Radical chemistry in the Pearl River Delta: observations and 1 modeling of OH and HO 2 radicals in Shenzhen 2018 2

. The ambient OH and HO 2 concentrations were measured continuously during the STORM (STudy of the Ozone 15 foRmation Mechanism) campaign at the Shenzhen site, located in the Pearl River Delta in China, in autumn 2018. The diurnal 16 maximum OH and HO 2 concentrations, measured by laser-induced fluorescence, were 4.5×10 6 cm -3 and 4.5×10 8 cm -3 , 17 respectively. The state-of-the-art radical chemical mechanism underestimated the observed OH concentration, similar to the 18 other warm-season campaigns in China. The OH underestimation was attributable to the missing OH sources, which can be 19 explained by the X mechanism. Good agreement between the observed and modeled OH concentrations was achieved when 20 an additional numerical X equivalent to 0.1 ppb NO concentrations was added to the base model. The modeled HO 2 could 21 reproduce the observed HO 2 , indicating the HO 2 heterogeneous uptake on HO 2 chemistry was negligible. Photolysis reactions 22 dominated the ROx primary production rate. The HONO, O 3 , HCHO, and carbonyls photolysis accounted for 29%, 16%, 16%, 23 and 11% during the daytime, respectively. The ROx termination rate was dominated by the reaction of OH + NO 2 in the 24 morning, and thereafter the radical self-combination gradually became the major sink of ROx in the afternoon. The atmospheric 25 oxidation capacity was evaluated, with a peak of 0.75 ⅹ 10 8 molecules cm -3 s -1 around noontime. A strong positive correlation 26 between O 3 formation rate and atmospheric oxidation capacity was achieved, illustrating the atmospheric oxidation capacity 27 was the potential tracer to indicate the secondary pollution. et al., 2012;Lu et al., 2019). A zero- 104 dimensional box model was used to conduct the radical closure experiment, and the overall framework was reported by Lu et 105 al. (2019). In this work, we conducted the radical closure experiment based on the Regional Atmospheric Chemical Mechanism 106 updated with the lasted isoprene chemistry (RACM2-LIM1), as Tan et al. (2017) described in detail. The model was 107 constrained to the measured meteorological, photolysis frequency, and the critical chemical parameters (CO, NO, NO 2, VOCs, 108 etc. ). The H 2 and CH 4 mixing ratios were set to 550 ppb and 1900 ppb, respectively. The model was operated in time-dependent 109 mode with a 5-min time resolution, and a 2-d spin-up time was used to reach steady-state conditions for long-lived species. As Lu et al. (2012) described, there are two types of radical closure experiment. One is the comparison of observed and 111 modeled radical concentrations, and the other is the comparison of radical production and destruction rates. The most 112 significant difference between the above is that the latter is conducted with the radical concentrations and k OH constrained. The 113 comparison of radical production and destruction rates, also called radical experimental budget, can test the accuracy of the 114 was crucial to exploring radical chemistry. The experimental budget for HO 2 273 and RO 2 radicals could not be conducted because RO 2 was not measured during this campaign. Herein, we showed the 274 simulated results by the base model. Figure 6 illustrates the diurnal profiles of ROx primary production rate ( P (ROx)) and 275 termination rate ( L (ROx)), and the contributions of different channels during the daytime. 276 August 2021; c Maximum over a period of time; d Maximum on some day. 303 Herein, we explored the AOC in Shenzhen based on the observed radical concentrations for the first time. As illustrated in 304 Fig. 7 (a), the diurnal profile of AOC exhibits a unimodal pattern, which is the same as the diurnal profile of OH concentration 305 and j (NO 2 ), with a peak around noontime. The diurnal peak of AOC was 0.75 × 10 8 molecules cm -3 s -1 . Comparatively, AOC 306 in this study can be comparable to those evaluated in Beijing (summer, 2018) and Hong Kong (autumn, 2012) (Li et al., 307 2018;Liu et al., 2021), but much lower than those evaluated in Hong Kong (summer, 2011) and Santiago (summer, 2005) (Xue 308 et al., 2016;Elshorbany et al., 2009).


Introduction 30
Severe ambient ozone (O3) pollution is one of China's most significant environmental challenges, especially in urban areas 31 (Shu et al., 2020;Li et al., 2019;Wang et al., 2020;Ma et al., 2019b;Wang et al., 2017). Despite the reduction in emissions of 32 Figure 2 gives an overview of the meteorological and chemical parameters from 05 October to 28 October 2018, when OH 139 and HO2 radicals were measured. The diurnal variations of the temperature (T), relative humidity (RH), j(O 1 D), and j(NO2) 140 followed a regular pattern from day to day. The overall meteorological conditions were characterized by high temperature 141 (about 20~30 ℃), high relative humidity (60~80%), and intensive radiation with j(O 1 D) up to 2.0 × 10 -5 s -1 and j(NO2) up to 142 6.0 × 10 -3 s -1 . The relative humidity and photolysis-frequency in this autumn campaign were similar to those in the summer 143 https://doi. org/10.5194/acp-2022-113 Preprint. Discussion started: 21 February 2022 c Author(s) 2022. CC BY 4.0 License. campaign conducted at Chengdu site (Yang et al., 2021). The temperature in this campaign was lower than that at Chengdu 144 site, but similar to that in the autumn campaign at Heshan site located in PRD as well (Tan et al., 2019;Yang et al., 2021). 145 The concentration of CO showed weak diurnal variation, indicating there was the non-obvious accumulation of 146 anthropogenic emissions on a regional scale. NO concentration peaked at 12 ppb during morning rush hour when the traffic 147 emission was severe, and thereafter, O3 concentration started to increase with the decreasing of NO concentration. The maxima 148 of O3 hourly concentration were high up to 120 ppb. According to the updated National Ambient Air Quality Standard of China 149 (GB3095-2012), O3 concentration exceeded the Class-II limit values (hourly averaged limit 93 ppb) on several days (6,7,8,150 and 26 October) when the environmental condition was characterized by high temperature and low relative humidity. NO2 151 concentration was high at night because of the titration effect of O3 with NO. 152 Along with the high O3 concentration on 6, 7, 8, and 26 October, high HCHO concentration was also recorded during the 153 corresponding periods, indicating HCHO was mainly produced as secondary pollutions because of the active photochemistry 154 in this campaign. Isoprene, mostly derived from biogenic emissions and mainly affected by temperature, peaked around 155 noontime. Tan et al. (2019) reported the median concentration of HCHO and isoprene concentrations were 6.8 ppb and 0.6 ppb 156 during 12:00-18:00 at Heshan site. Similarly, the median concentration of HCHO and isoprene concentrations in this study 157 were 4.9 ppb and 0.4 ppb during the corresponding periods, respectively. As a proxy for traffic intensity, the toluene to benzene 158 ratio (T/B), which is below 2, means the traffic emissions are the major sources of VOCs (Brocco et al., 1997). The T/B 159 gradually dropped from 07:00 until it reached the minimum value at 09:00, indicating traffic emission contributed more to 160 VOCs during morning rush hour than during other periods. However, the T/B, which varied within a range of 7-12, was above 161 2, and thus VOCs emission during this campaign was mainly from other sectors such as those involving solvent evaporation.

Observed and modeled OH and HO2 radicals 166
The OH and HO2 radicals were measured during 05-28 October 2018. The timeseries of the observed and modeled HOx 167 concentrations are displayed in Fig. S1 (a and b) in the Supplementary Information. Data gaps were caused by the rain, 168 calibration, and maintenance. The daily maxima of the observed OH and HO2 concentrations varied in the range of (2-9) × 10 6 169 cm -3 and (2-14) × 10 8 cm -3 , respectively. As in previous campaigns, the largest OH concentrations appeared around noontime 170 and showed a high correlation with j(O 1 D), a proxy for the solar UV radiation driving much of the primary radical production 171 (Tan et al., 2019). 172 The diurnal maximum of the observed and modeled OH concentration was 4.5 × 10 6 cm -3 and 3.5 × 10 6 cm -3 . There was an 176 https://doi.org/10.5194/acp-2022-113 Preprint. Discussion started: 21 February 2022 c Author(s) 2022. CC BY 4.0 License. agreement between the diurnal profiles of the observed and modeled OH concentrations within their errors of 11% and 40%, 177 respectively. A systematic difference existed with the decreasing of NO concentration. The model could reproduce the observed 178 OH concentrations well only in the early morning before 10:00. However, the model would underestimate the observed OH 179 concentration after 10:00 when NO concentration dropped 2 ppb. The OH concentrations observed in the environments with 180 low NO levels were underestimated by the state-of-the-art models at Backgarden (summer) and Heshan (autumn) sites in PRD 181 as well, and the OH underestimation was identified to be universal at low NO conditions in China (Lu et al., 2013;Lu et al., 182 2012;Ma et al., 2019a;Tan et al., 2017;Yang et al., 2021). 183 the primary production rate. 227 The OH production rate matched well with the destruction rate only in the early morning to about 10:00. Thereafter, the OH 228 destruction rate was larger than the production rate, which could explain the underestimation of OH concentration by the model. 229 The discrepancy between the OH production and destruction rates was attributed the missing OH source. The biggest additional 230 OH source was approximately 4.6 ppb h -1 , which occurred at about 12:00, when the OH production and destruction rates were 231 11.9 ppb h -1 and 16.5 ppb h -1 , respectively. The unknown OH source could explain about one third of the total OH production 232 rate, indicating the exploration of missing OH source was significant to study the radical chemistry. Details are given below 233

Quantification of missing OH sources
In this study, we tested this unclassical X mechanism. Good agreement between observations and simulations of both OH 266 and HO2 was achieved when a constant mixing ratio of 0.1 ppb of X was added into the base model. As shown in Fig. 5, the 267 model with X mechanism could agree with the observed OH concentrations even at low NO conditions. Unclassical OH 268 recycling was identified again in this study. However, X is an artificial species that behaves like NO, and thus the nature of X 269 is still unknown to us. Further exploration on this unclassical OH recycling is needed to improve our understanding of radical 270 chemistry. 271

Sources and sinks of ROx 272
The detailed analysis of radical sources and sinks was crucial to exploring radical chemistry. The experimental budget for HO2 273 and RO2 radicals could not be conducted because RO2 was not measured during this campaign. Herein, we showed the 274 simulated results by the base model. Figure 6 illustrates the diurnal profiles of ROx primary production rate (P(ROx)) and 275 termination rate (L(ROx)), and the contributions of different channels during the daytime.

279
The grey areas denote nighttime.

280
The ROx primary production and termination rates were basically in balance for the entire day, with maxima of 4 ppb h -1 281 around noontime. The ROx primary production rate was similar to those at Heshan (4 ppb h -1 ) and Wangdu (5 ppb h -1 ) sites, 282 but lower than those at Backgarden (11 ppb h -1 ), Yufa (7 ppb h -1 ), and Chengdu (7 ppb h -1 ) sites (Lu et al., 2013;Lu et al., 283 2012;Tan et al., 2017;Tan et al., 2019;Yang et al., 2021). During daytime, the main constitution of P(ROx) was OH and HO2 284 primary production rate. HONO and O3 photolysis mainly dominated the OH primary production rate, and HCHO photolysis 285 constituted the major HO2 primary production rate. P(ROx) was dominated by photolysis reactions, in which the photolysis of 286 HONO, O3, HCHO, and carbonyls accounted for 29%, 16%, 16%, and 11% during the daytime. In the early morning, HONO 287 photolysis was the most important primary source of ROx, and the contribution of O3 photolysis became progressively larger 288 and was largest at noontime. A large discrepancy between the ratio of HONO photolysis rate to O3 photolysis rate in 289 summer/autumn and that in winter occurs generally. The vast majority of OH photolysis source is attributed to HONO 290 photolysis in winter because of the higher HONO concentration and lower O3 concentration. About half of L(ROx) came from 291 OH termination, which occurred mainly in the morning, and thereafter, radical self-combination gradually became the major 292 sink of ROx in the afternoon. OH + NO2, OH + NO, and OH + others contributed 35%, 5%, and 9% to L(ROx), respectively. 293 HO2 + HO2 and HO2 + RO2 accounted for 8% and 16% in L(ROx). 294

AOC evaluation 295
AOC controls the abundance of precursors and the production of secondary pollutants (Yang et al., 2020b;Elshorbany et al., 296 2009). It is necessary to quantify AOC for understanding photochemical pollution. The AOC has been evaluated in previous 297 studies, as shown in Table 1. Overall, the AOC values in summer are higher than those in autumn and winter, and the values 298 https: //doi.org/10.5194/acp-2022-113 Preprint. Discussion started: 21 February 2022 c Author(s) 2022. CC BY 4.0 License. at lower latitudes are higher than those at higher latitudes for the same reason. The vast majority of AOC in previous studies 299 are evaluated based on the non-observed radical concentrations. 300 Herein, we explored the AOC in Shenzhen based on the observed radical concentrations for the first time. As illustrated in 304 Fig. 7 (a), the diurnal profile of AOC exhibits a unimodal pattern, which is the same as the diurnal profile of OH concentration 305 and j(NO2), with a peak around noontime. The diurnal peak of AOC was 0.75 × 10 8 molecules cm -3 s -1 . Comparatively, AOC 306 in this study can be comparable to those evaluated in Beijing (summer, 2018) and Hong Kong (autumn, 2012) (Li et al., 307 2018;Liu et al., 2021), but much lower than those evaluated in Hong Kong (summer, 2011) andSantiago (summer, 2005)   The correlation between F(O3) and AOC during the daytime (08:00-18:00) was explored, as shown in Fig. 7 (b). A strong 324 positive correlation between F(O3) and AOC was found, indicating AOC plays a significant role in driving secondary pollution. 325 Most data points in this campaign focused on low AOC (the median values below 0.7 × 10 8 molecules cm -3 s -1 ) and low F(O3) 326 (the median values below 17 ppb h -1 ) regimes. The correlation between F(O3) and AOC by fitting the median was denoted by 327