Online measurements were conducted at four urban and six rural sites in Henan Province, China, from 1 January to 29 February 2020, including Sanmenxia (U-SMX), Zhoukou (U-ZK), Zhumadian (U-ZMD), and Xinyang (U-XY), as well as rural locations including Anyang (R-AY), Xinxiang (R-XX), Jiaozuo (R-JZ), Shangqiu (R-SQ), Nanyang (R-NY), and Puyang (R-PY). Descriptions and the spatial distribution of these 10 sites can be found in Table S1 and Fig. S1 in the Supplement.

An aerosol and gas monitor (MARGA, Metrohm, Switzerland) was used to analyze the hourly water-soluble ions (Na+, NH4+, K+, Mg2+, Ca2+, Cl-, NO3-, and SO42-) in PM2.5, as well as gaseous species (NH3, HNO3, HCl, and HONO), at 10 sampling sites. The MARGA instrument is widely used (Chen et al., 2017; Stieger et al., 2019; Twigg et al., 2022). A detailed description of the instrument and QA/QC can be found in Sect. S1 in the Supplement. In brief, the atmospheric sample passes through a PM2.5 cut-off head, and both particles and gases enter a wet rotating dissolution device for diffusion. Subsequently, the particles in the sample undergo hygroscopic growth and condensation in an aerosol supersaturated vapor generator, followed by collection and ion chromatographic analysis. The gases in the sample are oxidized by H2O2 in the dissolution device and absorbed into a liquid solvent. They then entered the gas sample collection chamber for ion chromatographic quantification. The range of minimum detection limits for water-soluble ions was between 0.002 µgm-3 (Cl-) and 0.081 µgm-3 (NH4+). Uncertainties of 20 % are assumed for the detection of NH3 and NH4+, while uncertainties of 10 % are assumed for other components (Wang et al., 2020; Wang et al., 2022). In addition, a detailed description of HONO measurement using this system can be found in Sect. S2. Overall, the limit of detection for HONO was 4 pptv, and the uncertainty was estimated to be ± 20 %.

The data for NO2 and SO2 were obtained from a series of instruments provided by Thermo Fisher Scientific (USA). The hourly concentrations of organic carbon (OC) in PM2.5 were analyzed using a carbon analyzer (Model 4, Sunset Laboratory, USA). Detailed descriptions of the NO2, SO2, and carbon analyzers can be found in Sect. S3. The smart weather stations (LUFFTWS500, Sutron, Germany) were utilized for synchronized observation of meteorological parameters including pressure, temperature (T), and relative humidity (RH).

The temporal variations of NH3, NH4+, and TNHx at 10 sampling sites in the pre-COVID-19 pandemic period (PC; 1 to 23 January 2020) and during the COVID-19 pandemic period (DC; 24 January to 29 February 2020) are presented in Fig. , with their average concentration listed in Table . In general, rural sites exhibited higher concentrations of NH3, NH4+, and TNHx compared to urban sites, except for the R-NY site. This finding is consistent with previous studies conducted in Zhengzhou (Wang et al., 2020), Shanghai (Chang et al., 2019), and Quzhou (S. Feng et al., 2022), owing to the intense agricultural ammonia emissions. The highest concentrations of NH3 and TNHx were recorded at site R-JZ, with average values of 25.3 ± 11.5 and 40.8 ± 20.1 µgm-3, respectively. Site R-AY had the highest NH4+ concentration, measuring 19.3 ± 12.9 µgm-3. Note that the current study area exhibited higher NH3 levels compared to other regions (Table S6), which probably was attributed to the highest NH3 emissions of Henan Province in China, primarily from nitrogen fertilizer application and livestock farming (Wang et al., 2018; Ma, 2020).

Compared to the PC, NH3 concentrations increased in the DC at all sites. Notably, rural sites experienced more significant increases in NH3 concentrations than urban sites, which was similar to the trend in Shanghai (Xu et al., 2022). The largest increases in NH3 concentrations were observed at R-SQ (71 %, 7.3 µgm-3) and U-ZK (37 %, 4.8 µgm-3) for rural and urban sites, respectively. In contrast, the concentrations of NH4+ and TNHx decreased in the DC, with the largest reduction at the rural site, R-PY (51 %, 12.9 µgm-3), and the urban site, U-ZMD (48 %, 9.3 µgm-3). Regarding TNHx, rural sites exhibited larger reductions, with site R-SQ experiencing the largest decrease of 37 % (4.7 µgm-3).

Figure  illustrates the spatial distribution and the diurnal variation of NH3 concentrations at the 10 sites before and during the pandemic. NH3 concentrations in most sites exhibited a unimodal trend in the PC that NH3 concentrations gradually increased after sunrise, reaching a peak around noon (11:00–12:00 GMT+8), and then decreased to a trough around 16:00–17:00 (GMT+8). This diurnal pattern is similar to NH3 variations observed in urban areas of Houston, USA, as a result of the natural emissions from vegetation and soil during photosynthesis (Gong et al., 2011). However, other studies have recorded a significant NH3 peak during the early morning between 08:00–10:00 (GMT+8) (Ellis et al., 2011; Meng et al., 2018; Gu et al., 2022), suggesting the influence of vehicle emissions (Gong et al., 2011; Gu et al., 2022), residual NH3 mixing, soil or plant emissions (Ellis et al., 2011), and dew volatilization (Wentworth et al., 2016; Huang et al., 2021b). Therefore, the NH3 in urban and rural areas of this study was probably less affected by NH3 emissions from vehicles, different from the recent studies in megacities of China (e.g., Beijing and Shanghai) (Gu et al., 2022; Wu et al., 2023; Y. Zhang et al., 2020). In addition to the transport from agricultural emissions, urban NH3 in this region might also originate from other non-agricultural sources, such as wastewater treatment, coal combustion, household waste, urban green spaces, and human excrement (Chang et al., 2019).

During the COVID-19 pandemic, the diurnal variation of NH3 in both urban and rural sites still maintained a unimodal distribution. The peak values in urban sites remained consistent with PC levels, further demonstrating that the influence of vehicles on NH3 in urban areas was limited. Notably, the peak time of NH3 in rural sites shifted 1–2 h earlier compared to the trend in the PC. Ammonia in rural areas primarily originates from nitrogen fertilizer application, livestock, and poultry breeding (T. Feng et al., 2022; Meng et al., 2018), which are significantly influenced by T and RH (Liu et al., 2023). Table S7 and Fig. S5 reveal that there was an increased T and a decreased RH at rural sites in the DC than the PC, which could accelerate the evaporation of NH3 and thus potentially lead to earlier peak NH3 concentrations.

The increased NH3 accompanying decreased NH4+ in the DC suggests that the gas–particle partitioning of NH3/NH4+ may determine the elevated NH3 concentrations. Meteorological parameters, including RH and T, play a crucial role in the gas–particle partitioning of NH3 (Liu et al., 2023; Xu et al., 2020). Therefore, the higher T and lower RH in the DC (Table S7 and Fig. S5) favored the conversion of NH4+ to NH3, resulting in a decrease in ε(NH4+)([NH4+]/([NH3]+[NH4+]) compared to those in the PC (Table S7).

NH3 primarily enters particles to neutralize acidic ions (Wang et al., 2020; Xu et al., 2020; Liu et al., 2017; Ye et al., 2011; Wells et al., 1998). Accordingly, the concentrations of required ammonia (Required-NHx) and excess ammonia (Excess-NHx) were calculated based on the acidic substances as follows (Wang et al., 2020): 4Required-NHx=17×([SO42-]48+[NO3-]63+[Cl-]35.5+[HNO3]64+[HCl]36.5)-17×([Na+]23+[K+]39+[Ca2+]20+[Mg2+]12),5Excess-NHx=TNHx-Required-NHx, where [W] represents the concentration of the substance (µgm-3). The significant linear fitting (R2 is greater than 0.96, and the slope is close to 1) in Fig. S6 demonstrates that the anions and cations at each site were close to the equilibrium state. Therefore, the organic acids in PM2.5 may have less effect on NH3 and NH4+ and were not considered in Eq. ().

As shown in Fig.  and Table S8, compared to those in the PC, the concentration of Required-NHx in the DC significantly decreased (ranging from 37 % at site R-JZ to 58 % at site R-PY), while the concentration of Excess-NHx increased (ranging from 9 % at site R-AY to 78 % at site R-SQ). The reduction in the concentrations of sulfate and nitrate (Fig. S7) was responsible for the decrease in the concentration of Required-NHx. To sum up, in addition to meteorological conditions, the substantial reduction in anthropogenic emissions of SO2, NOx, and other pollutants in the DC led to a decrease in acidic substances (e.g., sulfate and nitrate) in particles, in turn resulting in more gas-phase NH3 concentration remaining in the atmosphere.

Diurnal patterns of particle pH in the PC and DC at 10 sites are summarized in Fig. , with their average values listed in Table S9. PM2.5 shows consistent moderate acidity, with mean values in the range of 4.2–5.1, which were close to the values in previous studies (Table S9). Compared to the PC, the particle pH at 10 sites increased significantly in the DC, with the highest increase of 0.5 (U-ZK) and 0.3 (R-PY) at urban and rural sites, respectively; this is the subject of an in-depth discussion in the following text.

To explore the dominant factors that determine the local particle pH level and result in the high pH during the DC, sensitivity tests of pH to chemical species (i.e., TNHx, TH2SO4, TNO3, TCl, TNa, K+, Ca2+, and Mg2+) and meteorological parameters (i.e., T and RH) were performed. A given range for a variable (i.e., TNHx) with corresponding average values of other parameters (i.e., TH2SO4, TNO3, TCl, TNa, K+, Ca2+, Mg2+, T, and RH) was input into the model and simulated to compare its effects on pH. As shown in Fig. S8, pH increases with an increase in the cation concentrations (i.e., TNHx, Na+, K+, Ca2+, and Mg2+) as well as a decrease in the anion concentrations (i.e., TH2SO4, TNO3, and Cl-), T, and RH. According to the average values of input data during the PC (blue line in Fig. S8) and DC (red line in Fig. S8) at U-ZK and R-PY sites, respectively, the changes in pH (ΔpH in Fig. ) indicate that the decrease in TNHx concentration and the increase in T in the DC led to a decrease in pH values (ΔpH: 0.09 at U-ZK and 0.08 at R-PY sites) compared to the PC. However, this effect was outweighed by the decrease in TH2SO4 (ΔpH: 0.07 and 0.8 at U-ZK and R-PY sites, respectively) and TNO3 (ΔpH: 0.05 and 0.4 at U-ZK and R-PY sites, respectively) concentrations as well as the increase in K+ (ΔpH: 0.03 at U-ZK and 0.2 at R-PY site) and Mg2+ (ΔpH: 0.01 at U-ZK and 0.04 at R-PY site) concentrations in the DC, resulting in an overall increase in pH values in the DC. Furthermore, the relationship between particle pH with the concentrations of Required-NHx and Excess-NHx, which considers all chemical components, is investigated to examine the dominant factor on the increasing pH in the DC. As shown in Fig. , the higher Excess-NHx concentrations in the DC led to higher increases in pH values (ΔpH: 1 at U-ZK and 0.5 at R-PY site) than those in the PC (ΔpH: 0.3 at U-ZK and 0.2 at R-PY site); thus Excess-NHx concentrations may be the key factor in promoting the pH values.

The observed HONO concentrations decreased by 18 % and 54 % at U-ZK (0.8 ppb) and R-PY (0.9 ppb) sites in the DC, respectively, compared to those (1.0 and 2.2 ppb) in the PC. Moreover, all the known HONO production sources rates including Pemi, POH+NO, Pground, Pground+hv, Paerosol, Paerosol+hv, and Pnitrate (Fig. , Figs. S9 and S10) show a decreasing trend from the PC to DC, with total reductions of 38 % (from 30 % to 45 % in the scenario with the minimum and maximum uncertainty, respectively) and 79 % (from 77 % to 82 % in the scenario with the minimum and maximum uncertainty, respectively) for U-ZK and R-PY, respectively. At the U-ZK, Pground+hv decreased the most (84 %), while at the R-PY, Pnitrate had the largest decrease of about 85 %, which was speculated to be related to the decrease in NOx and NO3- concentration in the DC. Note that the reduction rates in the overall known source and almost individual sources were greater than the reduction rates in HONO concentrations (Figs.  and ); thus we hypothesized that there should be other sources capable of promoting HONO production.

There were positive correlations between HONO with SO2, Excess-NHx, SO42-, and pH (Fig. S12) indicating that Reaction () might form an amount of HONO and contribute to less reduction in the observed HONO concentrations. Considering that Reaction () mainly reacts in the liquid phase, the calculated reaction rates of Reaction () under the conditions of RH > 60 % in the PC and DC periods are illustrated in Figs.  and S12. Despite the decrease in NO2 and SO2 concentrations in the DC, the increase in particle pH, increasing HSO3- concentration in the aqueous phase, promoted the Reaction () reaction rates by 58 % and 59 % at U-ZK and R-PY (Fig. ), respectively. Consequently, the enhanced Reaction () might prevent a large decrease in HONO (18 % at U-ZK and 53 % at R-PY) under the conditions of a significant reduction in vehicle emissions and a decline of 66 % and 69 % in NO2 concentrations at U-ZK and R-PY, respectively.

According to sensitivity tests of pH (Fig. S8) and Reaction () (Fig. S12), pH increases with an increase in the concentrations of cations (TNHx, TNa, K+, Ca2+, and Mg2+) and OC as well as a decrease in the anion concentrations (TH2SO4, TNO3, and Cl-), T, and RH. Reaction () is positively correlated with AWC, NO2, SO2, pH, and pressure and negatively correlated with T. Therefore, two extreme scenarios (i.e., the maximum and minimum rate scenarios) were evaluated to estimate the uncertainty of pH and Reaction () based on the measurement uncertainties at the U-ZK and R-PY sites. Figure S13 suggests that the two extreme scenarios can lead to -10 % to 7 % and -71 % to 125 % uncertainties at the U-ZK site and -10 % to 7 % and -78 % to 123 % uncertainties at the R-PY site for pH and Reaction (), respectively. Even considering the above uncertainty in Fig. , it can still be observed that during the DC period, the decrease in HONO was less than that of NO2, and the rate of the Reaction () increased.

Considering the conclusions of this study are based solely on observational data, there are certain limitations. For example, only the changes in the Reaction () reaction of PM2.5 were calculated, without considering variations in components, pH values, and Reaction () reaction rates of coarse particles. Additionally, although this study selected scenarios with RH >60 % to calculate Reaction () to ensure the presence of a liquid phase, there might still be some errors in calculating Reaction (). Furthermore, due to thermodynamic model calculations of pH values, changes in the mixed state of particle components, and the omission of organic acids, alongside the absence of gaseous HNO3 and HCl in this study, these factors may lead to inaccuracies in pH value simulations and uncertainty in Reaction () calculations (Pye et al., 2020; Haskins et al., 2018; Nah et al., 2018). Therefore, there is a certain degree of uncertainty in the conclusions regarding the growth of Reaction () reactions in this paper. Nevertheless, by calculating the changes in the reactions in Reaction (), this study provides a possible explanation for the relatively small decrease in HONO during the epidemic period.