Impacts of an unknown daytime nitrous acid source on its daytime concentration and budget , as well as those of hydroxyl , hydroperoxyl , and organic peroxy radicals , in the coastal regions of China

Introduction Conclusions References

a 60-250 % increase of OH, HO 2 and RO 2 near the ground in the major cities of the coastal regions of China, and a 5-48 % increase of OH, HO 2 and RO 2 in the daytime meridional-mean mixing ratios within 1000 m above the ground.When the additional HONO sources were included, the photolysis of HONO was dominated in the OH production rate in Beijing, Shanghai and Guangzhou before 10:00 LST with a maximum of 10.01 [7.26 due to the P unknown ] ppb h −1 in Beijing, whereas the reaction of HO 2 + NO (nitric oxide) was dominated after 10:00 LST with a maximum of 9.38 [7.23] ppb h −1 in Beijing.The whole RO x cycle was accelerated by the additional HONO sources, especially the P unknown .The OH production rate was enhanced by 0.67 [0.64] to 4.32 [3.86] ppb h −1 via the reaction of HO 2 + NO, and by 0.85 [0.69] to 4.11 [3.61] ppb h −1 via the photolysis of HONO, and the OH loss rate was enhanced by 0.58 [0.55] to 2.03 [1.92]  via the reaction of OH + CO (carbon monoxide) in Beijing, Shanghai and Guangzhou.

Introduction
The hydroxyl radical (OH) is the dominant oxidant in the troposphere, initiating daytime photochemistry, removing the majority of reactive gases, and leading to the formation of secondary products (e.g.ozone (O 3 ), peroxyacyl nitrates (PANs) and aerosols) that can affect air quality, climate, and human health (Stone et al., 2012).OH is formed primarily through the photolysis of O 3 , nitrous acid (HONO), hydrogen peroxide (H 2 O 2 ) and the reactions of O 3 with alkenes (Platt et al., 1980;Crutzen and Zimmermann, 1991;Atkinson and Aschmann, 1993;Fried et al., 1997;Paulson et al., 1997).Recent field experiments have found the contribution of the photolysis of HONO to daytime OH production can reach up to 56, 42, and 33 % in urban, rural and forest areas, respectively (Ren et al., 2003;Kleffmann et al., 2005;Acker et al., 2006), more than that of the photolysis of O 3 .However, most current air quality models fail to predict observed HONO concentrations, underestimating daytime HONO in particular (Czader et al., 2012;Gonçalves et al., 2012;Li et al., 2011), due to the incomplete knowledge of HONO sources.
It is generally accepted that the photolysis of HONO (Reaction R2) in the early morning could be a major source of OH.After sunrise, HONO mixing ratios are usually below the detection limit due to the strong photolysis of HONO.However, many field experi-Figures
OH + NO → HONO (R1) The P unknown was calculated by Su et al. (2008) at Xinken (Guangzhou, China), with a maximum of 4.90 ppb h −1 .Spataro et al. (2013) proposed a P unknown value of 2.58 ppb h −1 in Beijing.In fact, P unknown values, ranging from 0.06 to 4.90 ppb h −1 have been obtained from many field studies across the globe, as shown in Fig. 1, suggesting P unknown could contribute greatly to the daytime production of OH and hydroperoxyl radical (HO 2 ).
Based on these mechanisms outlined above, some modeling studies have been carried out to simulate HONO concentrations (e.g.An et al., 2011;Czader et al., 2012;Introduction

Conclusions References
Tables Figures

Back Close
Full   2010) used a relatively high emission ratio of 2.3 % for HONO / NO 2 to compute the direct emissions of HONO, which could have overestimated the HONO concentrations in the air (An et al., 2013).Czader et al. (2012) added Reactions (R6), (R7) and HONO emissions into the CMAQ model.The HONO simulations matched well with observations at night, but were significantly lower than observations at noon.Wong et al. (2013) reported good agreement between simulated and observed daytime HONO when HONO emissions, photolytically enhanced daytime formation mechanisms on both aerosols and the ground, and Reaction (R7) were included.However, according to our recent studies (Tang et al., 2014), this result depended heavily on the selection of uptake coefficients of NO 2 heterogeneous chemistry.Overall, the topic of HONO sources remains under discussion today, and so it is a challenge for modelers to decide which mechanism(s) to be coupled into an air quality model.
To investigate the importance of the mechanisms described above, correlation tests between the P unknown and NO 2 , HNO 3 , irradiation or the photolysis frequency of NO 2 [J(NO 2 )] were conducted in field experiments (Acker et al., 2007;Sörgel et al., 2011;Villena et al., 2011;Wong et al., 2012).Many of these studies demonstrated that there is a clear dependency of the P unknown on irradiation/J(NO 2 ) during daytime, particularly at noon.Rohrer et al. (2005) proposed that the photolytic HONO source at the surface of the chamber strongly depended on light intensity.Acker et al. (2007) summarized field experiments in several European countries and showed a strong correlation (R 2 = 0.81) between the P unknown and J(NO 2 ).Wong et al. (2012) also indicated that the P unknown showed a clear symmetrical diurnal variation with a maximum around Introduction

Conclusions References
Tables Figures

Back Close
Full noontime, closely correlated with actinic flux (NO 2 photolysis frequency) and solar irradiance; the correlation coefficient was over 0.70.Besides irradiation/J(NO 2 ), good correlations between the P unknown and NO 2 mixing ratios have been found from both field and laboratory studies, supporting the viewpoint that NO 2 is the primary precursor of HONO.Through estimating the P unknown , Acker et al. (2007) speculated that the daytime HONO levels might be explained by a fast electron transfer onto adsorbed NO 2 .Sörgel et al. (2011) indicated that the conversion of NO 2 most likely accounted for light-induced HONO formation, about an order of magnitude stronger than HONO formation during nighttime.High correlations between the P unknown and NO 2 mixing ratios have also been found (e.g.R 2 = 0.77 in Qin et al. (2006), Villena et al. (2011), and R 2 = 0.62 in Elshorbany et al., 2009), indicating that the photosensitized conversion of NO 2 is more likely to be the daytime HONO source.This is the reason why some researchers have adopted the HONO / NO 2 ratio as a HONO emission factor to assess its implications (Elshorbany et al., 2012).
Based on the studies introduced above, the P unknown calculated from field experiments may be a practical method to help quantify the daytime HONO source.In this study, field experiment data from 13 different field campaigns across the globe were used to express the P unknown as a function of NO 2 mixing ratios and J(NO 2 ) (see Sect. 2.2).We then added the P unknown into the WRF-Chem model to assess the impacts of the P unknown on the concentrations and production and loss rates of HONO, OH, HO 2 , and organic peroxy radical (RO 2 ).
2 Data and methods

Observed data
Observed air temperature (TA), relative humidity (RH), wind speed (WS) and direction (WD) near the ground were obtained from the National Climatic Data Center, China Introduction

Conclusions References
Tables Figures

Back Close
Full NO 2 in Beijing were obtained from the Beijing Atmospheric Environmental Monitoring Action carried out by the Chinese Academy of Sciences (Li et al., 2011;Wang et al., 2014), except those in Guangzhou, which were sourced from Qin et al. (2009).HONO observations were conducted using an annular denuder at the campus of Peking University (PKU) (39 • 59 N, 116 • 18 E) in Beijing on 17-20 August 2007 (Spataro et al., 2013) and a long path absorption photometer at the Backgarden (BG) supersite (23 • 30 N, 113 • 10 E), about 60 km northwest of Guangzhou on 3-31 July 2006 (X.Li et al., 2012).The measurement systems are described in detail in Spataro et al. (2013) and X. Li et al. (2012).OH and HO 2 were measured by laser induced fluorescence at the BG supersite on 3-30 July 2006 (Lu et al., 2012).

Parameterization of HONO sources
Besides HONO gas-phase production from Reaction (R1), three additional HONO sources (HONO emissions, Reaction (R4) (nighttime), and the P unknown ) were coupled into the WRF-Chem model in this work.HONO emissions were calculated using [0.023 f DV denotes the nitrogen oxides (NO x ) emission ratio of diesel vehicles to total vehicles, and f TS is the NO x emission ratio of the traffic source to all anthropogenic sources (Li et al., 2011;An et al., 2013;Tang et al., 2014).Reaction (R4) was inserted into the Carbon-Bond Mechanism Z (CBM-Z) during nighttime only.The heterogeneous reaction rate was parameterized by k = a D g A s (Jacob, 2000), where a is the radius of aerosols, ν is the mean molecular speed of NO 2 , D g is a gas-phase molecular diffusion coefficient taken as 10 −5 m 2 s −1 (Dentener and Crutzen, 1993), and A s is the aerosol surface area per unit volume of air, calculated from aerosol mass concentrations and number density in each bin set by the Model for Simulating Aerosol Interactions and Chemistry (MOSAIC).Hygroscopic growth of aerosols was considered (Li et al., 2011).Introduction

Conclusions References
Tables Figures

Back Close
Full Previous studies have shown P unknown ∝ NO 2 •J(NO 2 ).To quantify the relationship between the P unknown and NO 2 mixing ratios and irradiation, daytime P unknown , NO 2 mixing ratios and J(NO 2 ), based on all the available data sets from 13 different field campaigns across the globe, were plotted in Fig. 2. As expected, good correlation (R 2 = 0.75) between the P unknown and NO 2 mixing ratios was obtained (Fig. 2a).Furthermore, the correlation between the P unknown and NO 2 • J(NO 2 ) was increased to 0.80, with a linear regression slope of 19.60 (Fig. 2b), so the P unknown cloud be expressed as a function of NO 2 mixing ratios and J(NO 2 ), i.e., P unknown ≈ 19.60 × NO 2 × J(NO 2 ).This formula is very similar to et al. (2008), and Wong et al. (2012) as an additional daytime source of HONO through analysis of observed data, where S/V a is the aerosol surface area-to-volume ratio, S/V g is the ground surface area-to-volume ratio, α is a fitting parameter, and Q s is solar visible irradiance.

Model setup
Used in this study was the WRF-Chem model version 3.2.1 (Grell et al., 2005;Fast et al., 2006), with the CBM-Z (Zaveri and Peters, 1999) and the MOSAIC (Zaveri et al., 2008).The detailed physical and chemical schemes for the simulations can be found in Tang et al. (2014).Two domains with a horizontal resolution of 27 km were employed in this study: domain 1 covered East Asia, whereas domain 2 covered the coastal regions of China, including the Beijing-Tianjin-Hebei region (BTH), the Yangtze River delta (YRD), and the Pearl River delta (PRD) (Fig. 3).There were 28 vertical model layers from the ground to 50 hPa, and the first model layer was ∼ 28 m above the ground.Meteorological initial and boundary conditions were obtained from the National Centers for Environmental Prediction (NCEP) 1 Full Monthly anthropogenic emissions in 2006/2007 and biogenic emissions were the same as those used by Li et al. (2011) and An et al. (2013).Six simulations (cases R, R wop , and R p performed for the entire months of August 2007 and July 2006) with a spin-up period of seven days were conducted in this study to assess the P unknown effects on the concentrations and budgets of HONO, OH, HO 2 , and RO 2 .Case R only considered Reaction (R1) as a reference; Case R wop included case R with HONO emissions, and Reaction (R4) only at night; case R p contained case R wop with the P unkonwn [≈ 19.60 × NO 2 × J(NO 2 )].The P unknown and Reaction (R4) were added to the CBM-Z, and diagnostic variables (i.e.production and loss rates of HONO, OH, HO 2 , RO 2 , O 3 , and other species) were inserted into the CBM-Z to quantify the P unknown impacts on the budgets of HONO, OH, HO 2 , and RO 2 (Wang et al., 2014).

Comparison of simulations and observations
Simulations of TA, RH, WS and WD were compared with observations, as shown in Wang et al. (2014).The statistical metrics, i.e. mean bias (MB), mean error (ME), root-mean-square error (RMSE), normalized mean bias (NMB), normalized mean error (NME), index of agreement (IOA), and correlation coefficient (CC), were comparable with those of Wang et al. (2010) 1 for case R), comparable to the values of 30.2 % for NMB, 55.8 % for NME and 0.91 for IOA reported in L. Li et al. (2012) using the CMAQ model.When HONO emissions, Reaction (R4) and the P unknown were included, the NMB, NME and IOA increased to −2.20 %, 66.10 % and 0.80, respectively (Table 1 for case R p ).The NO 2 fluctuations were generally captured (Fig. 4) but Introduction

Conclusions References
Tables Figures

Back Close
Full the simulated amplitude of NO 2 was underestimated in some cases (Fig. 4).This underestimation could be related with the uncertainty of NO x emissions.For case R, the NMB, NME and IOA for NO 2 were −13.50 %, 42.10 % and 0.57, respectively (Table 1), similar to the results of Wang et al. (2010) using the CMAQ model (NMB of −33.0 %, NME of 50.0 %, and IOA of 0.61).Compared with case R, NO 2 simulations were further underestimated for case R p (Table 1 for case R p ) due to the underestimation of NO x emissions in Guangzhou.
HONO simulations with the gas-phase production only (case R) were always substantially underestimated compared with observations (Fig. 5), similar to the results of Sarwar et al. (2008), Li et al. (2011) and An et al. (2013).When HONO emissions and Reaction (R4) were included, HONO simulations were significantly improved, especially at night (Fig. 5 and Table 2 for case R wop ).For Beijing, the nighttime RMSE and NME were reduced by 0.90 × 10 6 molecules cm −3 and 44.70 %, whereas the NMB and IOA were increased by 50.00 % and 0.29, respectively (Table 2).For Guangzhou, the nighttime RMSE and NME were reduced by 0.44 × 10 6 molecules cm −3 and 32.90 %, and the NMB and IOA were enhanced by 58.80 % and 0.18, respectively.When the P unknown was included, daytime HONO simulations were considerably improved (Fig. 5 and Table 2 for case R p ).Compared with case R wop , the daytime NME in Beijing was reduced by 19.60 %, and the NMB and IOA in Beijing were increased to −24.30 from −62.00 % and 0.73 from 0.64, respectively (Table 2); the daytime NME in Guangzhou was reduced by 8.10 %, and the NMB in Guangzhou was increased to −61.20 from −76.50 % (Table 2).Simulated diurnal variations of OH and HO 2 showed consistent patterns with the observed data (Fig. 6).When HONO emissions and Reaction (R4) were considered (case R wop ), OH and HO 2 enhancements were minor in most cases compared with case R (Fig. 6 and Table 3), but the P unknown led to noticeable improvements in OH simulations on 5-12 July 2006 (Fig. 6).Substantial overestimation of OH mixing ratios on 20-25 July 2006 (Fig. 6) needs further investigation.Compared with case R, the NME was reduced by 79.60 %, whereas the NMB was increased by 105.40 %, and the Introduction

Conclusions References
Tables Figures

Back Close
Full IOA was improved to 0.84 from 0.79 (Table 3).When the P unknown was considered, HO 2 simulations were substantially improved (Fig. 6), the IOA was improved to 0.61 from 0.54 and the CC was improved to 0.66 from 0.57 (Table 3).However, HO 2 simulations were still substantially underestimated (Fig. 6).One of the major reasons for the HO 2 underestimation could be related to the considerable underestimation of anthropogenic volatile organic compounds (VOCs) (Wang et al., 2014).

P unknown simulations and its impacts on production and loss rates of HONO
High P unknown values were found in the coastal regions of China (Fig. 7), especially in the BTH, YRD and PRD regions due to elevated emissions of NO x (Zhang et al., 2009).
For the BTH region, the largest daytime-mean P unknown values reached 2.5 ppb h −1 in Tianjin (Fig. 7a).Elevated daytime-mean P unknown values were found in the YRD region, with a maximum of 2.0 ppb h −1 in Shanghai (Fig. 7b).Daytime-mean P unknown values reached 1.2 ppb h −1 in Guangzhou and Shenzhen of the PRD (Fig. 7c).The simulated P unknown values in the PRD region were lower than those in the BTH and YRD regions.One major reason is the underestimation of daytime NO 2 mixing ratios in the PRD (Fig. 4b).
For case R, daytime HONO production was primarily from the reaction of OH and nitric oxide (NO) (Reaction R1), with a maximum production rate of 0.69 ppb h −1 in Beijing, 1.20 ppb h −1 in Shanghai, and 0.72 ppb h −1 in Guangzhou near noon due to high OH mixing ratios (Fig. 8a, c, e).The loss rate of HONO was 0.62 ppb h −1 in Beijing, 0.25-0.84ppb h −1 in Guangzhou) (Fig. 8a, c, e).HONO emissions contributed 0.04-0.62ppb h −1 to HONO production (Fig. 8a, c, e).Simulated P unknown values ranged from 0.42 to 2.98 ppb h −1 in Beijing, from 0.18 to 2.58 ppb h −1 in Shanghai, and from 0.06 to 1.66 ppb h −1 in Guangzhou (Fig. 8a, c, e).The simulated P unknown values in Beijing (Fig. 8a) were in good agreement with the results of Spataro et al. (2013), with an average unknown daytime HONO production rate of 2.58 ppb h −1 in the studied summer period.However, the simulated P unknown values in Guangzhou (Fig. 8e) were lower than the 2.36-4.90ppb h −1 reported by Su et al. (2008), due mainly to the underestimation of the daytime NO 2 mixing ratios in the PRD region.The additional HONO sources produce more HONO, which subsequently photolyzes to yield more OH.Therefore, the formation of HONO through Reaction (R1) was greatly enhanced, with a maximum of 4.70 18] ppb h −1 in Guangzhou (Fig. 8b, d, f).The HONO loss rate via dry deposition ranged from 0.28 to 0.45 ppb h −1 (not shown), roughly equivalent to the contribution of HONO emissions, suggesting that dry deposition of HONO cannot be neglected in high NO x emission areas.

P unknown impacts on concentrations of OH, HO 2 and RO 2
Incorporation of the P unknown into the WRF-Chem model led to substantial enhancements in the daytime-mean mixing ratios of OH in the coastal regions of China, e.g.60-190 % in the BTH region, 60-210 % in the YRD region, and 60-200 % in the PRD region (Fig. 9a).The maximum enhancement of HO 2 reached 250 % in the BTH region, 200 % in the YRD region, and 140 % in the PRD region (Fig. 9b).Similarly, a maxi-Introduction

Conclusions References
Tables Figures

Back Close
Full Vertically, the P unknown enhanced the monthly meridional-mean daytime (06:00-18:00 LST) mixing ratios of OH, HO 2 and RO 2 by 5-38, 5-47 and 5-48 %, respectively, within 1000 m above the ground in the coastal regions of China (Fig. 10).The vertical enhancements of OH, HO 2 and RO 2 at the same latitude were roughly uniform within the 1000 m (Fig. 10) due to strong vertical mixing in the daytime.Different P unknown distributions led to distinct differences in the enhancements of OH, HO 2 and RO 2 , with a maximum located near 35 • N (Fig. 10).

P unknown impacts on the budgets of OH, HO 2 and RO 2
OH radicals are produced mainly through the photolysis of HONO and O 3 , the reactions between O 3 and alkenes, and the reaction of HO 2 + NO (Fig. 11).For case R, the predominant contribution to P(OH) [production rate of OH] was the reaction of HO 2 + NO, with a diurnal peak of 4.04 ppb h −1 in Beijing, 1.52 ppb h −1 in Shanghai, and 3.91 ppb h −1 in Guangzhou at noon (Fig. S1a, c, e in the Supplement).The photolysis of HONO and O 3 were the second and third most important sources of OH.The former was dominant (1.09 ppb h −1 in Beijing, 0.65 ppb h −1 in Shanghai, and 0.71 ppb h −1 in Guangzhou) mainly in the morning, while the latter was dominant (0.91 ppb h −1 in Beijing, 0.52 ppb h −1 in Shanghai, and 1.20 ppb h −1 in Guangzhou) at noon (Fig. S1a, c, e).
Compared with the three OH sources above, the contributions of the reactions of O 3 + alkenes and others were small, lower than 0.15 ppb h −1 (Fig. S1a, c, e).When the additional HONO sources were added, the photolysis of HONO became the most important source of OH in Beijing and Guangzhou before 10:00 LST, and in Shanghai before 12:00 LST; the diurnal peaks were 10.01 to the 3.10 ppb h −1 reported by Elshorbany et al. (2009).Another important source was the reaction of HO 2 + NO, with a diurnal maximum conversion rate reaching 9.38 [7.23] (Figs. 11b,d,f and S1b,d,f).The diurnal maximum loss rates were 1.98 ppb h −1 in Beijing, 1.12 ppb h −1 in Shanghai, and 1.70 ppb h −1 in Guangzhou for case R (Fig. S1b, d, f), whereas these values were 5.61 [4.38 due to the P unknown ] ppb h −1 in Beijing, 2.00 [1.00] ppb h −1 in Shanghai, and 2.65 [1.02] ppb h −1 in Guangzhou for case R p (Fig. 11b, d, f).The reactions of OH + VOCs to form HO 2 and RO 2 were the second most important loss path of OH, with a diurnal maximum of 0.75-1.73ppb h −1 for case R (Fig. S1b, d, f) and 1.57 [0.82 due to the P unknown ] to 5.37 [4.05] ppb h −1 for case R p in Beijing, Shanghai and Guangzhou (Fig. 11b, d, f).The third most important OH loss path was the reaction of OH + CO to form HO 2 ; the diurnal maximum rates were 0.46-1.47ppb h −1 for case R (Fig. S1b, d, f) and 0.93 [0.49due to the P unknown ] to 3.58 [2.86] ppb h −1 for case R p in Beijing, Shanghai and Guangzhou (Fig. 11b, d, f).
The averaged radical conversion rates in the daytime (06:00-18:00 LST) are illustrated in Fig. 12. OH radicals are produced mainly via the photolysis of O 3 , HONO and hydrogen peroxide (H 2 O 2 ), and the reactions between O 3 and alkenes, after which OH Introduction

Conclusions References
Tables Figures

Conclusions References
Tables Figures

Back Close
Full

Conclusions
The relationship between the P unknown , NO 2 mixing ratios and J(NO 2 ) was investigated using available data from 13 field studies across the globe.The formula P unknown ≈ 19.60 × NO 2 × J(NO 2 ) was obtained, and then the additional HONO sources (i.e. the P unknown , HONO emissions and nighttime hydrolysis conversion of NO 2 on aerosols) were inserted into the WRF-Chem model, to assess the P unknown impacts on the concentrations and budgets of HONO and RO x in the coastal regions of China.The results showed that: Introduction

Conclusions References
Tables Figures

Back Close
Full 8.The additional HONO sources produced an increase of 0.31 [0.28 due to the P unknown ] to 1.78 [1.64]Overall, the above results suggest that the P unknown significantly enhances the atmospheric oxidation capacity in the coastal regions of China by increasing RO x concentrations and accelerating RO x cycles, and could lead to considerable increases in concentrations of inorganic aerosols and secondary organic aerosols and further aggravate haze events in these regions.
The Supplement related to this article is available online at doi:10.5194/acpd-15-807-2015-supplement.Introduction

Conclusions References
Tables Figures

Back Close
Full  Full  Full  Full  Full ppb h −1 via the reaction of OH + NO 2 and by 0.31 [0.28] to 1.78 [1.64] ppb h −1 Discussion Paper | Discussion Paper | Discussion Paper |

2NO 2 +
H 2 O → HONO + HNO 3 (R4) Ammann et al. (1998) found HONO formation via the heterogeneous reduction of NO 2 on the surface of soot (Reaction R5), and Reaction (R5) can be enhanced by irradiation Discussion Paper | Discussion Paper | Discussion Paper | (Monge et al., 2010): Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | and L. Li et al. (2012) using the fifth-generation Pennsylvania State University/National Center for Atmospheric Research Mesoscale Model (MM5) and Zhang et al. (2012a) using the WRF model.For O 3 in Beijing of the BTH region and Guangzhou of the PRD region, the NMB, NME and IOA were −22.80 %, 58.70 % and 0.79, respectively (Table Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper |

Figure 1 .Figure 3 .
Figure 1.Summary of observed HONO mixing ratios at noon (black font) and the calculated unknown daytime HONO source (blue font) from field studies.
Li et al. (2010)008)12).Sarwar et al. (2008)incorporated Reactions (R4), (R9) and HONO emissions into the Community Multiscale Air Quality (CMAQ) model, but still underestimated HONO mixing ratios during daytime.Li et al. (2010)considered both aerosol and ground surface reactions, and HONO emissions in the WRF-Chem model (Weather Research and Forecasting model coupled with Chemistry), and found that HONO simulations were significantly improved.However,Li et al. ( ppb h −1 in Beijing, 2.63[1.15]ppbh−1 in Shanghai, and 4.88[1.43]ppbh −1 in Guangzhou near noon (Fig. 11a, c, e).The contributions of the photolysis of O 3 , the reactions of O 3 + alkenes and others to P(OH) showed minor changes in comparison with case R (Figs. 11a, c, e and S1a, c, e).Kanaya et al. (2009), who also conducted similar studies at Mount Tai (located in a rural area) of China, suggested that the reaction of HO 2 + NO was the predominant OH source, with a daytime average of 3.72 ppb h −1 , more than the 1.38 ppb h −1 of the photolysis of O 3 .Hens et al. (2014) reported similar results in a boreal forest, in which the dominant contributor to OH was the reaction of HO 2 + NO, ranging from 0.23 to 1.02 ppb h −1 during daytime.The production rates of OH in our study were higher than in Kanaya et al. (2009) and Hens et al. (2014) due to higher NO x emissions in urban areas than in rural areas.The dominant loss rate of OH was the reaction of OH + NO 2 for both cases R and R p The second and third largest sources of OH were the photolysis of HONO [0.65 ppb h −1 (16.05 %) in ppb h −1 via the reaction of OH + CO and 0.10 [0.09] to 0.63 [0.59] ppb h −1 via the reaction of CH 3 O 2 + NO in the HO 2 production rate, and 0.67 [0.61] to 4.32 [4.27] ppb h −1 via the reaction of HO 2 +NO in the HO 2 loss rate in Beijing, Shanghai and Guangzhou.Similarly, the additional HONO sources

Table 1 .
Model performance statistics for O 3 and NO 2 in Beijing in August 2007 and Guangzhou in July 2006.

Table 3 .
Model performance statistics for OH and HO 2 in Guangzhou in July 2006.