The significant contribution of HONO to secondary pollutants during a severe winter pollution event in southern China

Nitrous acid (HONO) can strongly affect atmospheric photochemistry in polluted regions through the production of hydroxyl radicals (OHs). In January 2017, a severe pollution episode occurred in the Pearl River Delta (PRD) of China, with maximum hourly PM2.5, ozone, and HONO levels reaching 400 μg m−3, 150 ppb, and 8 ppb, respectively, at a suburban site. The present study investigated the sources and processes generating such high HONO concentrations and the role of HONO chemistry in this severe winter episode. Four recently reported HONO sources were added to the Community Multiscale Air Quality (CMAQ) model, including RH-dependent (relative humidity) and light-enhancing effects on heterogeneous reactions, photolysis of particulate nitrate in the atmosphere, and photolysis of HNO3 and nitrate on surfaces. The revised model reproduced the observed HONO and significantly improved its performance for O3 and PM2.5. The model simulations showed that the heterogeneous generation on surfaces (with RH and light effects) was the largest contributor (72 %) to the predicted HONO concentrations, with the RH-enhancing effects more significant at nighttime and the light-enhancing effects more important in the daytime. The photolysis of total nitrate in the atmosphere and deposited on surfaces was the dominant HONO source during noon and afternoon, contributing above 50 % of the simulated HONO. The HONO photolysis was the dominant contributor to HOx production in this episode. With all HONO sources, the daytime average O3 at the Heshan site was increased by 24 ppb (or 70 %), compared to the simulation results without any HONO sources. Moreover, the simulated mean concentrations of TNO3 (HNO3+ fine particle NO−3 ) at the Heshan site, which was the key species for this haze formation, increased by about 17 μg m−3 (67 %) due to the HONO chemistry, and the peak enhancement reached 55 μg m−3. This study highlights the key role of HONO chemistry in the formation of winter haze in a subtropical environment.


Introduction
Nitrous acid (HONO) can significantly affect atmospheric photochemistry through its photolysis producing hydroxyl radicals (OHs) (Reaction R1) and subsequent reactions of OH with other gases (Alicke et al., 2003;Kleffmann et al., 2005). OH radicals oxidize volatile organic compounds (VOCs) and convert nitric oxide (NO) into nitrogen dioxide (NO 2 ) without consuming ozone (O 3 ), leading to the generation of O 3 (Reactions R2 to R6). The oxidation of oxides of nitrogen (NO x = NO+NO 2 ), sulfur dioxide (SO 2 ), and VOCs by OH and O 3 also produce secondary aerosols, which are the key components of haze (e.g., Pathak et al., 2009Pathak et al., , 20112 X. Fu et al.: The significant contribution of HONO Cao et al., 2012 The impact of HONO on atmospheric photochemistry varies under different environmental conditions. In general, the effect of HONO is more significant in polluted conditions than clean conditions. For example, calculations constrained by field measurements of HONO have suggested that the contributions of HONO photolysis to the daytime HO x (HO x = OH + HO 2 ) production can reach 56 % in an urban area (Ren et al., 2003) and 87 % at a roadside (Yun et al., 2017), which are higher than the contributions in rural and forested areas (30 % to 40 %) (Acker et al., 2006;Kleffmann et al., 2005). Simulations by chemistry transport models which considered major sources of HONO showed that the maximum enhancements of O 3 concentrations due to HONO were mostly less than 10 ppb in the US and other Western countries (Sarwar et al., 2008;Czader et al., 2012;Li et al., 2010;Goncalves et al., 2012), but more significant impacts have been reported in China due to more intense NO x and VOC emissions. For example, the reported maximum hourly O 3 enhancement can be more than 30 ppb in Beijing (Li et al., 2011;Xu et al., 2006) and up to 25 ppb in Hong Kong . The previous studies mostly focused on summertime. Limited attention has been paid to the winter season when the observed HONO concentrations can also be high (Hou et al., 2016;Li et al., 2018a;Xu et al., 2015). The source and formation mechanisms of HONO are still not fully understood. Most previous studies suggest that a heterogeneous reaction of NO 2 on the surface is dominant, especially at night (Li et al., 2012;Zhang et al., 2016). In the daytime, the long-known gasphase reaction of NO and OH explains less than 10 % of the daytime HONO production (Sarwar et al., 2008;Li et al., 2010;Zhang et al., 2016). Other daytime sources include direct traffic emissions Liang et al., 2017), humidity-and light-dependent heterogeneous generation (Finlayson-Pitts et al., 2003;Ndour et al., 2008;Monge et al., 2010), soil emissions (Oswald et al., 2013;Su et al., 2011;Meusel et al., 2018), photolysis of particle nitrate in the atmosphere (Ye et al., 2016b(Ye et al., , 2017, and photolysis of deposited HNO 3 and nitrate on the ground (Ye et al., 2016a;Zhou et al., 2011).
During 4-8 January 2017, a severe winter air pollution event occurred in the Pearl River Delta (PRD), a region long known to suffer from photochemical pollution due to its fast industrialization (Chan and Yao, 2008;Wang et al., 1998Wang et al., , 2003Xue et al., 2014;Zhang et al., 2008;Zheng et al., 2010). During this multiday episode, an hourly peak value of ∼ 150 ppb for O 3 and ∼ 400 µg m −3 for PM 2.5 were observed, and the HONO levels reached 8 ppb, with an average value of 2.9 ppb at night (18:00 to 06:00 LT) and 2.4 ppb in daytime (07:00 to 17:00 LT). The HONO values were among the highest ever reported in China (Bernard et al., 2016;Huang et al., 2017;Li et al., 2012Li et al., , 2018aQin et al., 2009;Su et al., 2008;J. Wang et al., , 2013Xu et al., 2015). It is of great interest to find out how such a high level of HONO was produced and what impact it had on the radical levels and secondary pollutants during this severe winter pollution event.
The present study utilizes the Community Multiscale Air Quality (CMAQ) model with up-to-date HONO sources, including those in the original CMAQ model (gas-phase generation, heterogeneous reaction, and vehicle emissions) and four newly added sources (RH-enhancing effects on heterogeneous reactions, light-enhancing effects on heterogeneous reactions, photolysis of particulate nitrate in the atmosphere, and photolysis of HNO 3 and nitrate adsorbed on surfaces). The updated model was then used to analyze the contributions of different HONO sources and also to quantify the contributions of HONO to secondary pollutants during this severe winter pollution event. Our study reveals the very large impact of HONO on winter-time chemistry at this subtropical site.  (Appel et al., 2017) was used in this study. One-way triple nesting domains were used with their horizontal resolutions being 36, 12, and 4 km, respectively, and the innermost domain (domain 3) covers the PRD region (Fig. 1). These domains are based on a Lambert projection with two true latitudes of 25 and 40 • N. The objective simulation period was 4 to 8 January 2017, with 6 days before as a spin-up time.
The Weather Research and Forecasting (WRF) Model version 3.7 was applied to generate the meteorological fields for the CMAQ simulations. The physical options used in the WRF model were the Lin microphysics scheme (Lin et al., 1983), the Rapid Radiative Transfer Model for GCMs (RRTMG) shortwave and longwave radiation scheme (Mlawer and Clough, 1998;Mlawer et al., 1997), Noah land surface scheme (Chen and Dudhia, 2001), YSU PBL scheme (Hong et al., 2006), and Kain-Fritsch cumulus scheme (Kain, 2004). To improve the meteorological modeling performance, the nudging was performed using the NCEP Au- tomated Data Processing (ADP) data (ds351.0 and ds461.0) and surface observation data from China Meteorological Administration . Table S1 in the Supplement summarizes the statistical performance for the meteorological predictions.
The anthropogenic emission input was generated based on three emission inventories covering different regions. For the PRD region, a local emission inventory with high resolution for 2010 (Pan et al., 2014) was used. For other regions in China, the emission data for 2013 were from Ma et al. (2017). For other Asian countries, the INTEX-B dataset ) was used. HONO emissions from transportation sources were calculated based on the HONO / NO x ratios and NO x emissions from the transportation sources in the anthropogenic emission inventory. The HONO / NO x ratios were set as 0.8 % and 2.3 % for gasoline and diesel engines, respectively Gutzwiller et al., 2002). Natural biogenic emissions were estimated by the Model of Emissions of Gases and Aerosols from Nature (MEGAN) (Guenther et al., 2006).

Parameterization of HONO sources
In addition to the direct anthropogenic emissions, the default CMAQ model has two more HONO sources, including the gas-phase homogeneous reaction of NO and OH (Reaction R7) and the heterogeneous reactions of NO 2 on surfaces (Reaction R8) (Sarwar et al., 2008). The heterogeneous formation of HONO on the particle, urban, and leaf surfaces was estimated with a reaction rate k = 5 × 10 −5 × (S/V) as measured by Kurtenbach et al. (2001) In the present study, we incorporated four additional HONO sources, as described below.
1. RH-enhancing effects on the heterogeneous reaction of NO 2 on surfaces The default heterogeneous reaction rate was based on measurements at a relative humidity of 50 %. However, previous field and lab studies found that surface adsorbed water played a key role in the heterogeneous conversion and the reaction rate was highly dependent on the RH level (Finlayson-Pitts et al., 2003;Stutz et al., 2004;Qin et al., 2009). The laboratory measurements over an RH range of 0 % to 80 % conducted by Finlayson-Pitts et al. (2003) showed that the heterogeneous conversion rate increased much faster when RH ≥ 50 % than that when RH < 50 %. Based on this result, the RH dependence of the heterogeneous reaction was considered through scaling the default reaction rate by a factor of f RH in this study, as shown in the following equation:

Light-enhancing effects on the heterogeneous reaction of NO 2 on surfaces
The default reaction rate coefficient for the heterogeneous reaction was based on measurements under dark conditions. However, it has been reported that sunlight significantly boosts the heterogeneous generation of HONO (Ndour et al., 2008;Monge et al., 2010;Stemmler et al., 2007). To consider the photo-enhancing effect, we applied a higher reaction rate at daytime Czader et al., 2012), as shown in the following equation: where light intensity means the total downward irradiance at the surface, measured in watts per meter squared (W m −2 ).

Photolysis of particulate nitrate in the atmosphere
Evidence from recent aircraft observations and laboratory measurements suggested that particulate nitrate in the atmosphere can undergo photolysis to produce HONO and NO 2 (Reaction R9) (Ye et al., 2016b(Ye et al., , 2017. Ye et al. (2017) reported the photolysis rates ranging from 6.2 × 10 −6 to 5.0 × 10 −4 s −1 , with a median of 8.3 × 10 −5 s −1 at noontime tropical conditions. The reported value was much higher than the photolysis rate of gaseous HNO 3 (∼ 7 × 10 −7 s −1 ) under the typical tropical noontime conditions (Finlayson-Pitts and Pitts, 2000;Ye et al., 2017). Similar to the methodology of Sarwar et al. (2008), the photolysis rate of particulate nitrate was estimated as the following equation: where J HNO3−CMAQ is the photolysis rate of gaseous HNO 3 calculated online in CMAQ.

Photolysis of HNO 3 and nitrate deposited on surfaces
Field observations and lab studies also indicated that the photolysis of HNO 3 and nitrate deposited on the surface could be an important daytime HONO source (Reaction R10) (Zhou et al., , 2011Baergen and Donaldson, 2013). Ye et al. (2016a) tested the photolysis of HNO 3 and nitrate deposited on various natural and artificial surfaces in the laboratory, and reported the photolysis rates ranging from 6×10 −6 to 3.7×10 −4 s −1 , with a median of 3.4 × 10 −5 s −1 .
This reaction was incorporated into the CMAQ model by assuming deposited HNO 3 and nitrate on surfaces to equal the accumulation of dry deposition since the last precipitation event, referring to the method of Sarwar et al. (2008). The photolysis rate of deposited HNO 3 and nitrate was estimated as the following equation:

Simulation cases
Eight simulations were conducted considering different HONO sources, including the following.

Observation data
Field observations of HONO and other major air pollutants were conducted at the Heshan site (22 • 42 50 N, 112 • 55 17 E) in the PRD region (Fig. 1). Hourly HONO concentration was measured using a Long Path Absorption Photometer (LOPAP) (QUMA, Model LOPAP ® -03) (Heland et al., 2001). The same instrument was employed by our group in several previous field campaigns (Zha et al., 2014;Xu et al., 2015;Liang et al., 2017;Yun et al., 2018). The reader referred to these sources (e.g., Yun et al., 2018) for description of measurement principle. Following our previous practice, the instrument background was determined with synthetic air four times a day, and calibrations with a nitrite solution standard were conducted every 3 days. The time resolution of this instrument was 10 min. The detection limit was 7 ppt with an accuracy of ±20 %. The sample inlets were placed at the roof of a four-floor building, at a height of about 15 m above the ground.
Other instruments whose data are used in the present paper have been summarized in Yun et al. (2018) with references provided for each instrument. Briefly, PM 2.5 concentrations were determined by a multiangle absorption photometer (Thermo Scientific, Model 5012). Sulfate, nitrate, and ammonium in PM 2.5 were measured by a gas and aerosol collector coupled with ion chromatography (GAC-IC) system. Gas HNO 3 concentrations were also measured by this GAC-IC system. O 3 concentrations were measured by a UV photometric analyzer (Thermo Scientific, Model 49i). NO 2 concentrations were measured using a chemiluminescence instrument (Thermo Scientific, Model 42i) coupled with a photolytic converter (Droplet Measurement Technologies, model BLC). The sample inlets for these instruments were placed at the same height as LOPAP.
Additionally, hourly PM 2.5 , O 3 , and NO 2 observation data at 56 official monitoring sites (Fig. 1) in the PRD region  were obtained from the Ministry of Ecology and Environment (MEE). It should be noted that NO 2 concentrations in the national network were measured using the catalytic conversion method, which overestimates NO 2 , especially during the period with active photochemistry and at the location away from the primary emission sources (Xu et al., 2013). The NO 2 observation data were adjusted based on the method of Zhang et al. (2017) (also see footnote of Table 1).

Observed pollution in this winter episode
During 4-8 January 2017, a severe pollution episode was observed in the PRD region. As shown in Fig. S1, the episode average PM 2.5 concentrations observed at the Heshan site reached 142 µg m −3 , which is nearly twice the respective standard of China (75 µg m −3 for daily average PM 2.5 concentration). The hourly PM 2.5 concentration peaked at 382 µg m −3 , which was among the highest PM 2.5 concentrations reported so far in the PRD region (Tan et al., 2009;Wang et al., 2012;Yue et al., 2015). O 3 levels were also high, with a peak of ∼ 150 ppb. Figure S2 presented the spatial distribution of observed PM 2.5 and O 3 on the two heaviest polluted days (5 and 6 January) based on the inter-polation of the observation data at the 56 official monitoring sites. It can be seen that the whole western PRD region suffered from severe pollution. On 5 January, the highest PM 2.5 pollution occurred in Foshan city and its surrounding area, with a peak of 448 µg m −3 . On 6 January, the pollution expanded to wider areas. Figure 2 compared the simulated HONO mixing ratios by the default CMAQ model (GEH) and the revised CMAQ model (GEHRLPD) with the observations at the Heshan site for 4-8 January. For the HONO simulation, the default CMAQ significantly underestimated the levels, with a normalized mean bias (NMB) of −71.6 %. The maximum underestimation was 6 ppb in the early morning of 5 January. The simulated HONO levels in the afternoon were lower than 0.1 ppb in the default CMAQ model. By considering the four additional HONO sources, the model performance was improved considerably, with an NMB of −22.5 %. The simulated average daytime HONO mixing ratios increased from 0.37 to 1.44 ppb, and the simulated nighttime HONO values increased from 1.10 to 2.63 ppb. The revised model reproduced the HONO diurnal variation. Underestimation was found on the mornings of 5-6 and 8 January. One possible reason was that some HONO sources were not considered in this study. For example, previous studies have proposed that deposited HONO can be reserved in dew water on the ground or vegetation during the night and re-released when the dew water evaporates in the morning (He et al., 2006). Here we have simply estimated HONO released from dew in the morning on 5 January during 00:00-08:00 LT using the method described in He et al. (2006). The total HONO dry deposition was 1.28 × 10 −6 moles m −2 . At 09:00 LT, if 80 % of accumulated HONO was emitted to the surface model layer (about 0-30 m), the total released HONO amount would be 0.76 ppb. This suggests that dew evaporation may be a considerable source, but other sources may also exist. Overestimation was seen on the nights of 4 and 7 January, which was possibly related to some missing HONO sinks, e.g., the uptake of HONO on the ground surface (VandenBoer et al., 2014). Additionally, the overestimation of NO 2 could also partially explain the overestimation of HONO on the night of 7 January.

Evaluation of model performance
The inclusion of the four additional HONO sources also changed the model performance for other air pollutants. As shown in Fig. S1, the O 3 prediction by the default CMAQ model was not satisfactory, especially for the peak value. For example, on 6 January, the observed maximum 1 and 8 h O 3 mixing ratios at the Heshan site reached 147 and 125 ppb, respectively, exceeding the respective standard of China (∼ 93 and 75 ppb for maximum 1 and 8 h O 3 concentrations, respectively). The simulated levels by the default CMAQ model were just 64 and 52 ppb, respectively, which failed to predict the noncompliance of O 3 for this episode. In contrast, the revised model was better to reproduce the magnitude. For the episode average, the NMB values decreased from −32.9 % to 5.5 % for the maximum 8 h O 3 . Therefore, it is crucial to include these additional HONO sources into air quality models. For NO 2 , the revised model could reproduce its temporal variation in general, although underestimation was seen for some peak NO 2 values. This underprediction could also partially explain the underestimation of HONO peak values. For PM 2.5 , improvements also can be seen, especially for the peak values. For the sulfate, nitrate and ammonium components in PM 2.5 , as shown in Fig. S3, the revised model performed well in reproducing their temporal variations, with correlation coefficients (R values) of 0.5, 0.7, and 0.7, respectively. Compared to the default CMAQ model, the NMB values decreased from −29.8 % to −10 %, −53.8 % to −41.2 %, and −32.9 % to −14.1 % for fine particle sulfate, nitrate and ammonium, respectively. Underestimation could be seen for particle nitrate, especially on 5 January. The possible reason was underestimation of the emissions of NH 3 or other alkaline species (e.g., Ca, K, Na), which led to gas HNO 3 not being converted to particle nitrate sufficiently. The model performance for total nitrate (TNO 3 = gaseous HNO 3 + fine particle nitrate) was satisfactory, with an NMB of −3.2 % and R of 0.8.
Regionally, as shown in Table 1, the simulation results of the revised CMAQ model were in better agreement with the observations, with the NMBs decreasing from −33.3 % to 0.7 %, −8.1 % to 4.7 %, and 20.6 % to 9.4 % for 8 h maximum O 3 , hourly average PM 2.5 , and hourly average NO 2 values, respectively.
We have conducted a sensitivity test to compare the simulations with those considering more recent emission by linearly adjusting the emissions based on the ratio of 2017 emissions to 2010 emissions for China reported in a very recent  paper (Zheng et al., 2018), but no improvements were indicated for HONO, O 3 , and PM 2.5 at the Heshan site (Fig. S4). Despite the uncertainty, we think that the current emission setting is acceptable for simulations of the present case.
The above results indicated the satisfactory performance of the revised CMAQ model to simulate HONO and other major pollutants. It could be used to analyze the formation and impact of HONO in this episode. Figure 3 presents the contributions of different HONO sources to the HONO mixing ratios at the Heshan site. Heterogeneous generation on surfaces (including RH and light effects, H + R + L) was the largest contributor, representing 72 % of the average predicted HONO for the whole day. It contributed up to 81 % of the average predicted HONO values during the nighttime (18:00 to 06:00 LT) and ∼ 52 % during the daytime (07:00 to 17:00 LT). These findings on the dominant contribution from heterogeneous generation were mation of the high HONO concentrations in this pollution episode. During the nighttime and early morning (18:00 to 08:00 LT), RH-enhancing effects on the heterogeneous generation were significant, due to high relative humidity (70 % to 90 %) (Fig. S5). The heterogeneous generation of HONO had a ∼ 2-fold increase by considering RH-enhancing effects, compared to that under the uniform relative humidity of 50 %. From 09:00 LT to the afternoon, light-enhancing effects became important, contributing approximately 25 % to HONO levels, due to strong radiation with a maximum of ∼ 450 W m −2 (Fig. S5). In addition to enhancing the heterogeneous generation, the existence of light also increased HONO values through the photolysis of particle nitrate in the atmosphere and total nitrate deposited on the surface. Due to the high total nitrate concentrations in this pollution episode, these two photolysis sources were dominant during noon and afternoon (11:00 to 17:00 LT), contributing 31 % and 36 % to the HONO levels, respectively. Vehicle emissions contributed approximately 8 % of the daily average HONO values. The gas-phase homogeneous reaction was the smallest source, contributing approximately 3 % and 7 % of the daily and daytime average HONO values, respectively. Our model simulations suggest that the three additional light-dependent sources (L + P + D) could be an important part of missing daytime sources for the present case.

Impacts of HONO chemistry on HO x and O 3
The HO x (OH + HO 2 ) radical plays a key role in the atmospheric photochemical process. The photolysis of HONO can produce OH radicals, and a fast conversion exists between OH and HO 2 radicals. Therefore, the photolysis of HONO can affect the abundance of HO x radicals. Figure 4 presents the average diurnal OH and HO 2 variations based on the simulations with and without the HONO sources (NO_HONO vs. GEHRLPD). The diurnal pattern and magnitude of the simulated OH and HO 2 mixing ratios in the GEHRLPD case were comparable to the previous observations in the PRD region and other areas (Lu et al., 2012;Mao et al., 2010 (Czader et al., 2012;Mao et al., 2010;Tham et al., 2016;Xue et al., 2016;Li et al., 2018c), the contribution of HONO photolysis to HO x was much larger in the present study, mainly due to the high HONO concentrations in this winter episode. For example, Xue et al. (2016) and Li et al. (2018c) reported daytime average contributions of less than 30 % from HONO photolysis at two sites in Hong Kong. Tham et al. (2016) showed a contribution of less than 50 % in the morning at a site in the North China Plain. In these cases, the HONO levels were relatively low, with the peak HONO levels ranging among 1-3 ppb. The high HONO concentrations at Heshan increased the contribution of HONO photolysis to the formation of HO x in this winter pollution episode. A similar high contribution can be found in other winter studies with high HONO concentrations, e.g., Elshorbany et al. (2010). The enhanced HO x formation due to HONO chemistry could increase O 3 concentrations. As shown in Fig. 6, the simulated daytime average O 3 mixing ratio at the Heshan site in the GEHRLPD case was approximately 24 ppb (70 %) higher than that in the NO_HONO case, with a peak increase up to 60 ppb. The O 3 enhancement was much higher than the previous simulation results for the summer cases (Sarwar et al., 2008;Li et al., 2010;Zhang et al., 2016). In addition to higher HONO concentrations, another possible reason was  that the VOC / NO x ratio was lower in winter than that in summer, due to less biogenic VOC emissions and more NO x emission from fuel combustion in winter. For example, Zou et al. (2015) reported VOC / NO x values higher than 10 : 1 in summer and about 5 : 1 for winter based on 1-year observation in Guangzhou, which is about 60 km from the Heshan site. The lower VOC / NO x ratio increased the sensitivity of O 3 concentrations to HO x changes through Reactions (2) to (4). Therefore, the enhancement of O 3 concentrations was more significant (Li et al., 2018b). Figure 7 presents the spatial distribution of simulated surface O 3 concentrations at 15:00 LT in the NO_HONO and GEHRLPD cases. With HONO sources, the increase of the regional average O 3 mixing ratio at 15:00 LT was 34 % (∼ 17 ppb). The above results indicated the significant impacts of HONO chemistry on the atmospheric oxidation capacity and O 3 pollution in the PRD region during this heavy winter episode.

Impacts of HONO chemistry on secondary inorganic aerosol formation
Secondary inorganic aerosols, including nitrate, sulfate, and ammonium, contributed approximately 50 % to PM 2.5 concentrations during this episode (Fig. S6). Among them, nitrate was the dominant component, with peak concentrations reaching ∼ 110 µg m −3 , which was much higher than the sulfate and ammonium concentrations. Particle nitrate can be generated through the partition of HNO 3 to particle phase. The major formation pathway of HNO 3 is the oxidation of NO 2 by OH during the daytime (Reaction R11). HNO 3 can also be produced via heterogeneous reactions of N 2 O 5 on the particle surface at night (Reactions R12 to R14). Therefore, HONO chemistry can accelerate the formation of nitrate through the enhancement of atmospheric oxidation capacity (HO x and O 3 ).
OH + NO 2 → HNO 3 (R11) Considering the uncertainty of the partition of gaseous HNO 3 to particle nitrate (see Sect. 3.2), we assessed the impact of HONO chemistry on the sum of HNO 3 and fine particle NO − 3 (TNO 3 ). As shown in Fig. 8, the average TNO 3 concentrations at the Heshan site increased by about 17 µg m −3 (67 %), and the peak enhancement reached 55 µg m −3 at 10:00 LT on 6 January, when the observed PM 2.5 and nitrate concentration was at its respective highest level. Figure 9 presents the distribution of simulated average surface TNO 3 concentrations in the NO_HONO and GEHRLPD case. With HONO sources, the regional average TNO 3 concentrations were increased by 8.4 µg m −3 for the entire episode. The absolute enhancement was more significant in the areas with higher www.atmos-chem-phys.net/19/1/2019/ Atmos. Chem. Phys., 19, 1-14, 2019 aerosol concentrations. The highest increase for TNO 3 was above 25 µg m −3 for the episode average. Meanwhile, the HONO chemistry also accelerated the formation of other secondary inorganic aerosols. Particle sulfate can be generated through the partition of H 2 SO 4 to the particle phase. The major formation pathways of H 2 SO 4 include the gaseous oxidation of SO 2 by OH and the aqueous oxidation of S(IV) by O 3 , H 2 O 2 , and other oxidants. The partition of gaseous NH 3 to particle ammonium was based on the H 2 SO 4 −HNO 3 −NH 3 thermodynamic equilibrium. The simulated enhancements of the average sulfate and ammonium concentrations at the Heshan site were 32 % and 33 %, respectively. The above results indicated that the HONO chemistry also aggravated the particulate pollution during this episode.

Conclusions
This study has identified the major contributors to the observed high HONO levels during a severe winter pollution episode and highlighted the importance of HONO chemistry in the combined photochemical and haze pollution in a subtropical region. Including up-to-date HONO sources in the widely used CMAQ model significantly improved its capability in simulating ambient concentrations of HONO and other major pollutants (e.g., O 3 and PM 2.5 ). The model simulations suggested a predominant contribution from NO 2 heterogeneous reactions enhanced by humidity and solar radiation. The high HONO concentration significantly increased the atmospheric oxidation capacity and the levels of ozone and secondary aerosols, especially total nitrate. This study highlights the key role of HONO chemistry in the formation of winter haze in a subtropical environment and indicates the critical need to include and update HONO sources in regional air quality models in order to predict ozone and other secondary pollutants during heavy pollution events in southern China and similar regions. Additional efforts are needed to improve the current representation of HONO sources, such as evaporation of dew and a more accurate simulation of deposited nitrate.
Data availability. Model codes and input data are available from Xiao Fu. The measurement data used in this study are available from Tao Wang (cetwang@polyu.edu.hk) and Dingli Yue (dingliyue@163.com).
Author contributions. TW and XF designed the research; XF performed model simulations and analysis; LZ and QL assisted in model simulations; YZ, DY, ZW, HY, MX, CY, and WW conducted measurements of trace gases and aerosol; JZ processed the PRD emission data; RH processed the weather data; XF and TW wrote the manuscript. All authors contributed to discussion and commented on the paper.
Competing interests. The authors declare that they have no conflict of interest.
Special issue statement. This article is part of the special issue "Regional transport and transformation of air pollution in eastern China". It is not associated with a conference.