean soil NOx emissions sustain O3 accumulation without significant anthropogenic interference. In contrast to global studies that emphasize the role of soil HONO emissions in particulate matter pollution (Li et al., 2026), our regional observations demonstrate that soil Nr emissions on the Tibetan Plateau directly and strongly regulate atmospheric oxidation capacity and O3 cycling, with profound cascading effects on regional air quality and climate systems.