Different physicochemical behaviors of nitrate and ammonium during transport: a case study on Mt. Hua, China

. To understand the chemical evolution of aerosols in the transport process, the chemistry of PM 2 . 5 and nitrogen isotope compositions on the mountainside of Mt. Hua ( ∼ 1120 m above sea level, a.s.l.) in inland China during the 2016 summertime were investigated and compared with parallel observations collected at surface sampling site ( ∼ 400 m a.s.l.). The PM 2 . 5 exhibited a high level at the mountain foot site (MF; average 76 . 0 ± 44 . 1 µgm − 3 ) and could be transported aloft by anabatic valley winds, leading to the gradual accumulation of daytime PM 2 . 5 with a noon peak at the mountainside sampling site (MS). As the predominant ion species, sulfate exhibited nearly identical mass concentrations at both sites, but its PM 2 . 5 mass fraction was moderately enhanced by ∼ 4 % at the MS site. The ammonium variations were similar to the sulfate variations, the chemical forms of both of which mainly existed as ammonium bisulfate (NH 4 HSO 4 ) and ammonium sulfate ((NH 4 ) 2 SO 4 ) at the MF and MS sites, respectively. Unlike sulfate and ammonium, nitrate mainly existed as ammonium nitrate (NH 4 NO 3 ) in ﬁne particles and exhibited decreasing mass concentration and proportion trends with increasing elevation. This ﬁnding was ascribed to NH 4 NO 3 volatilization, in which gaseous HNO 3 from semi-volatile NH 4 NO 3 subsequently reacted with dust particles to form nonvolatile salts, resulting in signiﬁcant nitrate shifts from ﬁne particles into coarse particles. Such scavenging of ﬁne-particle nitrate led to an enrichment in the daytime 15 N of nitrate at the MS site compared with to the MF site. In contrast to nitrate, at the MS site, the 15 N in ammonium depleted during the daytime. Considering the lack of any signiﬁcant change in ammonia (NH 3 ) sources during the vertical transport process, this 15 N depletion in ammonium was mainly the result of unidirectional reactions, indicating that additional NH 3 would partition into particulate phases and further neutralize HSO − 4 to form SO 2 − 4 . This process would reduce the aerosol acidity, with a higher pH (3 . 4 ± 2 . 2) at the MS site and lower ones (2 . 9 ± 2 . 0) at the MF site. Our work provides more insight into physicochemical behaviors of semi-volatile nitrate and ammonium, which will facilitate the improvement in the model for a better simulation of aerosol composition and properties.


Introduction
Atmospheric particulate matter measuring equal to or less than 2.5 µm in aerodynamic diameter (PM 2.5 ) is a worldwide air pollution burden that can deteriorate the urban air quality and induce adverse human health effects that contribute to lowered life expectancies (Shiraiwa et al., 2017;Lelieveld et al., 2015;Fuzzi et al., 2015;Wang et al., 2016). Recent studies have disclosed that the mechanisms underlying these effects are profoundly dependent on particle properties, e.g., the size, concentration, mixing state, and chemical compositions of particles Liu et al., 2021;Guo et al., 2014). Thus, since 2013, China has issued strict emission directives to mitigate haze pollution. Consequently, the annual PM 2.5 concentration in China fell by approximately one-third from 2013-2017 . These controls notwithstanding, the PM 2.5 levels in most cities in China still exceed the least-stringent target of the World Health Organization (2021; 35 µg m −3 ), especially in rural areas and small cities (Lv et al., 2022;Li et al., 2023).
Near-surface particulate matter (PM) can also be transported to the upper air, and this process critically impacts radiative forcing, cloud precipitation, and the regional climate by scattering/absorbing solar radiation and by influencing aerosol-cloud interactions (Van Donkelaar et al., 2016;Andreae and Ramanathan, 2013;Fan et al., 2018). Past assessments of these effects have been characterized by large uncertainties (Carslaw et al., 2013); for example, Bond et al. (2013) found that black carbon climate forcing varied from +0.17 to +2.1 W m −2 with a 90 % uncertainty. Such massive uncertainties are mainly due to our limited knowledge regarding the spatiotemporal distribution, abundance, and compositions of airborne PM (Seinfeld, 2016;Raes et al., 2000). In addition, aerosols may undergo aging during the vertical transport process, causing increasingly complex compositions and changes in aerosol properties. Despite these factors, to date, vertical observations remain comparatively scarce compared to surface measurements. Therefore, to obtain an improved understanding of the fundamental chemical and dynamical processes governing haze development, more field observations of upper-layer aerosols are necessary, as these measurements could provide updated kinetic and mechanistic parameters that could serve to improve model simulations.
Currently, various monitoring approaches have been developed and applied to measure vertical aerosols, e.g., satellite remote sensing and in situ lidar methods; these approaches can be used to obtain the pollution concentration profiles (Van Donkelaar et al., 2016;Reid et al., 2017). To accurately measure chemical compositions, aircraft and unmanned aerial vehicles (UAVs) equipped with a variety of instruments can be utilized in short-term sampling campaigns (Lambey and Prasad, 2021;Zhang et al., 2017), but these tools are unsuitable for long-term continued observations due to their high operational costs. In cases of near-surface vertical urban atmosphere observations, techniques involving tethered balloons, meteorological towers, and skyscrapers are usually adopted (Zhou et al., 2020;Xu et al., 2018;Fan et al., 2021). However, the vertical application range of these methods is limited to only ∼ 500 m, thus hardly meeting the requirements of research conducted above the boundary layer. Therefore, high-elevation mountain sites have long been regarded as suitable places for long-term research on the upper-layer aerosol (including its composition, chemicalphysical properties, formation processes, etc.; Dzepina et al., 2015;Zhou et al., 2021;Wang et al., 2013), which are conducive to better understanding of the haze episodes in the lower troposphere. Although the fixed observation position is the key drawback of this monitoring approach, it has still been widely selected for use in various vertical observation campaigns, e.g., in past studies conducted in Salt Lake Valley (Baasandorj et al., 2017), in Terni Valley (Ferrero et al., 2012, and on Mt. Tai (Meng et al., 2018;. Mt. Hua adjoins the Guanzhong basin of inland China, where haze pollution has been a persistent environmental problem (Wu et al., 2020a(Wu et al., , 2021Wang et al., 2016). In our previous studies conducted at the mountaintop of Mt. Hua, we found that air quality was significantly affected by surface pollution, and distinctive differences were found in the aerosol compositions and size distributions at the mountaintop compared to those measured at lower elevations near ground level Li et al., 2013). With the implementation of strict emission controls, the atmospheric environment in this region has changed dramatically from the SO 2 /sulfate-dominated previous environment to the current NO x /nitrate-dominated environment (Baasandorj et al., 2017;Wu et al., 2020c). However, the fundamental chemical and dynamical processes driving this PM 2.5 -loading explosion are unclear under the current atmospheric state with increasing O 3 and NH 3 levels. To better rationalize these processes, in this work, 4 h integrated aerosol samples were synchronously collected on the mountainside and at the lowerelevation land surface, and the chemical components and stable nitrogen isotope compositions of nitrate and ammonium were analyzed in the collected PM 2.5 samples. We compared the chemical compositions and diurnal cycles between the two sampling sites and then discussed the changes in the chemical forms of secondary inorganic ions during their vertical transport from lower to higher elevations. Our study revealed that nitrate and ammonium exhibited distinct physicochemical behaviors during the aerosol-aging process.

Sample collection
In this campaign, the PM 2.5 samples were synchronously collected at two locations in the Mt. Hua area during the period from 27 August to 17 September 2016. One sampling site was located on a building belonging to the Huashan Meteorological Bureau (34 • 32 N, 110 • 5 E; 400 m above sea level, a.s.l.) at the foot of Mt. Hua. Surrounded by several traffic arteries and dense residential and commercial buildings, as shown in Fig. 1b, this site is an ideal urban station for studying the impacts of anthropogenic activities on local air quality and is referred to hereafter as the MF site. The mountainous sampling site (34 • 29 N, 110 • 3 E; 1120 m a.s.l.) was located approximately 8 km from the city site horizontally (Fig. 1c) at an elevation of 720 m above the average Huashan town level of ∼ 400 m a.s.l. This site was situated on a mountainside that experiences little anthropogenic activity due to its steep terrain and is abbreviated hereafter as the MS site. Furthermore, this location adjoins one of the larger valleys of Mt. Hua; therefore, the measurements taken at this location were strongly affected by the lower-elevation air pollutants transported upwards by the valley winds. At both measurement sites, the PM 2.5 aerosol samples with a 4 h interval in were collected onto prebaked (at 450 • C for 6 h) quartz filters using high-volume (1.13 m 3 min −1 ) air samplers (Tisch Environmental, Inc., USA). All air samplers were installed on the roofs of buildings, approximately 15 m above the local ground surface. Furthermore, size-resolved aerosol sampling was synchronously conducted at two sites during summertime (10-22 August, 2020). These samples, with nine size bins (cutoff points were 0.43, 0.65, 1.1, 2.1, 3.3, 4.7, 5.8, and 9.0 µm, respectively), were collected using an Anderson sampler at an airflow rate of 28.3 L min −1 for ∼ 72 h. After sampling, the filter samples were stored in a freezer (at −18 • C) prior to analysis.
The hourly PM 2.5 , NO x , and O 3 mass concentrations were detected at the mountainside sampling site using an E-BAM, a chemiluminescence analyzer (Thermo Fisher Scientific, model 42i, USA) and a UV photometric analyzer (Thermo Fisher Scientific, model 49i, USA), respectively. At the MF site, only PM 2.5 was monitored, using another E-BAM, while the data of the other species were downloaded from the Weinan Ecological Environment Bureau (http://sthjj.weinan.gov.cn/, last access: 8 July 2021). Meteorological data characterizing both sampling sites throughout the whole campaign were obtained from the Shaanxi Provincial Meteorological Bureau (http://sn.cma.gov.cn/, last access: 8 July 2021).

Chemical analysis
Four punches (1.5 cm diameter) of each aerosol sample were extracted into 10 mL Milli-Q pure water (18.2 M ) under sonication for 30 min. Subsequently, the extracts were filtered with 0.45 µm syringe filters and water-soluble ions were detected (Na + , NH + 4 , K + , Mg 2+ , Ca 2+ , SO 2− 4 , NO − 2 , NO − 3 , and Cl − ) by using ion chromatography. The detection limit for these nine ions was < 0.01 µg mL −1 . A DRI model 2001 thermal-optical carbon analyzer was used, following the IM-PROVE_A temperature protocol to analyze the organic car-bon (OC) and elemental carbon (EC) in each PM 2.5 filter sample (in 0.526 cm 2 punches). For more details regarding the utilized methods, readers can refer to our previous studies (Wu et al., 2020a).
To quantify the stable nitrogen isotope compositions of nitrate (δ 15 N-NO − 3 ) and ammonium (δ 15 N-NH + 4 ) in PM 2.5 samples, the filter samples were pretreated, as described for the water-soluble ion analysis. The ammonium in the extracts (approximately half of the resulting solution) was oxidized by hypobromite (BrO − ) to nitrite (NO − 2 ), which was subsequently reduced by hydroxylamine (NH 2 OH) in a strongly acidic environment. The above product (N 2 O) was then analyzed by a commercially available purge-and-cryogenic-trap system coupled to an isotope ratio mass spectrometer (PT-IRMS). A bacterial method (Pseudomonas aureofaciens, a denitrifying bacterium without N 2 O reductase activity) was used to convert the sample NO − 3 into N 2 O, which was ultimately quantified through PT-IRMS. As revealed in previous studies (Fang et al., 2011), the presence of NO − 2 in aerosols may interfere with the denitrifier method when measuring δ 15 N. Nonetheless, NO − 2 generally composed tiny portions in most of our samples and, on average, contributed < 1.0 % to NO − 3 +NO − 2 . Thus, we believe that the proportion of NO − 2 in the sample was too small to affect the resulting δ 15 N measurements, based on the discussions reported by Wankel et al. (2010). More details regarding the analytical artifact and quality control protocols can be found elsewhere (Wu et al., 2021;Liu et al., 2014).

Concentration-weighted trajectory (CWT) analysis
CWT is a powerful tool used here to reveal the potential spatial sources responsible for the high PM 2.5 loadings measured on Mt. Hua; this method has been used previously in similar studies (Wu et al., 2020c, b). In this study, the CWT analysis was conducted using the Igor-Pro-based tool coupled with hourly PM 2.5 concentrations and 12 h air mass backward trajectories that were simulated by using the Hybrid Single-Particle Lagrangian Integrated Trajectory (HYS-PLIT) model (Petit et al., 2017).

2.4
Theoretical calculations of the partial pressures of NH 3 and HNO 3 and the dissociation constant of NH 4 NO 3 To obtain the product of the partial pressures of NH 3 and HNO 3 , the NH 4 NO 3 deliquescence relative humidity (DRH) was first calculated using Eq. (1). The average DRH of NH 4 NO 3 between the two sites was 65.0 ± 2.9 %, which is slightly lower than the atmospheric RH (66.0 ± 19.3 %).
As the works by Wexler and Seinfeld (1991) and Tang and Munkelwitz (1993) revealed, aerosols are multicomponent mixtures, meaning the aerosol DRH is always lower than the DRH of the individual salts in the particles. Thus, the actual DRH of the aerosols observed in this study would be lower than the calculated DRH of NH 4 NO 3 . Based on these analyses, the particles would be deliquescent most of the time, but for simplification, we always assumed that NH 4 NO 3 was in an aqueous state, corresponding to the following dissociation Reaction (R1): According to the approach illustrated in the referenced work (Seinfeld, 2016), the equilibrium constant of the dissociation reaction can be described as Eq. (2).
where K AN (mol 2 (kg 2 atm 2 ) −1 ) is the equilibrium constant of Reaction (R1) (this value is temperature-dependent and can be calculated by Eq. 3), γ NH 4 NO 3 is the binary activity coefficient for NH 4 NO 3 (γ NH 4 NO 3 = γ NH 4 γ NO 3 ), and m NH + 4 and m NO − 3 are the molalities of NH + 4 and NO − 3 , respectively. To calculate γ NH 4 NO 3 and m NH + 4 m NO − 3 , the activity coefficients of the corresponding ions and the aerosol water content were assessed using the E-AIM (IV) model (http: //www.aim.env.uea.ac.uk/aim/model4/model4a.php, last access: 2 November 2021). Combining Eqs. (2) and (3), we obtained the product of the partial pressures of NH 3 and HNO 3 (P HNO 3 P NH 3 ), obtaining an average of ∼ 15.2 ± 26.0 ppb 2 (where ppb is parts per billion) at the MF site. This value was within the range of values (1.0 to 37.7 ppb 2 ) measured by the in situ Gas and Aerosol Composition monitor (IGAC) in the summer of 2017 in Xi'an, a metropolitan city located in the Guanzhong basin of inland China that has suffered from serious haze pollution (Wu et al., 2020b). Thus, we believe that P HNO 3 P NH 3 variations can be assessed using the above method (to a certain extent). For simplification, the dissociation constant of dry NH 4 NO 3 particles (K p ; ppb 2 ) was applied in this study, which can be calculated as a function of temperature, using Eq. (4), as was revealed by Mozurkewich (1993). Despite not considering the aerosol properties (e.g., acidity and mixing state) that may induce the shift of the NH 4 NO 3 equilibrium states, this assessment Table 1. Mass concentrations of species in the PM 2.5 samples, pH, and the meteorological conditions at the two sampling sites.

Mountain foot Mountainside
(ii) Mass concentration of species and ALWC (µg m −3 ) and pH 14.1 ± 9.5 22.2 ± 12.1 Aerosol liquid water content (ALWC) and pH are predicted by the thermodynamic model (E-AIM (IV)). method was also applied in a similar work conducted by Lindaas et al. (2021).

Results and discussion
3.1 Overview of PM 2.5 at both sites 3.1.1 Meteorological conditions and temporal variations in PM 2.5 concentrations The temporal variations in the 4 h PM 2.5 mass concentrations, water-soluble ions, and meteorological factors measured at the two sampling sites are illustrated in Fig. 2, and the comparisons of the above variables are summarized in Table 1. The average temperature (T ) and relative humidity (RH) at the MF site were 23.2 ± 4.2 • C and 68.9 ± 18.2 % (Table 1), respectively, and these values were characterized by marked diurnal variations, as shown in Fig. 2a. However, relatively cold and moist weather frequently occurred at the MS site, which exhibited less pronounced diurnal T and RH variations, with variations approximately 8 • C and 6 % lower than the mean values derived at the MF site, respectively. Windy weather (wind speed of 3.2±2.0 m s −1 ) also prevailed at this sampling site, with gusts above 10.0 m s −1 ; this condition is conducive to the dissipation of pollutants. Overall, the PM 2.5 concentrations measured at the MF site varied from 22.8 to 245.6 µg m −3 , with a mean value of 76.0 ± 44.1 µg m −3 , approximately corresponding to the Grade II level (75 µg m −3 ) of the National Ambient Air Quality Standard in China. Even so, the PM 2.5 levels at Huashan town (i.e., at the MF site) were still higher than those measured in many typical megacities in the summertime, e.g., Xi'an (37 µg m −3 in 2017; Wu et al., 2020a) and Beijing (46.3 µg m −3 in 2016; Lv et al., 2019). Noticeably, stagnant meteorological conditions with increasing RH (> 77 %) and relatively low wind speeds (< 2.0 m s −1 ) occurred during the relatively late stage of observation, leading to a buildup of high PM 2.5 loadings (78.7 to 245.6 µg m −3 ). Such typical haze events last approximately 4 d (12 to 16 September, 2016), indicating that aerosol pollution is still severe in rural towns, despite the notable air quality improvements recorded in most Chinese urban areas. A similar temporal PM 2.5 pattern was seen at the MS site, where the average PM 2.5 concentration (47.0 ± 38.0 µg m −3 ) was only 0.62-fold of that at the MF site and was within the range of that measured at the summit of Mt. Tai (37.9 µg m −3 in 2016; Yi et al., 2021) and on Mt. Lushan (55.9 µg m −3 in 2011; Li et al., 2015) in the summertime. As shown in Fig. 2d, a multiday episode (mean PM 2.5 of 106.3 µg m −3 ) also appeared at the MS site during the period from 12 to 15 September, corresponding to the days on which high surface pollution was recorded. This was indicative of the potential impacts of surface pollution on air quality in mountainous areas.

Diurnal variation in PM 2.5
As shown in Fig. 2c and d, regular diurnal PM 2.5 variations were seen throughout the whole campaign, especially at the MS site. To reveal the differences in the daily changes in PM 2.5 between the two sampling sites, the mean diurnal cycles of hourly PM 2.5 and the boundary layer height (BLH) are depicted in Fig. 3. At the MF site, the PM 2.5 concentration was moderately enhanced during the nighttime, with a daily maximum (88.2 ± 53.0 µg m −3 ) observed at 06:00 local standard time (LST). After sunrise, PM 2.5 exhibited a decreasing trend until ∼ 15:00 LST, corresponding to thermally driven boundary layer growth. Conversely, the aerosol concentrations at the MS site immediately increased as the boundary layer uplifted in the early morning and peaked at 14:00 LST, when the MS site was located completely within the interior of the boundary layer. Proverbially, anabatic valley winds prevail in mountainous regions during the daytime. Thus, the aerosol-rich air at the MF site may be transported aloft by the prevailing valley breeze, leading to significantly enhanced PM 2.5 levels at the MS site in short time periods. This finding was further verified by the similar diurnal NO 2 pattern identified at the MS site, as illustrated in Fig. S1 in the Supplement. In the forenoon period, continuous enhancement in the NO 2 level was observed at the MS site, with a daily maximum of 14.4 ± 53.0 µg m −3 (at 11:00 LST); this maximum was ∼ 7-fold the early morning NO 2 concentration. However, O 3 exhibited indistinctive variations during this period, and this was indicative of less NO 2 being generated from photochemical reactions. As mentioned above, there are no obvious anthropogenic emission sources around the MS site; therefore, our observations indicate the remarkable transport of pollutants from the lower ground surface to higher elevations during the daytime.
Moreover, the PM 2.5 concentrations at the MS site exhibited less nighttime variation, with a modest abatement (Fig. 3b). The nocturnal BLH usually remained below the elevation of the MS site; thus, the surface PM 2.5 may have contributed less to the aerosol levels at the MS site at night. To identify the potential spatial sources of nocturnal PM 2.5 at the MS site, a high-elevation CWT analysis was conducted. As illustrated in Fig. 4, the CWT values in the daylight hours were mostly concentrated over the sampling site, consistent with our discussion above. However, relatively high nighttime CWT loadings were distributed on Mt. Hua and in its surrounding regions, indicating that regional transport may be a major source of PM 2.5 at the MS site at night. Thus, the constituents and variations in nocturnal PM 2.5 at the MS site may mainly be the results of regional features. To verify the feasibility of vertical transport of air parcel, the WRF-Chem (Weather Research and Forecasting model coupled with chemistry) model was applied here to simulate wind filed and the divergence that represents the expansion rate of the air mass in a unit of time. From Fig. 5a, the southerly winds prevailed at mountain foot area during the whole campaign, which would blow the pollutants into the valley. And these pollutants at low-elevation can be transported to the upper layer by the updrafts, as indicated by the positive values of vertical divergence at the MF area that decreased with enhanced elevation (Fig. 5b). Besides that, we also analyzed the organic compounds in PM 2.5 samples, e.g., levoglucosan, BkF, and IP+BghiP, which are major tracers for the emissions from biomass burning, coal combustion, and vehicle exhausts, respectively (Wang et al., 2009;Wu et al., 2020a;Wang et al., 2007). From Fig. S2, the indistinctive divergences of diagnostic ratios and proportion of these organic tracers were found among both sampling sites, suggesting an insignificant change in the corresponding emission sources during the transport.
3.2 Characterization of water-soluble ions in PM 2.5 3.2.1 Comparisons of water-soluble ions between the two sites Figure 6 shows the fractional contributions of the chemical compositions to the PM 2.5 at both sampling sites. As summarized in Table 1, the water-soluble ion level (WSI; 24.0 ± 15.0 µg m −3 ) at the MF site was comparable to that of organic matter (OM; OM = 1.6 × OC; Wang et al., 2016), with a fractional contribution of ∼ 31 % to PM 2.5 (Fig. 6). At the MS site, the WSI exhibited lower values (19.5 ± 16.0 µg m −3 ), yet the proportion was moderately enhanced by ∼ 6 %. Notably, this elevated contribution of WSIs was mostly attributed to sulfate and ammonium. Similar patterns in which the secondary inorganic ions (sulfate, nitrate, and ammonium; SNA) mass fraction increased with altitude within the mixing height have also been observed in Terni Valley (central Italy; Ferrero et al., 2012) and Salt Lake Valley (USA; Baasandorj et al., 2017). Among the SNA components, sulfate was the predominant species, exhibiting slight mass concentration differences between the two sampling  sites (10.1±6.4 µg m −3 versus 9.0±7.1 µg m −3 ). However, a ∼ 4 % enhancement in the mass fraction of sulfate was measured at the MS site. Ammonium also exhibited a similar feature, accounting for ∼ 5 %-7.5 % of the PM 2.5 . These sulfate and ammonium mass concentration homogeneities across the two sites were indicative of the further formation of these two ions during transport. Unlike sulfate and ammonium, nitrate and its proportions showed opposite trends, decreasing with elevation; this was consistent with most of the measured components. The above variation features of SNA among two sites were found for most of the time periods in the campaign, except for 12-13 September, which showed a higher SNA concentration at the MS site ( Fig. 2e and f). On these 2 d, the MS site remained outside the boundary layers (a.s.l.; ∼ 550 m), suggesting less of an effect of the surface pollutants on the aerosol upper layers, while the precursor masses (∼ 12.3 µg m −3 for SO 2 and 8.4 µg m −3 for NO 2 ) were insufficient to form so much SNA at the MS site. Thus, the higher SNA aloft on the above 2 d may mostly be driven by regional or long-range transport. This can be verified by the CWT analysis, for which high loadings were mainly distributed in the western and southwestern areas of Mt. Hua (Fig. S3a), and the cities on the air mass transport pathways (e.g., Xi'an and Weinan) also suffered from moderate haze pollution on these 2 d (Fig. S3b). On account of the different sources of PM 2.5 between these 2 d and the remaining periods, the samples during 12-13 September were excluded in the following discussion, whereas the residual SNA data still exhibited similar variations, as mentioned above, and the divergence in nitrate mass concentration and fractional contribution to PM 2.5 among two sites even became more pronounced (Fig. S4). Moreover, distinct nitrate size distributions were also observed between the different sites in the summertime of 2020 (Fig. S5). From Fig. S5, we can note that the nitrate at a low elevation was enriched in the fine mode, with a minor peak in the coarse fraction. However, the high-elevation nitrate exhibited a bimodal pattern, with two equivalent peaks in the fine and coarse fractions, and was well correlated with coarse-mode calcium but poorly correlated with ammonium (R 2 = 0.51, p < 0.05). To our knowledge, ammonium nitrate, a major form of fine-mode particulate nitrate, can be easily volatilized and converted into gasphase NH 3 and HNO 3 (Pakkanen, 1996;Harrison and Pio, 1983). Thus, the gaseous HNO 3 volatilized from fine PM may react with coarse-mode cations (e.g., Ca 2+ , Mg 2+ , and Na + ) to form nonvolatile salts, leading to that significant nitrate shifts from fine particles to large particles. A similar phenomenon was also found in our previous study conducted at the summit of Mt. Hua . Nonvolatile sulfate was predominantly found in the fine fraction at both sampling sites, which may support this concept. More evidence for this hypothesis is presented below in Sect. 3.3.
The diurnal cycles of the 4 h sulfate, nitrate, and ammonium are illustrated in Fig. S6. As shown in Fig. S6, the total SNA concentration at the MF site exhibited a morning peak from 08:00-12:00 LST; this variation was quite different from that of PM 2.5 . Such a difference between the total SNA and PM 2.5 at the MF site could partially be attributed to the lower sampling resolution and enhanced formation of SNA in the morning. The diurnal total SNA pattern identified at the MS site coincided with the PM 2.5 pattern, exhibiting a daily maximum reaching ∼ 21.2 ± 19.9 µg m −3 (from 12:00-16:00 LST), which is a 1.2-fold increase compared to that measured at the MF site. Among the SNA components, morning peaks of nitrate and ammonium (from 08:00-12:00 LST) were also observed at the MF site. These nitrate and ammonium concentrations at the MF site can contribute to those at the MS site through vertical transport, leading to a significant enhancement in the nitrate and ammonium concentrations aloft, with afternoon peaks during 12:00-16:00 LST. Even so, the maximum nitrate concentration at the MS site (6.5 ± 7.4 µg m −3 ) was still lower than that measured at the MF site (8.9 ± 6.8 µg m −3 ) due to the NH 4 NO 3 volatilization under the transport process, while ammonium exhibited the opposite trend. This finding was consistent with the above discussion. Unlike nitrate and ammonium, similar diurnal variations in sulfate were observed between the two sampling sites, with daily maxima observed from 12:00-16:00 LST at both sites. The major sulfate formation pathway during the daytime in summer is the photooxidation of SO 2 with an OH radical, and the formation rate facilitated by this process is much lower than that of the nitrate formation process (Seinfeld, 2016;Rodhe et al., 1981). Thus, sulfate formation may occur continuously during vertical transport, leading to smaller difference in the diurnal cycle of sulfate between the two sites.

Chemical forms of SNA at both sites
As shown in Fig. 6, the water-soluble ions considered here mainly included sulfate, nitrate, and ammonium, which usually exist in the form of ammonium salts (NH 4 HSO 4 , (NH 4 ) 2 SO 4 , NH 4 NO 3 , and so on). In the H 2 SO 4 -HNO 3 -NH 3 thermodynamic system, H 2 SO 4 and HNO 3 are neutralized by NH 3 under ammonia-rich conditions and mainly exist as (NH 4 ) 2 SO 4 and NH 4 NO 3 in aerosols. Conversely, H 2 SO 4 is converted to HSO − 4 in environments with relatively low NH 3 availabilities. Thus, NH 4 HSO 4 and NH 4 NO 3 may be the dominant aerosol components under such environmental conditions (Rodhe et al., 1981;Seinfeld, 2016). To reveal the major SNA forms at the different sampling sites considered herein, a theoretical ammonium concentration was calculated according to thermodynamic equilibrium with the atmospheric sulfate and nitrate levels. The theoretical ammonium levels were calculated as follows: where [SO 2− 4 ] and [NO − 3 ] represent atmospheric concentrations (µg m −3 ). When (NH 4 ) 2 SO 4 and NH 4 NO 3 are the dominant species, the NH + 4 theory can be calculated using Eq. (5). In contrast, Eq. (6) suggests that NH 4 HSO 4 and NH 4 NO 3 are abundantly present in the analyzed aerosols. Figure 7 compares the measured NH + 4 concentrations with the theoretical NH + 4 concentrations derived by the two equations above. As illustrated in Fig. 7a, the slope of the observational NH + 4 values against the theoretical NH + 4 values calculated using Eq. (6) was much closer to unit at the MF site than at the MS site, meaning that NH 4 HSO 4 and NH 4 NO 3 were the major chemical forms of SNA at the MF site. However, the opposite pattern was revealed at the MS site; thus, the upper aerosols were characterized by abundant (NH 4 ) 2 SO 4 and NH 4 NO 3 . We also found that the diurnal variations in the SNA chemical forms in PM 2.5 , which we mainly concerned in this study, were insignificant at both sampling sites. Based on observational data collected during the 2020 summertime, the NH 3 level at the MF site (36.0±68.0 ppb) was ∼ 9-fold that at the MS site (4.1 ± 2.5 ppb). Under such an abundant NH 3 environment, the S(VI) was in the major form of NH 4 HSO 4 , but appeared as (NH 4 ) 2 SO 4 in a relatively low NH 3 environment, which was somewhat unexpected. As can be inferred from earlier studies (Seinfeld, 2016;Shi et al., 1999), the NH 3 Henry's law coefficients generally increase in value as the temperature decreases. Therefore, the lower temperatures measured at the MS site would create a more favorable en-vironment for ammonia, thus shifting its partitioning toward the particulate phase. The HSO − 4 transported from the MF site would thus be further neutralized to SO 2− 4 by this additional ammonium during transport, leading to the significant difference observed in the chemical forms of SNA between the two sites. Moreover, as the chemical component changed from NH 4 HSO 4 to (NH 4 ) 2 SO 4 , the aerosol acidity moderately decreased, showing a higher bulk PM 2.5 pH (3.4 ± 2.2) at relatively clean upper layer and a lower value (2.9±2.0) at heavily polluted grounds (Table 1). However, previous studies generally recognized that the aerosol would become more acidic when the air parcels were transported from polluted to cleaner/more remote regions (Liu et al., 1996;Nault et al., 2021). As shown in Table 1, the change in aerosol liquid water content (ALWC) has an indistinctive difference among both sampling sites (t test; p > 0.05). Thus, we think that such a reduced aerosol acidity with increasing elevation in our study was mainly due to the change in the chemical component, which was caused by the different physicochemical behaviors of the semi-volatile species of nitrate and ammonium during transport. More discussion is included in the following section.

Physicochemical behaviors of nitrate and ammonium during transport
According to the above discussion, a conceptual model illustrating the physicochemical behaviors of nitrate and ammonium during vertical transport was proposed to explain the chemical composition differences between the two sites. As shown in Fig. 8, surface air parcels containing abundant NH 4 HSO 4 and NH 4 NO 3 particles can be transported to the upper atmosphere by the prevailing valley winds, and during this process, the volatile NH 4 NO 3 is easily converted to gaseous NH 3 and HNO 3 . Subsequently, heterogeneous reactions of the gaseous HNO 3 with fugitive dust occur, thus forming nonvolatile salts and resulting in the accumulation of nitrate on the coarse-mode particles. However, as the temperature decreased, the NH 3 that volatilized from the fine Figure 7. Comparison of the calculated and observed NH + 4 concentrations at both sampling sites. particles or was derived from the surface could re-enter the particulate phase through the gas-particle partition. Therefore, (NH 4 ) 2 SO 4 would be formed in the aerosol phase and would gradually replace NH 4 HSO 4 .
To investigate the likelihood of NH 4 NO 3 volatilization during the transport process, the dissociation constant of NH 4 NO 3 (K p ) and the partial pressures of gas-phase NH 3 and HNO 3 were calculated in this study. More details regarding the calculation steps of the above factors can be found in Sect. 2.4. Based on the thermodynamic principles presented by Stelson and Seinfeld (1982), when the product of the partial pressures of NH 3 and HNO 3 (P HNO 3 × P NH 3 ) is greater than K p , then the equilibrium of the system shifts toward the aerosol phase, thus increasing the NH 4 NO 3 formation. In contrast, a relatively low P HNO 3 × P NH3 /K p value (< 1) suggests that the NH 4 NO 3 dissociation is induced and that NH 4 NO 3 is transferred to the gas phase. Figure 9 depicts the ratio of the product of the partial pressures of NH 3 and HNO 3 with different ambient temperatures. As shown in Fig. 9, approximately 85 % of the samples collected at both sampling sites were located within the region with P HNO 3 × P NH 3 /K p of less than 1, demonstrating a common NH 4 NO 3 dissociation phenomenon during the observed period. For the samples with P HNO 3 × P NH 3 /K p ratios < 1, the mean value of the MS site ratios was approximately half that of the MF site ratios, indicating that the NH 4 NO 3 dissociation may be more likely at higher elevations than that at lower elevations. This finding was inconsistent with the aircraft observations of the wildfire smoke plumes collected by Lindaas et al. (2021), who revealed that P HNO 3 × P NH 3 /K p exhibited an increasing trend within 1-3 km (a.s.l.). As we already know, the abundant NH 3 and NO x can be emitted by the wildfire, which would be transported aloft and lead to a higher NH 3 and HNO 3 mixing ratio compared to that at lower elevation. This may drive a higher P HNO 3 × P NH 3 /K p ratio at the upper layers of aircraft observations in the western U.S. (Lindaas et al., 2021).
Moreover, the nitrogen isotope compositions of nitrate and ammonium in PM 2.5 that can retain invaluable information regarding physicochemical processes (Wiedenhaus et al., 2021;Elliott et al., 2019) were thus measured to further verify the conceptual model. As previously mentioned, unlike daytime pollutants, nocturnal pollutants exhibited different sources between the two sampling sites. Thus, their nitrogen isotope compositions were more complicated and less comparable. However, for simplicity, only the daytime samples were analyzed here, based on the hypothesis that the sources of the high-elevation pollutants were the same as those of the pollutants collected at the MF site. As shown in Fig. 10, a discrepancy in the δ 15 N value of nitrate (δ 15 N-NO − 3 ) featuring more 15 N-enriched NO − 3 was observed at the MS site, with a p value of less than 0.05. This finding can be ascribed to the evaporation of a portion of the particulate NH 4 NO 3 due to a dissociation shift in equilibrium; in this shift, the lighter 14 N was preferentially incorporated into the atmosphere, leading to 15 N enrichment in the remaining nitrate. A similar phenomenon was also revealed by Wiedenhaus et al. (2021), who thought that the ammonium nitrate dissociation may be an important reason for the accumulation of 15 N in aerosol particles. Additionally, Freyer et al. (1993) revealed that gas-phase isotopic exchanges between NO and NO 2 result in the enrichment of the heavier 15 N isotope in the more oxidized form and may further affect δ 15 N-NO − 3 through nitrate formation reactions. The above isotopic exchange between NO 2 and NO x can be roughly described as follows: where K and f NO 2 are the temperature-dependent exchange constant and mole fraction of NO 2 , respectively. Based on trace gas observations, the f NO 2 values of the air aloft were very high due to the frequently undetectable NO concentration, indicating a rather limited isotopic exchange between NO 2 and NO. Therefore, the evaporation of particulate NH 4 NO 3 has been a significant factor affecting the measurement of a higher δ 15 N-NO − 3 at the MS site than at the MF site in our observations. According to the above analysis, the ammonium at the MS site should theoretically be more and more enriched in δ 15 N with the continuous NH 4 NO 3 volatilization. However, our observation of δ 15 N-NH + 4 did not correspond to the above pattern, namely the ammonium at the MS site depleted in δ 15 N compared to that at the MF site (p < 0.05; Fig. 10). Given the unchanged NH 3 sources, as verified in Sect. 3.1.2, such seemingly unreasonable observations were mainly caused by the gas-to-particle conversion of ammonia. In this process, the reversible phaseequilibrium reactions between NH 3 (g) and HNO 3 (g)/HCl(g) would yield positive enrichment the in δ 15 N of aerosol NH + 4 (Walters et al., 2019); nevertheless, unidirectional reactions involving NH 3 (g) and SO 2− 4 /HSO − 4 favored 15 N depletion in the particle form, as revealed by Heaton et al. (1997). Thereby, the lower δ 15 N-NH + 4 values at the MS site were mostly driven by those irreversible reactions, rather than the reversible equilibrium ones. This result further confirmed our  conjecture that the additional NH 3 would partition into particulate phases and further neutralize the acidic NH 4 HSO 4 , leading to an increasing pH at the MS site compared to that at the MF site. Taken together, this compelling evidence verifies that fine-mode nitrate and ammonium exhibit distinctly different physicochemical behaviors during their transport.

Conclusions and atmospheric implications
In this study, aerosol samples were collected at 4 h intervals on the mountainside of Mt. Hua, and the OC, EC, watersoluble ions and isotope compositions of nitrate and ammonium were measured and compared with simultaneous observations taken at a lower-elevation site (MF site). The particle mass at the MF site was approximately 1.5-fold that at the MS site, and distinctly different diurnal cycles were observed between the two sampling sites. Based on the BLH variation, we revealed that near-surface PM 2.5 could be transported to the upper layers by the mountain-valley breeze, leading to the gradual accumulation of pollutants on the mountainside during the daytime.
Sulfate, the predominant species found among ions at both sampling sites, exhibited nearly identical mass concentrations at the two sites but had a moderately enhanced mass fraction at the MS site. Such homogeneity was also observed in ammonium, which mainly existed as NH 4 HSO 4 + NH 4 NO 3 and (NH 4 ) 2 SO 4 + NH 4 NO 3 at the MF and MS sites, respectively. This observation indicated the further formation of ammonium during the transport process. Unlike sulfate and ammonium, nitrate at the MS site exhibited abated trends in both its concentration and proportion, mainly due to the volatilization of NH 4 NO 3 . With the help of nitrate and ammonium nitrogen isotopes, we proposed a conceptual model to illustrate the different behaviors of nitrate and ammonium during vertical transport; in this model, the semi-volatile NH 4 NO 3 in surface air parcels was easily converted into gaseous NH 3 and HNO 3 . Subsequently, heterogeneous reactions occurred between the gaseous HNO 3 and fugitive dust, forming nonvolatile salts and leading to a significant nitrate shift from fine particles to coarse particles. In addition, the decreasing temperature was favorable for NH 3 partitioning toward the particle phase, and the addition of ammonium further neutralized HSO − 4 to form SO 2− 4 . This process would reduce the aerosol acidity, with bulk PM 2.5 pH increasing from 2.9 ± 2.0 at the MF site to 3.4 ± 2.2 at the MS site.
Over the past decade, the relative abundance of NH 4 NO 3 has been enhanced in most urban areas of China because strict emission directives have been promulgated to abate the emission and environmental impacts of SO 2 (Xie et al., 2020;Song et al., 2019). In this work, we observed that NH 4 NO 3 volatilization was a ubiquitous phenomenon for particles during transport, resulting in a shift in part-wise nitrate from the fine mode to the coarse fraction; this shift has also been reported in the offshore areas of the UK (Yeatman et al., 2001). Thus, we think that considering only fine-fraction nitrate may result in the conversion rate of NO x to nitrate being partly underestimated at some times, especially in the summer. Moreover, the deposition velocity of coarse particles is usually faster than that of fine particles; therefore, the above process would appreciably elevate the deposition of N into the environment. Indeed, abundant NO 2 , O 3 , and NH 3 co-occurrence is common in the East Asian atmosphere, and under these conditions, secondary inorganic aerosols can be effectively produced, leading to a PM 2.5 loading explosion in the urban atmosphere of China (Wu et al., 2020c;Wang et al., 2016). Given this, harmonious reductions in NO 2 , O 3 , and NH 3 will be urgent in further mitigation strategies to improve air quality and alleviate other potential effects. Data availability. Meteorological data and hourly PM 2.5 , NO 2 , and O 3 mass concentrations are obtained as documented in Sect. 2 and are available at https://doi.org/10.5281/zenodo.7413640 (Wu, 2022). The data of the other species used in this study are available by contacting the corresponding author.
Author contributions. GW designed the experiment. CW, JiaL, and CC collected the samples. CW and CC conducted the experiments. CW and GW performed the data interpretation and wrote the paper. All authors contributed to the paper with useful scientific discussions or comments.

Competing interests.
The contact author has declared that none of the authors has any competing interests.
Disclaimer. Publisher's note: Copernicus Publications remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Acknowledgements. This work has been financially supported by the National Natural Science Foundation of China (grant nos. 42130704 and 42007202), Shanghai Science and Technology Innovation Action Plan (grant no. 20dz1204000), and ECNU Happiness Flower program. We thank Lang Liu from School of Public Policy and Administration, Northwestern Polytechnical University, Xi'an, China, for his support of model simulation in meteorological data during the campaign.
Financial support. This research has been supported by the National Natural Science Foundation of China (grant nos. 42130704 and 42007202) and the Science and Technology Innovation Plan Of Shanghai Science and Technology Commission (grant no. 20dz1204000).
Review statement. This paper was edited by Xavier Querol and reviewed by three anonymous referees.