Effect of ammonia on fine-particle pH in agricultural regions of China: comparison between urban and rural sites

Particle acidity is a fundamental property that affects atmospheric particulate chemistry. Synchronous online monitoring was performed at two urban sites, Zhengzhou (U-ZZ) and Anyang (U-AY), and three rural sites, Anyang (R-AY), Xinxiang (R-XX), and Puyang (R-PY) in Henan Province, during a haze episode to investigate the pH value and its driving factors in the agricultural regions of China. The pH values of particles calculated by ISORROPIA-II model at rural sites were slightly higher than those at urban sites, with the median (min–max) values of 5.2 (4.8–6.9, R-PY), 5.1 (4.7–6.5, R-AY), 4.9 (4.1–6.8, R-XX), 4.8 (3.9– 5.9, U-AY), and 4.5 (3.8–5.2, U-ZZ). Sensitivity tests show that TNHx (total ammonium, gas+ aerosol) followed by total sulfate were the most important factors that influenced the predicted pH. Generally, particle pH increased with a cation increase and decreases in anions, temperature, and relative humidity. Similar pH values (∼ 3.0) at the required NHx concentrations for the five sites indicated that the presence of excess NHx was likely important for the lower acidity of PM2.5 during the severe haze episodes in this region. Moreover, the concentrations of excess NHx may drive the higher pH values at rural sites, because of the higher excess NHx concentrations at rural sites than those at urban sites. The underlying influence of regional transport on local particle pH cannot be neglected, as it differs the chemical components of PM2.5 and meteorological conditions. Air masses transported from rural and agricultural regions may enhance the particle pH value in urban aerosols given the high pH of particles and high ammonia levels. These results suggest that ammonia is urgently needed to be involved in the regional strategy for the improvement of air quality in China.

Direct measurements on particle pH are challenging because of the small size and nonideality of chemical species in solvated aerosols. Therefore, thermodynamic models, such as E-AIM (http://www.aim.env.uea.ac.uk/aim/aim.php, last access: 2 October 2019) and ISORROPIA-II (http:// isorropia.eas.gatech.edu, last access: 2 October 2019) (Clegg et al., 1998;Nenes et al., 1998), which rely on the measurements of particulate and gaseous species, are widely used in estimating particle pH. Generally, the global distribution of fine particulate matter (PM 2.5 , aerodynamic diameter ≤ 2.5 µm) pH is bimodal with a population of particles having a mean pH of 1-3 and another population having an average pH closer to 4-5 (Pye et al., 2020). Table 1 shows S. Wang et al.: Effect of ammonia on fine-particle pH in agricultural regions of China that PM 2.5 in mainland China was moderately acidic, with pH values that ranged from 3.4 to 5.7 Guo et al., 2017;Liu et al., 2017;Shi et al., 2017Shi et al., , 2019Song et al., 2018;H. Wang et al., 2019), which were 3-5 units higher than those reported in other regions, such as Hong Kong, Singapore, USA, and Greece (Behera et al., 2013;Bougiatioti et al., 2016;Guo et al., 2015Guo et al., , 2016Pathak et al., 2004).
High atmospheric ammonia (TNH x , gas NH 3 plus particle NH 4 + ) is a dominant factor that drives the high PM 2.5 pH in megacities of China because it suppresses the production of particle hydronium (Cheng et al., 2016;Wang et al., 2016). Liu et al. (2017) argued that excess NH 3 and elevated aerosol water content (AWC) were responsible for the high pH in Beijing. Nevertheless, Guo et al. (2017) and Song et al. (2018) demonstrated that high levels of ammonia did not increase the PM 2.5 pH into a fully neutralized condition in Beijing and Xi'an, China. Weber et al. (2016) calculated that only large increases in NH 3 together with sulfate reductions can lead to an increase in pH. Correspondingly, a pH higher by about 1 unit than urban aerosols was observed in locations of intensive agriculture with high NH 3 concentrations in the southeastern USA (Nah et al., 2018). The primary sources for ammonia include agricultural emissions, such as livestock waste, N-fertilizer application, and biomass burning, as well as traffic and industrial emissions (Huang et al., 2012;Shen et al., 2011;. Field studies have reported that elevated NH 3 and NH 4 + concentrations were recorded at rural sites in the North China Plain (Meng et al., 2011(Meng et al., , 2018Shen et al., 2011;Wen et al., 2015). Therefore, studying in an agriculturally developed region is needed to obtain insight into the effect of ammonia on particle pH. Unfortunately, previous studies mainly concentrated on calculating the particle pH in the megacity of China, and few studies have focused on the agricultural regions of China. In addition to ammonia, particle pH can also be influenced by other chemical compositions and meteorological conditions, such as sulfate, AWC, temperature (T ), and relative humidity (RH). Sulfate is often the main acid component of aerosols, and largely determines the acidity of PM 2.5 . Sensitivity tests in Beijing suggested that sulfate, TNH x , and T were the common driving factors, and Ca 2+ and RH were the unique factors in special seasons . The pH sensitivity based on the 10-year record in Canada showed that chemical compositions had various effects on particle pH under different meteorological conditions; moreover, careful examination for any particular region is needed (Tao and Murphy, 2019).
Henan Province is situated in central China; it has a dense rural population and is a top-ranking province in China in terms of agricultural production and chemical fertilizer consumption (National Bureau of Statistics, 2016). The NH 3 emission inventory for Henan Province showed that approximately 1031.6 Gg NH 3 was released in Henan in 2015, thereby contributing to approximately 10 % of China's to-tal emissions (approximately 10 Tg) (Huang et al., 2012;. Livestock waste and N fertilizer application were major sources for ammonia emissions, which may increase TNH x concentrations at rural sites than in urban sites. Furthermore, Henan Province is a severely PM 2.5polluted region in China. In January 2018, a large-scale and long-lasting haze episode that caused PM 2.5 concentration to reach 400 µg m −3 occurred in this region (S. Wang et al., 2019b). An experiment was performed in two urban and three rural sites in Henan Province to investigate particle acidity and its driving factor. ISORROPIA-II model was utilized to estimate PM 2.5 pH using a high-time-resolution dataset. The novelty of the work addressed in the present study is that this study is the first on PM 2.5 acidity that compares urban and rural sites in the agricultural regions of China. The specific objectives of the study are as follows: (1) estimation and comparison of the PM 2.5 pH at the five monitoring sites, (2) identification of the factors that determine the pH, and (3) discussion of the effect of ammonia on pH. Our results are helpful to understand the factors that better determine particle acidity.

Site descriptions
Field sampling was synchronously conducted from 12 to 24 January 2018 at five sites (i.e., two urban sites located in the center of Zhengzhou (U-ZZ) and Anyang (U-AY), and three sites located in the rural areas of Anyang (R-AY), Xinxiang (R-XX), and Puyang (R-PY)). Locations of the five sites ( Fig. S1 in the Supplement) with brief descriptions are listed in Table 2. U-ZZ site is surrounded by busy roads, and one freeway is located 3 km to the south and one freeway is located 7 km to the east. Moreover, this site is near a coal-fired power plant, a gas-fired power plant, and several small-scale industries, such as pharmaceutical companies, electronic factories, and equipment manufacturing. U-AY site is surrounded by busy roads, and the Anyang steelworks is located 8 km to the west. R-AY site is surrounded by farmland and is 1 km west of the Jinggang'ao freeway and 1 km north of a belt freeway. R-XX and R-PY sites are surrounded by farmland without other prominent anthropogenic emission sources.

Instrumentation
The hourly mass concentrations of water-soluble inorganic ions (WSIIs) in PM 2.5 , such as NH 4 + , SO 4 2− , NO 3 − , Cl − , Na + , Mg 2+ , Ca 2+ , and K + and their gaseous precursors (i.e., NH 3 , HNO 3 , and HCl) were measured using an ambient ion monitor (URG-9000D, Thermal Fisher Scientific, USA) at the U-ZZ site and monitor for aerosols and gases (MARGA, Metrohm, Switzerland) at other sites. Both instruments have been successfully deployed in several other field experiments (Li et al., 2017;Shi et al., 2017;S. Wang et al., 2019b), and detailed information is available elsewhere (Markovic et al., 2012;Rumsey et al., 2014). As a brief summary, ambient air is drawn into the systems at a flow rate of 16.7 L min −1 . Subsequently, particles and gases are collected by two aerosol sample collectors through a wet parallel plate or wet rotating denuder. Aqueous solution samples are quantified by using two ion chromatography analyzers. Detection limits for all species during the sampling periods were less than 0.1 µg m −3 . Measurement uncertainties were assumed to 10 % for URG-9000D here by reference to 10 % of Markovic et al. (2012) and 12 % in Hu et al. (2014). For the MARGA instrument, Song et al. (2018) took 20 % as overall relative uncertainties for major species, and Rumsey et al. (2014) reported that the analytical biases were lower than 10 % for SO 4 2− , NO 3 − , and HNO 3 , but higher than 15 % for NH 3 and NH 4 + . Therefore, overall uncertainties were estimated to be 20 % for NH 3 and NH 4 + and 10 % for other species. The combined uncertainties for TNH x were calculated to be 14 % and 23 % for U-ZZ and other sites, respectively, and − + HNO 3(g) ) and TCl (Cl (aq) − +HCl (g) ) for all sites. Because of the complex interference during sampling periods at the five sites, such as the slope from the internal calibration, the solution volume, temperature, pressure, and airflow rate, the actual uncertainties might be different from the estimated value. Hourly elemental carbon (EC) and organic carbon (OC) concentrations in PM 2.5 were simultaneously obtained by the semi-continuous carbon analyzers (Model 4, Sunset Laboratory Inc, USA) in the same place, except for the U-ZZ site, of which data were provided by the Department of Environmental Protection of Henan Province. Detailed information on this instrument can be found in Panteliadis et al. (2015). Briefly, PM 2.5 collected by the device was oxidized to carbon dioxide and analyzed by a nondispersive infrared detector. The relative measurement uncertainty for this instrument was estimated to be 20 % Healy et al., 2013). Meteorological parameters, including T , RH, wind direction, and wind speed, were recorded by smart weather sensor (LUFFT-WS500, Sutron Corporation, Germany), with the accuracies of ±0.2 • C, ±2 %, < 3 • , and 0.1 m s −1 , respectively.

pH prediction
Particle pH was estimated using the ISORROPIA-II thermodynamic model. Input data, including RH, T , concentrations of K + , Ca 2+ , Mg 2+ , TNH x , total sulfate (TH 2 SO 4 , replaced by observed SO 2− 4 ), total sodium (TNa, replaced by observed Na + ), TCl, and TNO 3 , were used to calculate the particle hydronium ion concentration per volume of air (H + air ) and particle water associated with inorganics (AWC inorg ) by computing the equilibrium composition for the Na + -K + -Ca 2+ -Mg 2+ -NH 4 + -SO 4 2− -NO 3 − -Cl − -H 2 O aerosol system. Considering that forward mode is less sensitive to measurement error than the reverse mode and high RH levels were recorded in sampling periods, ISORROPIA-II was run in the forward model for the aerosol system in the metastable condition Guo et al., 2015;Hennigan et al., 2015). Moreover, we excluded data with RH < 30 %, a condition where predicted-to-observed partitioning of nitrate between the gas and particle phase are weakly correlated, and thus uncertainties in the pH prediction are expected to be high Guo et al., 2016). Aerosol pH was calculated according to the following formula (Bougiatioti et al., 2016): where the modeled concentrations for AWC inorg and H + air are µg m −3 , and AWC org is the particle water associated with the organics predicted using the following method: where m s is the mass concentrations of organic matter (OC× 1.6), ρ s is the organic density (1.35 g cm −3 ), and k org is the organic hygroscopicity parameter (0.06) .

pH uncertainty
Two extreme scenarios were evaluated to estimate the pH uncertainty based on the measurement uncertainties of input data (Murphy et al., 2017). Sensitivity tests of pH to input data (see Sect. 3.3) suggest that simulated pH increase with the cation concentrations (i.e., TNH x , TNa, K + , Ca 2+ , and Mg 2+ ) increasing as well as the anion concentrations (i.e., TH 2 SO 4 , TNO 3 , and TCl), T , and RH decreasing. Moreover, the pH value has a positive correlation with OC concentrations via Eqs. (1) and (2). Therefore, cation and OC concentrations were adjusted up to within their maximum positive uncertainties; anions, RH, and T were adjusted down within their maximum negative uncertainties, which represented the pH max case; for the pH min case, cations and OC concentrations were adjusted down, and anions, RH, and T were adjusted up. Figure 1 suggests that pH max cases lead to 6 % (slope − 1) errors for all sites, and pH min cases can result in approximately 10 % deviations. Thus, pH uncertainties were set to be −9 to +6 %, −13 to +6 %, −12 to +6 %, −11 to +6 %, and −10 to +5 % for U-ZZ, U-AY, R-AY, R-XX, and R-PY sites, respectively.

Model validation
The reliability of pH estimation using the ISORROPIA-II thermodynamic model depends on several assumptions, such as equilibrated gas and particle phases (Pye et al., 2020). Thus, the predicted and observed semi-volatile species (e.g., NH 3(g) /NH 4(aq) + , HCl (g) /Cl (aq) − , and HNO 3(g) /NO 3(aq) − ) are compared in Fig. S2 in the Supplement. Observed and predicted NH 4 + , Cl − , and NO 3 − exhibit significant correlations, have correlation coefficients (r) above 0.95 and slopes near 1 at the five sites. In addition, NH 3 concentrations are in good agreement (r > 0.95) with slopes between 0.89 (U-ZZ) and 1.13 (R-PY). These results suggest the excellent performance of ISORROPIA-II for modeling these species. The differences in the slopes of NH 4 + (Fig. S2b in the Supplement) and NH 3 (Fig. S2e in the Supplement) between U-ZZ and R-PY sites were probably attributed to the unbalanced charge of input WSIIs (see Text S1 in the Supplement for more details of calculation) with average equivalent ratios ( Fig. S3 in the Supplement) of 0.99 ± 0.13 (U-ZZ) and 1.20 ± 0.12 (R-PY). However, Song et al. (2018) argued that ion balance was not a key factor for pH calculation in the forward mode, because the forward-mode calculations account for additional constraints imposed by the partitioning of semi-volatile species. To verify this, we adjusted the measured NH 4 + concentration, which was most sensitive to pH modeling (see Sect. 3.3), to fit the ion balance, and reran the pH calculation. As shown in Fig. S4 in the Supplement, re-Figure 1. pH uncertainties of the five sites based on two extreme scenarios of measurement uncertainty. Cations and organic carbon concentrations (OC) were adjusted up to within their maximum positive uncertainties; anions, relative humidity (RH), and temperature (T ) were adjusted down within their maximum negative uncertainties, which represented the pH max case; for the pH min case, cations and OC concentrations were adjusted down, and anions, RH, and T were adjusted up. The color scale bar represents adjusted RH.
predicted NH 4 + and NH 3 concentrations have better performance (r = 0.99, 0.94 < slopes < 1.06) with input concentrations at two sites than before. However, re-calculated pH values change slightly, and these deviations are included in the range of pH uncertainty discussed in Sect. 2.4. Therefore, the little differences in NH x partitioning between the model and measurement among the five sites were acceptable for pH calculation.
Correlations between the predicted and measured HNO 3 and HCl are weak. Similar problems were found in the northeast USA and Beijing. These discrepancies were potentially due to measurement uncertainties brought about by low gas concentrations, the interference of coarse-mode particles, non-volatile cation measurement artifacts, uncertainties in the thermodynamic constants, and kinetic limitations to mass transfer , Haskins et al., 2018Pye et al., 2020;Liu et al., 2017). Pye et al. (2020) suggested that ISORROPIA-II yields a mean activity coefficient of (H + , Cl − ) that may result in a higher predicted HCl concentration. HNO 3 can be partition to both fine and coarse modes, thereby affecting predicted fine-mode nitrate concentrations (Nah et al., 2018). The best semi-volatile species for evaluation of pH modeling depend on the fraction of the gas phase. In this work, most HNO 3 and HCl are concentrated in the particle phase, and thus they are not suited to test the model .
The PM 2.5 pH was also calculated by the E-AIM (Version IV) model to evaluate the performance of ISORROPIA-II using the observed data (RH > 60 %) of the U-ZZ site as an example. Close correlation (r = 0.89) is found between two models with a slope of 0.95 (Fig. S5 in the Supplement). The pH values in ISORROPIA-II are 0.46±0.15 units higher than those in E-AIM. These values are comparable to the founding by Liu et al. (2017) and Song et al. (2018), which is possibly due to the differences of activity coefficient values between the ISORROPIA-II and E-AIM models (Pye et al., 2020). Overall, the predicted pH values using ISORROPIA-II are effective in this work.

Haze episodes
During the sampling periods, five monitoring sites simultaneously experienced a long-lasting and large-scale haze episode. Time series of the concentrations of major species (i.e., NH 3 , NH 4 + , SO 4 2− , and NO 3 − ) and meteorological parameters at the U-ZZ site as an example are presented in  Table 3. Three study cases were classified on the basis of similar meteorological conditions and WSIIs levels at the five sites. The WSIIs concentration, T , and RH gradually increased in the southern wind during Case 1 (12-14 January). Case 3 (21-25 January) was characterized by decreased WSII concentrations, T , and RH with the northeastern wind. Even though the total WSIIs (TWSIIs) concentrations were comparable, the chemical components of WSIIs were various between Cases 1 and 3 (Table 3). In particular, NO 3 − , NH 4 + , and NH 3 concentrations in Case 1 were higher than those in Case 3 at all sites. Conversely, SO 4 2− concentrations in Case 1 were lower than those in Case 3. C.  reported that the southern cities of Henan Province (e.g., Nanyang, Shangqiu, Zhoukou, and Zhumadian) had relatively higher ammonia emissions than the cities involved in this study. Moreover, the northeastern air masses from the Jing-Jin-Ji regions were easily enriched with sulfate (S. Wang et al., 2019a. Evidently, back-trajectory frequency analysis (see Text S2 in the Supplement for more details of the discussion) confirmed that sampling sites during Cases 1 and 3 ( Fig. S7a and c in the Supplement) were predominantly influenced by the southern and northeastern air masses, respectively.
The pH values (Table S2 in the Supplement) of the five sites during Cases 2 and 3 were comparable but slightly lower than those during Case 1. As shown in Fig. S9 in the Supplement, predicted pH values present better correlations with H + air concentrations rather than AWC. Particle hydronium ion aqueous concentration depends on both the presence of ions and the amount of particle AWC . Moreover, H + air was closely associated with the NH 3 mixing ratios, and higher NH 3 always corresponded to lower H + air . High correlations (r > 0.5) between H + air and TWSIIs (Table S3 in the Supplement) imply that local pH was significantly affected by the TWSIIs levels, and probably resulted in the high acidity during Case 2. H + air was also correlated with individual chemical species (e.g., TNH x , TH 2 SO 4 , TNO 3 , and TCl). Therefore, the diversity of pH in Cases 1 and 3 may be partly owing to the different proportions of particle-and gas-phase constituents that will be discussed in detail below. In addition, high correlations between RH and H + air (Table S3 in the Supplement) suggest that meteorological conditions have major effects on particle pH. Given the Case 2 was less affected by regional transport, the diurnal patterns of median pH values of the five sites (Fig. 4) indicate that pH values during nighttime were 0.3 (R-PY) to 0.5 (U-ZZ) units higher than those during the daytime. Similar results were also found in other cities (e.g., Beijing, Tianjin, southern Canada, and the USA) (Battaglia et al., 2017;Guo et al., 2015;Murphy et al., 2017;Shi et al., 2019), resulting from the diurnal trends of T and RH (Fig. 4f). After sunrise, high T facilitated the dissociation of particle-phase ammonium (e.g., NH 4 NO 3 ) and led to a rapid loss of AWC Saraswati et al., 2019).

Sensitivity tests of pH
To explore the dominant factors that determine the local particle pH level and resulting in the high pH at rural sites, sensitivity tests of pH to chemical species (i.e., TNH x , TH 2 SO 4 , TNO 3 , TCl, TNa, K + , Ca 2+ , and Mg 2+ ) and meteorological parameters (i.e., T and RH) were performed. Firstly, the real-time measured values of a variable (e.g., TNH x ) and average values of other parameters (i.e., TH 2 SO 4 , TNO 3 , TCl, TNa, K + , Ca 2+ , Mg 2+ , T , and RH) during Case 2 were input into ISORROPIA-II to investigate the sensitivity of local pH to this variable (i.e., TNH x ) , and results are listed in Table 4 and Fig. S10 in the Supplement. In addition, a given range for a variable for all sites with corresponding average values of other parameters was simulated to compare its effects on pH among five sites (Fig. 5 and    (Table S4 in the Supplement), which aims to better reflect the actual observation conditions. The degree of sensitivity was represented as the relative standard deviation (% RSD) of the re-calculated pH values. RSD calculates the absolute value of the coefficient of variation, which helps us to determine how small or large the standard deviation is when compared to the mean of the re-calculated pH data set. Therefore, when the re-calculated pH data set has a higher RSD value, this variable can change the pH more easily and is more important for local pH.
As shown in Table 4, the most important factor that influenced particle pH during Case 2 was TNH x , followed by TH 2 SO 4 at the five sites, and the U-ZZ site was also affected by TNO 3 . The acidity of PM 2.5 is governed by the phase partitioning of semi-volatile gases such as NH 3 , HNO 3 , and HCl (Pye et al., 2020). (NH 4 ) 2 SO 4 , NH 4 HSO 4 , and NH 4 NO 3 , which are mainly formed through neutralization reactions between NH 3 with SO 4 2− and HNO 3 successively according to the regime of local ammonia, are the most abundant components of PM 2.5 in winter haze in this region (S. Wang et al., 2019b;Pathak et al., 2009). Therefore, particle pH was driven by TNH x , TH 2 SO 4 , and TNO 3 . TNa, TCl, and crustal ions (i.e., K + , Ca 2+ , and Mg 2+ ) have less influence on the predicted pH values, because these species were at low concentrations, together accounting for lower than 10 % of TWSIIs. However, since the low volatility of these cations allows them to preferentially neutralize sulfates over NH 3 , the effects of crustal dust and sea spray on particle pH cannot be ignored when the mass fraction of these ions are high in typical pollution events (e.g., sandstorms) or areas (e.g., coasts) (Allen et al., 2015;Guo et al., 2018;Vasilakos et al., 2018). In addition to chemical species, T was more sensitive to local pH than RH at the five sites, which is similar to the finding in Beijing and Canada Tao and Murphy, 2019). This result can be explained by the remarkable influence of T on the solubility and dissociation constants of the partitioning of NH 3(g) /NH 4(aq) + (Hennigan et al., 2015). Figure 5 and Fig. S11 in the Supplement compare the sensitive degrees of pH to input data among different sites. Except for TNH x , the sensitivities of pH to TH 2 SO 4 , TNO 3 , and other ions at urban sites were more significant than those at rural sites, particularly at U-ZZ sites of 7.2 % and 14.8 % of RSD to TH 2 SO 4 and TNO 3 , respectively. To gain insight into the differences of pH sensitivity among the five sites, sensitivity tests of pH to TH 2 SO 4 , TNO 3 , and TNH x were calculated using the fixed TNO 3 (67.5 µg m −3 ) and TH 2 SO 4 (36.5 µg m −3 ) concentrations under the average meteorological conditions of five sites (i.e., T = 2.5 • C and RH = 60 %). These chosen values were close to the average values of the five sites during Case 2 (i.e., 36.4 ± 15.4 µg m −3 for SO 4 2− , 67.5±23.5 µg m −3 for TNO 3 , 2.5±1.5 • C for T , and 59.3 ± 14.0 % for RH), and other chemical species were set to be zero. As shown in Fig. 6, sensitivities of pH to TH 2 SO 4 and TNO 3 increase with the decrease in TNH x concentration, particularly when the TNH x concentrations are lower than 60 and 40 µg m −3 , respectively. Therefore, the pH changes tended to become more sensitive to TH 2 SO 4 and TNO 3 at the U-ZZ site ( Fig. 5e and f), with the lowest TNH x concentrations (46.8±14.7 µg m −3 ). All these results may be explained by the presence of excess ammonia that will be discussed below.
Sensitivity tests suggest (Fig. 5, Figs. S10 and S11 in the Supplement) that particle pH gradually grows with increased cation and decreased anion concentrations. Specifically, the TNH x concentration that increased from 25 to 90 µg m −3 can promote particle pH by 3.5 (U-ZZ) to 4.5 (R-AY) units. The crustal cations and sea spray have limited effects on pH, but a 10-fold increase in these species still can increase predicted pH values by about 0.1 units. TH 2 SO 4 and TNO 3 , which increased from 10 to 80 µg m −3 and 1 to 125 µg m −3 , can reduce the pH values by 1.5 (R-PY) to 4.0 (U-ZZ) and 0.2 (R-AY) to 1.4 (U-ZZ) units, respectively. In addition, a 20 • C (−5 to 15 • C) and 65 % (30 % to 95 %) increase drops the pH by approximately 1.3 and 2.7 units at the five sites, respectively. Corresponding to the observed data during Case 2 (Table 3), lower TNH x concentrations and higher T values were probably responsible for the higher acidity at urban sites than those at rural sites. Moreover, Fig. 6 shows that particle pH increase with decreasing TH 2 SO 4 (Fig. 6a) or increasing TNO 3 (Fig. 6b), particularly when the TNH x is at high concentration. These results indicate that predicted pH will be enhanced by increasing the TNO 3 / TH 2 SO 4 ratio, which is consistent with the observation in Beijing (Xie et al., 2020). All these may be due to the elevated ammonia partitioning with nitrate fraction increase, and nitrate-rich particles would absorb more water compared to the sulfate-rich particles (Xie et al., 2020). Based on the above, higher TNH x (Table 3) and TNO 3 /TH 2 SO 4 ratios (2.9 ± 0.6 and 1.7 ± 0.6 for Cases 1 and 3, respectively) during Case 1 may lead to higher pH values than those during Case 3, even though the TWSIIs levels were comparable. In the long run, the decreasing sulfate concentration in PM 2.5 accompanied with increasing nitrate concentration has been recorded during haze episodes in China, because strong actions were taken to reduce the coal consumption in recent years Wang et al., 2017). Considering the ammonia in the atmosphere of northern China might still be increasing (Liu et al., 2018), future acidity of PM 2.5 during winter haze episodes is expected to become increasingly less acidic in this area. A more careful analysis is needed to test this inference, however, because fine-mode particle remains a relatively constant pH of 0-2 despite a 70 % reduction in sulfates in the USA (Vasilakos et al., 2018;Weber et al., 2016). , TNH x (NH 3(g) + NH 4(aq) + ), TNO 3 (HNO 3(g) + NO 3(aq) − ), and TH 2 SO 4 (replaced by observed SO 4 2− ) among the five sites. A given range for a variable for all sites with corresponding average values of other parameters was used as input to the ISORROPIA-II model. The range of the x axis is close to the observed minimum and maximum values in Case 2 ( Table S4 in the Supplement). The color scale bar represents the pH values. The relative standard deviation (% RSD) and range (Range) represent the sensitivity degree of pH to this variable and range (min-max) of the re-predicted pH value in the test, respectively. The square plots on the graph represent the average values of each variable observed in Case 2 with standard deviation as an error bar. Table 4. Sensitivity of pH to input data. The real-time measured values of a variable and the average values of other parameters during Case 2 were input into the ISORROPIA II. The degree of sensitivity was represented as the relative standard deviation (% RSD) of the re-calculated pH values; higher RSD implied higher sensitivity of this factor to pH and thus is more important for local pH.

Effects of ammonia on pH
It has been suggested that unusually high levels of NH 3 can increase pH: on average, a 5-fold to 10-fold increase in the NH 3 levels leads to a 1-unit change in pH (Nah et al., 2018;Weber et al., 2016;Guo et al., 2017). To examine the effects of major indicators of ammonia (i.e., TNH x , Required-NH x , and Excess-NH x ; see Text S3 in the Supplement for more details of calculation) on aerosol acidity, particle pH was calculated by using a wide range of TNH x (25-130 µg m −3 ) and average values of other parameters during Case 2 (Ta-ble 3) of each site. Simultaneously, the concentrations of TNH x , Required-NH x , Excess-NH x , and corresponding pH values are illustrated in Fig. 7. An "S-curve" growth trend of pH with increasing TNH x appears with the inflection point around the Required-NH x concentration. In the case when the input TNH x concentration is lower than the Required-NH x , the growth rate of pH increases with increasing TNH x ; otherwise, the pH growth flattens out, which is similar to the findings of Weber et al. (2016) and Bougiatioti et al. (2016). In the process of increasing NH 3 concentration, NH 3 reacted with SO 4 2− and HNO 3 successively, during which large amounts of H + were consumed and pH values rapidly increased. Theoretically, the Required-NH x plus nonvolatile cations can neutralize all anions regardless of phase partitioning and bisulfate formation, and thus changes in TNH x around the Required-NH x concentrations have a significant impact on particle pH. Subsequently, dissolving Excess-NH x into the particles became difficult, and pH values increased slowly Seinfeld and Pandis, 2016). Therefore, the underlying reason why TNH x concentrations lower than 60 and 40 µg m −3 elevated the pH sensitivities (Fig. 6) to TH 2 SO 4 and TNO 3 is that the Excess-NH x is exhausted under the case of TH 2 SO 4 and TNO 3 that ranged from 10 to 100 µg m −3 , respectively. On the other hand, the pH values of the five sites at the Required-NH x concentrations are near 3.0, which coincides with the pH value of Beijing in the no-excess-NH x cases . Therefore, the presence of Excess-NH x in the aerosol likely was an important factor in the lower acidity of PM 2.5 , and it might have enhance pH values by 1.5 (U-ZZ) to 2 (R-PY) units. Considering that the Excess-NH x concentrations at rural sites (e.g., 30.1 ± 6.2 µg m −3 at R-PY) were higher than those at urban sites (e.g., 14.8 ± 4.1 µg m −3 at U-ZZ), Excess-NH x concentrations may drive the pH values of rural sites higher than those of urban sites.
3.5 Implications of regional transport From the above discussion, the differences of pH among three cases indicate that the underlying influence of regional transport on local particle pH cannot be neglected, as it differs the chemical components of PM 2.5 and meteorological conditions. In particular, the median pH values (Table S2 in the Supplement) of Case 1, during which the air masses were transported from the south of sampling regions, increased by 0.2-0.9 units over those during Cases 2 and 3. Aqueous formations of sulfate are strongly dependent on particle pH. Cheng et al. (2016) reported that the aqueous-phase sulfate production rates from NO 2 and O 3 oxidation of SO 2 had a positive correlation with particle pH during the Beijing haze events. When pH exceeded approximately 4.5 (higher than this value at rural sites in this work), NO 2 oxidation dominated the sulfate formation, and its reaction rate increased by 1 order of magnitude with the rise of pH by 1 unit. Thus, air masses transported from rural and agricultural regions may promote the sulfate formation in urban aerosols. In addition to sulfate, higher aerosol pH favors partitioning of TNO 3 toward aerosol NO 3 rather than gaseous HNO 3 , thus elevating the particle mass concentration (Nenes et al., 2020;Weber et al., 2016). Sampling regions in this study are located on a transport route for Beijing (MEP, 2017), thereby frequently affecting its local particle pH. The lifetimes of NH 3 (1-5 days or less) and NH 4 + (1-15 days) in the atmosphere are sufficient for transport to Beijing during a haze episode (Aneja, 2000;Lefer et al., 1999;Warneck, 1988).  have also observed that PM 2.5 pH from the southwest direction was generally higher than that from the northern direction in Beijing. Therefore, the particle pH in Beijing may be enhanced when southern air masses from this region accompany elevated-pH particles and high ammonia levels. Figure 7. (a-e) Particle pH corresponds to increasing TNH x at the five sites, used to examine the effects of major indicators of NH 3 (i.e., TNH x , Required-NH x , and Excess-NH x ) on aerosol acidity. Particle pH was calculated by using a wide range of TNH x (25-130 µg m −3 ) and average values of other parameters during Case 2 of each site. The concentrations of TNH x , Required-NH x , and Excess-NH x with corresponding pH values are marked by a hollow box, hollow circle, and arrow, respectively. The yellow and blue background colors correspond to the NH x -poor and NH x -rich regions, respectively.

Conclusions
An experiment was performed using a series of high-timeresolution instruments in two urban (i.e., U-ZZ and U-AY) and three rural sites (i.e., R-AY, R-XX, and R-PY) in Henan Province during a large-scale and long-lasting haze episode. The ISORROPIA-II model was used to investigate the pH value and its driving factors. PM 2.5 exhibited moderate acidity with median pH values of 4.5 (3.8-5.2), 4.8 (3.9-5.8), 4.9 (4.1-6.2), 5.1 (4.7-6.3), and 5.2 (4.8-6.5,) at U-ZZ, U-AY, R-XX, R-AY, and R-PY, respectively. The pH values at rural sites were slightly higher than those at urban sites.
The predicted pH values of PM 2.5 were significantly affected by the WSIIs levels, different proportions of particleand gas-phase constituents, and meteorological parameters. Sensitivity tests show that TNH x , followed by TH 2 SO 4 , were the most important factors influencing the predicted pH at the five sites. In addition, T was more sensitive to local pH than RH. Generally, particle pH rose with the increase in cation and the decrease in the anion, T , and RH. Further study suggests that predicted pH will be enhanced by increasing the TNO 3 /TH 2 SO 4 ratio. Therefore, elevated TNH x concentrations and lower T were probably responsible for the higher pH values at rural sites than those at urban sites, and higher TNH x and TNO 3 /TH 2 SO 4 ratios during Case 1 may lead to higher pH values than those during Case 3. An S-curve growth trend of pH to increasing TNH x was found with the inflection point around the Required-NH x concentration. Moreover, the pH values of the five sites at the Required-NH x concentrations are near 3.0, and the presence of Excess-NH x in the aerosol can enhance pH values by 1.5 (U-ZZ) to 2 (R-PY) units, and thus Excess-NH x was likely an important factor leading to the lower acidity of PM 2.5 during the severe haze episodes in this region. Considering that the Excess-NH x concentrations at rural sites were higher than those at urban sites, Excess-NH x concentrations may lead the pH values of rural sites to be higher than those of urban sites.
The underlying influence of regional transport on local particle pH cannot be neglected, as it differs the chemical components of PM 2.5 and meteorological conditions. Air masses transported from rural and agricultural regions with elevated pH particles and high ammonia levels may promote secondary particle formation in urban aerosols. Therefore, ammonia should be considered in the regional strategy for improving air quality in China.
Data availability. All data in this work are available by contacting the corresponding author Shasha Yin (shashayin@zzu.edu.cn).
Author contributions. SY and RZ designed and led this study. SY was responsible for all observations and data collection. LW, YL, CW, and WW interpreted the data and discussed the results. SW wrote the paper.
Competing interests. The authors declare that they have no conflict of interest.