A schematic of the Pas-QW used in this study is shown in Supplementary Figure S1,S2. The sampler consists of a quartz wool disk (14.0 cm diameter, 1.35 cm thickness, and 1.0 g mass), which was placed inside the pan and covered with a stainless-steel mesh (1 × 1 cm). For the uptake study, two batches of Pas-QW from different time periods were arranged in the field deployments for better contrast. The first batch of 15 passive samplers was set up and deployed in Yantai (YT) from January to April 2019, with every three parallel Pas-QW collecting at 15, 29, 44, 60, and 77 d intervals over the period. The second batch of 18 passive samplers was set up and deployed in Guangzhou (GZ) from October to December 2020, with every three parallel Pas-QW collecting at 15, 29, 41, 53, 62, and 74 d intervals (Supplementary Table S1,S2,S3). Neither of the sampling sites had any obvious point emission sources nearby. During field sampling, the AAS was operated continuously alongside passive sampling throughout the calibration exercise. The AAS was set at an identical height to deploy the passive sampler, and total suspended particle samples were collected on prebaked quartz fiber filters. The samples were collected for a 2–5-day duration at a stable flow rate of 300 L/min. Because of inevitable factors such as power interruption, heavy rainfall, and typhoons, the active sampling time did not completely match the retrieval time of the Pas-QW samplers. Therefore, the average concentrations of NO₃ ⁻ obtained from active sampling during the corresponding Pas-QW deployment time were used as the reference air concentrations during the calibration exercise. Detailed information on the calibration exercise is provided in Supplementary Text S2.

A total of 27 air-monitoring stations were selected as sampling sites in China from March to June in 2017, and 11 air-monitoring stations were selected in the Indochina Peninsula (ICP) from March to June in 2016. These included 25 urban sites, 10 rural or suburban sites, and three background sites. Detailed information and basic parameters of each sampling site are provided in Supplementary Table S4,S5. The air samplers were assembled at the deployment sites to avoid contamination during the transport process. After deployment, these quartz wool disks were retrieved, resealed, and transported back to Guangzhou for storage.

The quartz wool disk samples were soaked in Milli-Q water, ultrasonically extracted for 30 min, and then the extracts were filtered. Ionic species were measured via ion chromatography (Dionex ICS3000; Dionex Ltd., Sunnyvale, CA, United States) based on the reported analysis method (Zong et al., 2018). The analytical precision determined from the replicates was less than 5% for each ion. The detection limits were within the range of 0.01–0.05 μg/m³ for anions and 0.02–0.07 μg/m³ for cations. Blank values were used for calibration (Supplementary Text S3).

The δ¹⁵N and δ¹⁸O values of NO₃ ⁻ were determined by the N₂O isotope analysis method (Mcilvin and Altabet, 2005). Briefly, NO₃ ⁻ extracted from the filter was reduced to NO₂ ⁻ using cadmium powder, and then NO₂ ⁻ was further reduced to N₂O using sodium azide in an acetic acid buffer. δ¹⁵N and δ¹⁸O were analyzed using an isotope ratio mass spectrometer (MAT253; Thermo Fisher Scientific, Waltham, MA, United States). The detected δ¹⁵N and δ¹⁸O values (‰) were corrected by multi-point correction (r² = 0.999) based on international reference standards (IAEA-NO₃ ⁻, USGS32, USGS34, and USGS35) and reported in the standard delta notation relative to the Vienna Canyon Diablo Troilite standard. The standard deviation of the isotopic measurements for both standards and samples was less than 0.35‰ for δ¹⁵N and 0.70‰ for δ¹⁸O. In addition, NO₂ ⁻ concentrations in the quartz wool disk samples were usually less than 1% of that of NO₃ ⁻; therefore, they were neglected in the δ¹⁵N and δ¹⁸O analyses.

The Bayesian mixing model could use stable isotopes to identify the probability distribution of the contributions of each source to a mixture and explicitly explain the uncertainty, which is associated with multiple sources, fractionation, and isotopic signatures (Parnell et al., 2013). Recently, the Bayesian model (MixSIR) was improved by incorporating the isotopic fractionation of the equilibrium/Leighton reaction, thereby constructing a suitable NO_x source apportionment model (Zong et al., 2020). In our study, coal combustion (13.7 ± 4.6‰), biomass burning (1.0 ± 4.1‰), mobile sources (−7.2 ± 7.8‰), and microbial processes (−35.4 ± 10.7‰) were confirmed as the major contributors to NO_x in the CICP (Supplementary Text S4 and Supplementary Table S6). The details of the isotopic fractionation are discussed in Supplementary Text S5 and Supplementary Table S7.

The uptake profiles of NO₃ ⁻ over time are shown in Figure 1A for both YT and GZ, with raw data and detailed information presented in Supplementary Table S1 and Supplementary Text S2. Linear uptakes of NO₃ ⁻ for Pas-QW during the calibration exercise period were found by good linear correlations between the values of uptake volume and deployment time. The average sampling rate was 2.5 ± 1.2 m³/day, which was comparable to the recently reported sampling rates of sulfate based on Pas-QW (Wang et al., 2021). To reflect the sampling rate as much as possible, two sampling calibrations were carried out in different seasons in the North and South of China. The sampling time of YT is the northern winter with relatively low temperature, humidity, and light intensity, which can cause relatively low oxidant concentration and low NO_x oxidation rate in the atmosphere, thereby reducing the atmospheric NO₃ ⁻ concentration. The sampling time of GZ is the southern autumn characterized by low NO_x concentration, but the high oxidant concentration caused by the relatively high temperature, humidity, and light intensity can increase atmospheric NO₃ ⁻ concentration. Although the growth trend of nitrate collected in the two cities is slightly different, the capture by Pas-QW both increased linearly over the period and the difference in sampling rates was not statistically significant (p > 0.05). This suggested that the Pas-QW offered a feasible tool for monitoring studies and the sampling rate was stable and not greatly affected by atmospheric NO₃ ⁻ concentration. In addition, the sampling rate is not only affected by the physical and chemical properties of the target compound but also influenced by atmospheric conditions. Information on the correlation between sampling rates and environmental and meteorological parameters is given in Supplementary Table S2. The results showed that the correlation between the sampling rate and these basic parameters is not significant, indicating that the influence of changes on meteorological parameters may be weakened during the long-term sampling process and would not have a great impact on the time-weighted average concentration.

The isotopic information (δ¹⁵N and δ¹⁸O) of NO₃ ⁻ collected with Pas-QW and AAS is presented in Figure 1B and Figure 1C, respectively, with raw data listed in Supplementary Table S3. Only the values for GZ were compared to avoid the isotopic fractionation effects due to unavoidable factors such as transportation or long-term storage. A considerable discrepancy in δ¹⁵N-NO₃ ⁻ values was observed between Pas-QW and AAS at 29-, 41-, 53-, 62-, and 74-d intervals. Considering the uncertainty caused by the adsorption of nitric acid gas during sampling (Zhang and Mcmurry, 1992; Schaap et al., 2004), the δ¹⁵N-NO₃ ⁻ values collected by the AAS are generally slightly depleted than those collected by Pas-QW (Figure 1B). However, the δ¹⁵N-NO₃ ⁻ values collected by Pas-QW were linearly enriched over 74 days, indicating that nitrate mainly collected with alkali metals has been retained in the quartz wool disk due to the evaporation of semi-volatile ammonium nitrate (Zhang and Mcmurry, 1992; Schaap et al., 2004). Comparing the time-averaged δ¹⁵N-NO₃ ⁻ values collected by Pas-QW and the AAS during the sampling period, the correlation (δ¹⁵N_AAS-NO3− = -0.5×δ¹⁵N_{Pas-QW-NO3−} + 5.8, r = 0.6) between the two samplers is reasonable. Therefore, to ensure comparability, the δ¹⁵N-NO₃ ⁻ values collected by Pas-QW in this study have been corrected to those of AAS using this equation. As for the δ¹⁸O of nitrate aerosol, there was no significant difference in δ¹⁸O-NO₃ ⁻ values between Pas-QW and the AAS during the sampling period (Figure 1C). The two were in the same order of magnitude, and the variability of each period was between 0.9 and 1.2, indicating that the oxygen isotopes collected by Pas-QW were basically consistent with those collected by the AAS over the long-term sampling process.

To further verify the feasibility of Pas-QW, samples from twelve sampling sites (HRB, SY, XM, WH, SH, CD, ZZ, TY, SJZ, BJ, NJ, and HZ) were collected in March–June of 2017 and compared with those of the previous studies conducted in the same regions and in the same year. The mass size distribution of NO₃ ⁻ is dominated by the accumulation mode, and its concentration in PM_2.5 is about 75% of that in TSP (Lijiang, 2003; Wang et al., 2003; Fang et al., 2013). The results for the collection of target compounds from both fine PM and coarse PM using Pas-QW were comparable to those of most of the previous studies conducted on PM_2.5 samples, which have been corrected to the concentrations in TSP (Hong et al., 2018; Huang et al., 2018; Wang P et al., 2019; Wang Y. et al., 2019; Huang et al., 2019; Sun et al., 2019; Wu et al., 2019; Zhang et al., 2019; Wang et al., 2020; Dong et al., 2021; Liu et al., 2021; Luo et al., 2021). The comparison between the literature values and the measured values in this study is shown in Figure 1D. Two values at the same site are in the same order of magnitude, and the slope was close to 1 (R² = 0.6). This suggested that the detected concentrations of NO₃ ⁻ in this study were comparable to the concentrations reported in the literature. The Pas-QW provides a direct and effective measurement of atmospheric NO₃ ⁻ over a large scale.

Spatial information on NO₃ ⁻ concentrations in the CICP is provided in Figure 2 and Supplementary Table S4. During the sampling period, the temperature, relative humidity, and wind speed were within the ranges 9.1–35.1°C (21.5 ± 7.3°C), 36.0–84.0% (63.4 ± 14.0%), and 0.9–3.8 m/s (2.4 ± 0.8 m/s), respectively. The meteorological parameters of these sampling sites were comparable to the range of the calibration exercise time, except for a few sites. Among the 38 sampling sites, the average NO₃ ⁻ concentration in the ICP (6.4 ± 2.3 μg/m³) was slightly lower than that in China (7.8 ± 5.0 μg/m³). Higher values typically appeared in urban sites (8.8 ± 4.5 μg/m³), followed by rural sites (5.3 ± 2.8 μg/m³) and background sites (2.7 ± 0.4 μg/m³). Analysis of variance showed that there was a significant difference between the NO₃ ⁻ concentrations of different functional areas (p < 0.05). The lowest NO₃ ⁻ concentration was in Alxa, which was located in the northwest of China, with a vast area surrounded by desert. The highest NO₃ ⁻ concentration was in Hangzhou, located in the Yangtze River Delta in China. As a major commercial and financial center in mainland China, it has a prosperous economy, with thriving industrial activity and rapidly developing tourism (Tian et al., 2018; Jin et al., 2021). The spatial distribution of NO₃ ⁻ concentrations was largely attributed to the development of the regional economies, and cities with intensive human activities usually have higher NO₃ ⁻ concentrations. A similar phenomenon was found in the NO_x emission study derived from satellite observations (Supplementary Figure S3) (Ding et al., 2017).

In terms of its isotopic compositions, the average δ¹⁵N-NO₃ ⁻ values in China (4.9 ± 1.6‰) were lower than those in the ICP (6.4 ± 1.6‰). The values in background sites (6.1 ± 0.2‰) and rural sites (6.1 ± 1.3‰) are higher than those in urban sites (4.9 ± 1.8‰). The highest δ¹⁵N-NO₃ ⁻ value was in Taiyuan (9.5‰), which was located in central China, and the contribution of the coal industry to total production is more than one third (Yearbook S. S, 2018). The lowest δ¹⁵N-NO₃ ⁻ value was in Chongqing (2.5‰), located in the Sichuan Basin, with large populations and intensive agricultural activities (Yearbook S. S, 2018). The average δ¹⁸O-NO₃ ⁻ value in this study was 42.3 ± 8.5‰, which is consistent with that of the previous studies in the same season (Su et al., 2020; Zong et al., 2020). The values of background sites (41.5 ± 7.6‰), rural sites (40.2 ± 10.8‰), and urban sites (43.2 ± 7.3‰) were not significantly different (p > 0.05). The δ¹⁸O-NO₃ ⁻ values were largely affected by the different pathways of nitrate formation (Hastings et al., 2003; Chang et al., 2018; Fan et al., 2020). According to the global models, the contribution of the •OH generation pathway to nitrate production was equal to that of the N₂O₅ generation pathway in the troposphere below 1 km altitude, especially in spring and summer, and the dominant •OH generation pathway under high OH-level conditions may result in low δ¹⁸O-NO₃ ⁻ values (Alexander et al., 2020; Luo et al., 2020). In addition, recent studies have shown that the peroxy radical reaction is also an important pathway for NO to NO₂ conversion (Fang et al., 2011; Guha et al., 2017). These changes in the NO_x oxidation pathways may lead to generally low δ¹⁸O-NO₃ ⁻ values in the atmosphere.

In the atmosphere, NO₃ ⁻ is mainly produced by the conversion of NO_x, which exhibits different δ¹⁵N-NO_x signatures from its distinctive emissions. As previously mentioned, NO_x from coal combustion or biomass burning has higher δ¹⁵N values, while δ¹⁵N from mobile sources or microbial processes tends to be more negative (Hastings et al., 2009; Miller et al., 2017; Zong et al., 2018). During the transformation process of NO_x to NO₃ ⁻, the nitrogen isotopic fractionation effects on δ¹⁵N cannot be neglected (Walters et al., 2015; Walters and Michalski, 2015). Certifiably, the δ¹⁵N-NO₃ ⁻ values in this study had a significant correlation with temperature (r = 0.32, p < 0.05; Supplementary Table S5). According to the fractionation theory in the conversion of NOx to NO₃ ⁻, changes in temperature can result in the variation of nitrogen isotopic fractionation (Walters and Michalski, 2016), which may further affect the δ¹⁵N-NO₃ ⁻ values. To assess the impact of temperature on the δ¹⁵N-NO₃ ⁻ value, we calculated the nitrogen isotopic fractionation in these regions (Supplementary Text S5 and Supplementary Table S7). The fractionation factor of δ¹⁵N for the sampling sites varied from 8.8 to 10.6‰, with a mean value of 9.7 ± 1.8‰. The nitrogen isotopic fractionation between the sites with the highest temperatures (LAO-V) and that with the lowest temperatures (HRB) was approximately 1.8‰, which is lower than the difference in real nitrogen isotope values between the two regions (3.2 ‰). This indicated that although the variation of δ¹⁵N-NO₃ ⁻ values was affected by isotopic fractionation, source differences were still a vital factor affecting the regional distribution of δ¹⁵N-NO₃ ⁻ values. The low values in urban sites and high values in background and rural sites can be attributed to increased contributions of mobile sources and coal combustion, respectively. Regional differences were also observed in five Chinese megacities based on the isotopic information and modeling (Zong et al., 2020).

Previous studies have suggested that the oxidation of NO₂ by •OH (•OH + NO₂) in the gas phase and the hydrolysis of N₂O₅ on pre-existing aerosols (N₂O₅+H₂O, heterogeneous process) were considered to be the two main reactions controlling the formation of NO₃ ⁻ (Morin et al., 2009; Xiao et al., 2020). The conversion process of NO to NO₂ mainly includes O₃ oxidation and peroxy radical (HO₂ and its organic homolog RO₂) oxidation. According to the global models, we assumed that 15% of NO to NO₂ conversion was through HO₂/RO₂ oxidation and 85% by O₃ oxidation (Alexander et al., 2020). The details of oxidation pathways and balance equations are discussed in Supplementary Text S1. The use of δ¹⁸O-H₂O ranges from −27.5 to 0‰, and the terminal δ¹⁸O-O₃ value (δ¹⁸O-O₃ ^*) is 130.4 ± 12.9‰ (Fang et al., 2011; Vicars and Savarino, 2014). Through the temperature-dependent equilibrium isotope exchange between •OH and H₂O, the resulting minimum and maximum estimates of δ¹⁸O-OH were from −67.4‰ to −41.0‰ (Walters and Michalski, 2016). Using these assumptions and equations, the expected δ¹⁸O-NO₃ ⁻ value for the •OH oxidation pathway is 46.5–71.4‰, and for the N₂O₅ pathway, it is 88.7–113.5‰. In this study, the observed δ¹⁸O-NO₃ ⁻ values in most regions were less than 60‰, suggesting that NO_x oxidation by •OH during this period was, indeed, the dominant pathway for the formation of NO₃ ⁻ in the atmosphere.

The characteristics of isotope compositions and basic environmental parameters of 38 sampling sites are shown in Figure 3. The relatively high δ¹⁸O-NO₃ ⁻ values of China were mainly found in megacities (such as GZ, NJ, ZZ, and BJ), which were consistent with the O₃ distribution in China during 2016 (Li et al., 2018). Moreover, the δ¹⁸O-NO₃ ⁻ values share a significant positive correlation with NO₃ ⁻ concentrations (r = 0.36, p < 0.05), and the spatial distribution of the latter was mainly attributed to the development of regional economies. These evidence imply that the high O₃ concentrations induced by human activities increase the proportion of N₂O₅ heterogeneous reactions, affecting the regional distribution of δ¹⁸O-NO₃ ⁻ values. Using the linear regression method (Supplementary Table S5), we also found that the δ¹⁸O-NO₃ ⁻ values share a significant negative correlation with latitude (r = −0.43, p < 0.01). This indicated that in addition to anthropogenic factors, there are other factors that may affect the regional distribution of δ¹⁸O-NO₃ ⁻ values, especially when multiple functional areas were included over a large scale. The wind field in the CICP during the sampling period is provided in Supplementary Figure S4 and Supplementary Text S6. The CICP has a definitive monsoonal climate, with a prevailing approximately southerly wind from March to June. This leads to the transport of water vapor from the evaporation of tropical and subtropical oceans to higher latitudes and inland through meridional and zonal circulations (Zhang and Yao, 1994). Given the preferential condensation of heavy isotopes, oxygen isotopes in atmospheric precipitation continue to be lighter as water vapor condenses and evaporates (Dansgaard, 1964; Yuetsever, 1975). The decreasing trend of δ¹⁸O-NO₃ ⁻ values from south to north in the CICP was consistent with the variation of δ¹⁸O-H₂O values at latitude, indicating that changes in the oxygen isotopic fractionation of the oxidant induced by natural factors may have a significant impact on the δ¹⁸O-NO₃ ⁻ values over a large spatial scale.

The overall estimation showed that coal combustion was the most important source for NO_x, followed by biomass burning, mobile sources, and biogenic process, with the contribution of 38.6 ± 8.0%, 25.5 ± 1.9%, 22.6 ± 2.8%, and 13.3 ± 3.4%, respectively (Supplementary Figure S5). Hence, the fractional concentration of the total NO₃ ⁻ that was derived from CC, BB, MS, and BS was 2.8 ± 1.6 μg/m³, 1.9 ± 1.2 μg/m³, 1.7 ± 1.1 μg/m³, and 1.0 ± 0.7 μg/m³, respectively (Figure 4). This finding was in agreement with the characteristics of energy consumption in China, which showed that coal combustion is dominant in energy combustion, but the proportion is gradually decreasing (Supplementary Figure S6). According to the “Twelfth Five-Year Plan for National Environment Protection,” 95% of all power plants in China were installed with NO_x removal systems by the end of 2015 (Shang et al., 2016). NO_x emissions decreased from 28 million tons in 2012 to 21 million tons in 2017, with significant proportion of the decrease attributed to the reduction in emissions from power plants (Supplementary Figure S7) (Liu et al., 2015; Li et al., 2017). The cities in Beijing–Tianjin–Hebei (BTH) and Yangtze River Delta (YRD) have higher CC concentrations, which can be partly attributed to their large populations and substantial coal consumption. In 2017, the power consumption in Jiangsu, Zhejiang, and Hebei provinces ranked second, fourth, and fifth, respectively, and their thermal power generation all exceeded 280 TWh (Yearbook C. S, 2018). In addition, GY, SSPN, and most sites in the ICP also have higher CC concentrations. These may be due to the anthropogenic emissions from local households, inefficient combustion, and the impact on the regional transmission of pollutants in the Sichuan Basin (SC), South Asia, and the northwestern Pearl River Delta. In China, residential consumption of coal has been calculated to be approximately 2.7 billion tons in 2017, and the pollution load of these households without pollution control devices would be equivalent to that of 170 billion tons of coal-fired power plants based on the emission calculation (Zhang et al., 2017). The higher CC concentrations typically appeared in urban sites (3.2 ± 1.5 μg/m³), followed by rural sites (2.3 ± 1.5 μg/m³) and background sites (1.1 ± 0.5 μg/m³). This result was consistent with that of previous research, that is, coal combustion in residential stoves was a widespread source from urban to remote areas in the CICP (Chen et al., 2017; Song et al., 2021).

The model results showed that biomass burning was the second largest contributor to NO_x with the concentrations of 1.6 ± 0.5 μg/m³ and 2.0 ± 1.3 μg/m³ in the ICP and China, respectively. The higher BB concentrations were found in SC and YRD of China, which is consistent with the distribution of cultivated land (Supplementary Figure S8). In cultivated areas, agricultural wastes are usually stored as fuel for household cooking or heating, while other wastes are consumed by open burning in the fields. China ranks third in the world with 1.43 million square kilometers of arable land, and the NO_x emissions induced by biomass burning increased more than six-fold from 1990 to 2013 (Li et al., 2016). The ICP is a typical high-incidence area of global biomass burning and occupies the leading position of biomass combustion pollution emission in Asia (Streets et al., 2003). The higher BB concentrations in this study may be mainly attributed to the highly concentrated combustion before farming in spring (from March to April) and various types of combustion (burning which includes natural (wildfire) and man-made forest fire (clearing forest)) (Streets et al., 2003). The high BB concentrations were observed in both rural and urban sites, indicating that biomass fuel is used as a conventional fuel and its emission intensity could not be ignored.

The mobile source was another important contributor to NO_x, and high concentrations were observed in urban sites (2.1 ± 1.2 μg/m³), especially in the typical urban agglomerations such as BTH and YRD. According to automobile industry surveys, the average annual growth rate of car ownership in China reached 14% since 2009, and more than one-half of motor vehicles occurred in megacities (Yearbook C. S, 2018). Since the mass ratio of NO_x/SO_x from gasoline combustion and coal combustion has been estimated to be 13:1 and 1:2, respectively, the ratio of NO₃ ⁻/SO₄ ²⁻ has often been used as an indicator of automobile exhaust emissions and coal combustion (Wang et al., 2006). The mass ratio of NO₃ ⁻/SO₄ ²⁻ in this study ranged from 0.12 to 2.87, and a high NO₃ ⁻/SO₄ ²⁻ ratio in cities was consistent with the result obtained by the Bayesian model. These suggested that the explosive growth of car ownership in recent years may lead to vehicle exhaust emissions becoming an important source of NO_x pollution. In addition, it has been estimated that the NO_x emissions from ships in China reached 11.3% of the total emissions in recent years (Zhang et al., 2016). The high MS concentrations in this study were also observed in coastal areas, indicating that ship emissions from the Bohai Sea, the East China Sea, the South China Sea, and the Bay of Bengal may be another important mobile source. This finding was consistent with the conclusions of a maritime NO_x emission study over Chinese seas based on satellite observation (Ding et al., 2018).

Microbial activity in wetlands and soils can consume accumulated nitrogen and release large pulses of NO, especially in environments with relatively high temperature and humidity (Zong et al., 2020). Thus, compared with other sources, microbial processes are more uniformly distributed in the ICP. Under high natural productivity and intensive agricultural activities, the extensive use of nitrogen fertilizers in cultivated land can lead to high NO_x emission (Jaeglé et al., 2004). In addition, high MS concentrations were also observed in cities around the YRD and Bohai Bay, indicating that oceanic sources (especially in marine sediments and estuaries) may be another important component of microbial processes. Although the contribution of marine bacteria to NO_x has not been widely reported, marine sediments and estuaries can be part of denitrification areas (Wankel et al., 2006). Based on the similar activity mechanism of bacteria, the oceans around the CICP may also be one of the contributors to NO_x emissions.