A Comparative Study to Reveal the Influence of Typhoons on the 1 Transport , Production and Accumulation of O 3 in the Pearl River 2 Delta , China 3

The Pearl River Delta (PRD) region in South China is faced with severe ambient O3 pollution in autumn and summer, 14 which mostly coincides with the occurrence of typhoons above the Northwest Pacific. With increasingly severe O3 pollution 15 in the PRD under the influence of typhoons, it is necessary to gain a comprehensive understanding of the impact of typhoons 16 on O3 transport, production and accumulation for efficient O3 reduction. In this study, we analysed the general influence of 17 typhoons on O3 pollution in the PRD via systematic comparisons of meteorological conditions, O3 processes and sources on 18 O3 pollution days with and without typhoon occurrence (denoted as the typhoon-induced and no-typhoon scenarios, 19 respectively), and also examined the differences in these influences in autumn and summer. The results show that the approach 20 of typhoons was accompanied by higher wind speeds and strengthened downdrafts in autumn as well as the inflows of more 21 polluted air masses in summer, suggesting favourable O3 transport conditions in the typhoon-induced scenario in both seasons. 22 However, the effect of typhoons on the production and accumulation of O3 were distinct. Typhoons led to reduced cloud cover, 23 and thus stronger solar radiation in autumn, which accelerated O3 production, but the shorter residence time of local air masses 24 was unfavourable for the accumulation of O3 within the PRD. In contrast, in summer, typhoons increased cloud cover, and 25 weakened solar radiation, thus restraining O3 formation, but the growing residence time of local air masses favoured O3 26 accumulation. The modelling results using the Community Multiscale Air Quality (CMAQ) model for the typical O3 pollution 27 days suggest increasing contributions from the transport processes as well as sources outside the PRD for O3 pollution, 28 confirming enhanced O3 transport under typhoon influence in both seasons. The results of the process analysis in CMAQ 29 suggest that the chemical process contributed more in autumn but less in summer in the PRD. Since O3 production and 30 accumulation cannot be enhanced at the same time, the proportion of O3 contributed by emissions within the PRD was likely 31 to decrease in both seasons. The difference in the typhoon influence on O3 processes in autumn and summer can be attributed 32 to the seasonal variation of the East Asian monsoon. From the “meteorology-process-source” perspective, this study revealed 33


Calculation of the trajectories and air parcel residence time 158
To explore the potential effect of cross-regional transport on O3 pollution in the PRD, we applied the Hysplit model (Stein et 159 al., 2015) with the Global Data Assimilation System (GDAS) datasets as inputs to calculate 72-h backward trajectories reaching 160 the PRD at 14:00 LT for all O3 pollution days. The Modiesha site (23.1°N, 113.3°E; Fig. S1b), which is located in the central 161 part of the PRD, was the endpoint of backward trajectories, with its height set as 500 m above the ground. Its height was set 162 as 500 m above the ground to better represent the effect of long-range transport on O3 pollution, as well as to minimize the 163 disturbance of objects near the surface to the transport (Park et al., 2007). 164 165 Air parcel residence time (APRT), discussed by Huang et al. (2019), is the average number of hours that air parcels originated 166 from one place stay within a pre-defined domain, and long APRTs can be used to indicate good accumulation conditions for 7 locally sourced pollutants. To calculate APRTs in the PRD, we designed a 21×15 point matrix (resolution: 0.2°×0.2°) that 168 embraces the whole PRD (Fig. S4), and forward trajectories starting from these points were also calculated using the Hysplit 169 model. The height of all points was set as 100 m above the ground, which is close to the height of emissions to represent the 170 height of all local emissions and to reduce the disturbance of the surface, as well. The start times were set as 2:00, 8:00, 14:00 171 and 20:00 LT for all O3 pollution days. Afterwards, the length of time each trajectory remained within the administration 172 borders of the PRD, i.e., APRT, was calculated and attributed to its starting point. The APRT values of all points were averaged 173 for each scenario and were interpolated to obtain field results. APRTs in each point were averaged, and these averaged APRT 174 values in all points were interpolated using the Kriging method to obtain field results for the further comparisons. 175

CMAQ modelling: basic setups and modelling methods 176
We utilised the widely used 3D chemical transport model, the CMAQ model (version 5.0.2), to investigate the effects of 177 typhoons on O3 processes and sources. October 2015 and July 2016 featured the most severe O3 pollution under typhoon 178 influence among all representative months of the two seasons (Table S3), and thus, they were chosen as the period in the 179 CMAQ modelling (because there was no severe O3 pollution during the first 10 days of October 2015 and 3-5 November can 180 be classified as the no-typhoon O3 pollution days, we adjusted the modelling period in autumn to 11 October-10 November 181 2015) and all typhoon-induced and no-typhoon O3 pollution days in these two months served as representative O3 pollution 182 days under multiple scenarios in the comparisons. In detail, there were four typhoon-induced O3 pollution days (14-16 and 21 183 October 2015) and four no-typhoon O3 pollution days (28 October and 3-5 November 2015) in October 2015, whereas there 184 were four and six typhoon-induced and no-typhoon days in July 2016, respectively (typhoon-induced: 7-8 and 30-31 July 185 2016; and no-typhoon: 22-26 and 29 July 2016). The results of the daytime (9:00-17:00 LT) O3 PA and SA on the above O3 186 pollution days were averaged for each the typhoon-induced and no-typhoon scenarios in autumn (October 2015) and summer 187 (July 2016) and were used in the comparisons. 188

189
The main setups of the CMAQ modelling are presented as follows. Two-nested modelling domains with the resolutions of 36 190 and 12 km (denoted as d01 and d02, respectively) were set in this study (Fig. 2). Specifically, d02 covers the whole East and 191 Central China (EC-China), enabling us to evaluate the contribution of emissions in these areas to O3 pollution in the PRD. 192 There were 19 vertical layers in the CMAQ modelling, with about 10 layers within the PBL (about 0-1 km in heights; Guo et 193 al., 2016). The Weather Research and Forecasting (WRF) model (version 3.2) provided the meteorological fields used as inputs. 194 SMOKE (version 2.5) and MEGAN (version 2.10) were used to process the anthropogenic and biogenic emission files, 195 respectively. The anthropogenic emission inventory used in this study consisted of the following three parts: (1) emissions in 196 the PRD, which were provided by the Guangdong Environmental Monitoring Centre; (2) emissions in other areas of mainland 197 China, which were extracted from the MEIC inventory (He, 2012); and (3) emissions in other countries and regions in Asia,198 which were extracted from the MIX inventory (Li et al., 2017). The initial and boundary conditions of the d01 modelling were 199 obtained from the same-period results of the MOZART-4 global model (available at https://www.acom.ucar.edu/wrf-200 8 chem/mozart.shtml, last accessed: Dec. 2019), and those of the d02 modelling were extracted from the d01 modelling results. 201 The SAPRC07 gas-phase chemistry mechanism (Carter, 2010) and the AERO6 aerosol scheme were set in the CMAQ 202 modelling. In addition, the simulations of the two months were both started 10 days ahead to minimise the disturbance of the 203 bias of the initial conditions. The modelling performances of CMAQ and WRF were determined to be acceptable based on the 204 comparisons between the observational and modelling series of meteorological parameters, O3 MDA8, daily NO2 205 concentrations and the mixing ratios of non-methane hydrocarbons (NMHCs) in the PRD (for details, refer to Sect. 1 of in the 206 Supplement Information), which ensures the validity of the further analyses.

211
The PA tool in CMAQ was implemented to quantify the hourly contributions of O3 processes (or integrated process rate, IPR), 212 which includes vertical/horizontal transport (convection+diffusion), chemical process (net O3 production through gas-phase 213 reactions), dry deposition and cloud process. To explore the overall effect of typhoons on O3 transport and production in the 214 region, the mean PA results within the administration boundaries of the PRD were calculated and compared. First, we compared near-ground meteorological parameters in the PRD on the typhoon-induced and no-typhoon O3 pollution 238 days. The parameters from the ERA-Interim re-analysis (including the parameters of the first and second categories in Sect. 239 2.1) and the routine monitoring datasets (including air temperature, RH, wind speed, zonal and meridional wind speeds 240 measured at 14:00 LT of all O3 pollution days at 29 national meteorological sites within the PRD (Fig. S1a)) and the ERA-241 Interim re-analysis (including all near-surface parameters from the analysis and forecast fields introduced in Sect. 2.1, 242 extracted at the same time and the locations of sites as these in routine monitoring datasets) were used in the comparison 243 (since all O3 pollution days in October and over 60% of O3 pollution days in July were characterized with sunny, cloudy, or 244 overcast weathers with no rainfall in the PRD (Table S4, represented by the weather in Guangzhou), precipitation was not 245 considered in the comparisons).. The Mann-Whitney U test was applied to determine whether the above parameters were 246 significantly different (p < 0.05) in the two types of O3 pollution scenarios between typhoon-induced and no-typhoon O3 247 pollution days. 248 249 As is listed in Table 2, statistically significant differences between the typhoon-induced and no-typhoon scenarios existed for 250 most of the parameters, such as meridional (south-north) wind speed, cloud covers within various height ranges and net 251 surface solar radiationin both seasons, these parameters were significantly different for the two scenarios. It indicates that 252 the causes of O3 pollution may vary on typhoon-induced and no-typhoon O3 pollution days. Note that air temperature, one of 253 the parameters most closely related to O3 pollution in the PRD (Zhao et al., 2019), was not significantly different in the two 254 scenarios. We also found that the comparison in autumn and summer did not produce the same results: the typhoon-induced 255 days in autumn featured lower RH, stronger winds (especially north wind), reduced cloud cover (low, medium, high and 256 total) and stronger surface solar radiation, whereas in summer, these days had higher RH, weaker south winds, more cloud 257 cover (medium, high and total), weaker surface solar radiation and lower PBL heights. Therefore, the impact of typhoons on 258 O3 pollution differs in the two seasons, as well. In order to reveal the impact of typhoons on O3 transport, production, and 259 accumulation in the PRD, more detailed comparisons of the corresponding meteorological indicators are presented in the 260 following sections. 261 Table 2. The comparisons of meteorological parameters (all at 14:00 LT except for net surface solar radiation, which is the average value 262 for 9:00-17:00 LT) in the PRD for the three scenarios (no-typhoon, typhoon-induced, close typhoon-induced) in two seasons (autumn, 263 summer). RM, routine measurement; ERA, ERA-Interim re-analysis. All of the parameters are presented as "the mean value ± standard 264 deviation". The differences between parameters in the typhoon-induced or close typhoon-induced scenarios and the corresponding 265 typhoon-induced scenarios for the same season are given in parentheses, and "*" indicates p < 0.05, or statistically significant differences 266 between these parameters when the Mann-Whitney U test is used.

O3 transport conditions: comparison of wind speeds, backward trajectories and vertical air motions 268
The higher wind speeds and/or O3 levels in the transported air masses are, the more likely O3 transport plays an increasingly 269 important role in O3 pollution. In the PRD, O3 levels are closely linked to the type of air masses influencing the region, which 270 can be identified based on backward trajectories. According to Zheng  that are transported into the PRD along different paths and contribute to O3 pollution here, namely, the continental, coastal and 272 marine air masses (Fig. 3a). The continental and coastal air masses can bring O3 from EC-China to the PRD, and thus, they 273 are typically recognised as being polluted and contributing to relatively high O3 levels in the PRD. In contrast, the marine air 274 masses, originated from the South China Sea, are much cleaner. In this section, we studied the influence of typhoons on O3 275 transport by comparing wind speeds and 72-h backward trajectories in various the typhoon-induced and no-typhoon scenarios.

281
As is displayed in Fig. 3b-c, we identified the influence of continental air masses on the typhoon-induced O3 pollution days 282 in autumn, as well as mixed contributions from the continental and coastal air masses on the corresponding no-typhoon days. 283 However, for the former scenario, significantly increased wind speeds (Table 2) ensure more favourable conditions for the 284 transport of O3. In summer, the three types of air masses may all have affected O3 pollution in the typhoon-induced scenario, 285 while only the marine air masses influenced the PRD in the no-typhoon scenario (Fig. 3d-e). Since wind speeds did not vary 286 significantly (Table 2), the inflows of much more polluted air masses resulted in that typhoons also tended to increase the 287 contribution of transport to O3 pollution in the PRD in summer. The increasing influence of much more polluted air masses 288 (continental and coastal air masses) led by typhoon ensured that more O3 was transported to the PRD, thus typhoons also 289 tended to increase the contribution of transport to O3 pollution in the PRD in summer. In addition, the influence of different 290 air masses was also accompanied with variations in the prevailing winds in the PRD, that is, north winds and easterlies in the 291 typhoon-induced and no-typhoon scenarios in autumn, respectively, and southwest winds in the no-typhoon scenario in 292 summer (indicated by wind roses in Fig. S5). For the typhoon-induced scenario in summer, the dominate wind direction is 293 hard to determine. These variations in the local wind fields potentially result in the different spatial distribution of O3 294 concentrations in various scenarios. 295 13 296 Downdrafts are typically considered to be an important cause of typhoon-induced O3 pollution (Lam, 2018), but in which 297 scenarios downdrafts influence the PRD remains unclear. Thus, we explored the overall features of vertical air motions from 298 the surface layer to the tropopause in four the typhoon-induced and no-typhoon scenarios, and the ERA-Interim reanalysis 299 dataset (including the parameters of the third category all upper air parameters at multiple heights introduced in Sect. 2.1) was 300 utilised in the comparisons. The contours in Fig. 4 show the cross sections of mean vertical wind speed at 14:00 LT of all O3 301 pollution days corresponding to the typhoon-induced and no-typhoon scenarios of two seasons, which were made along the 302 113.2°E longitude line, from 26.0°N to 20.0°N along the 113.2°E longitude line (Fig. S4). On the typhoon-induced days in 303 autumn, downdrafts occurred over large areas above the PRD, especially above a height of ~4 km~700 hPa. Although updrafts 304 can still be found near the sea surface in this scenario, vertical wind speeds tended to be lower compared with those on the no-305 typhoon days in autumn, which also suggests the enhancement of downdrafts caused by typhoons. In summer, the influence 306 of downdrafts was found over the PRD under 850 hPa on the typhoon-induced O3 pollution days. However, overall, updrafts 307 prevailed above the land areas and downdrafts prevailed above the sea in both the typhoon-induced and no-typhoon scenarios 308 in summer, which is recognised as the structure of the East Asian summer monsoon cell (Chen et  scenario in summer were overall higher than these in the corresponding no-typhoon scenario. Therefore, the approach of 311 typhoons did not break the structure of the summer monsoon cell, but rather they further strengthened the vertical motions 312 above both land areas and sea. These analyses suggest that typhoons do not necessarily lead to downdrafts during O3 pollution 313 periods in the PRD and its adjacent areas; and in summer, vertical air motions affected by typhoons are more complicated than 314 expected owing to the existence of the East Asian summer monsoon. 315

316
We also explored the regions where downdrafts and updrafts occurred on a larger scale and their potential connections with 317 O3 levels. As is shown in Fig. 5, though updrafts appeared in the PRD at 850 hPa on the typhoon-induced days in autumn, 318 downdrafts dominated in the region at 700 and 500 hPa. For the areas to the north of the PRD, the important role of downdrafts 319 was found at all three heights. In contrast to the no-typhoon days in autumn, downdrafts tended to cover much larger areas in 320 this scenario. Moreover, these areas at 850 and 700 hPa generally featured higher O3 mixing ratios as well as lower RH (Fig.  321 S6) than others, which is a sign of possible direct downward O3 transport (Roux et al., 2020;Wang et al., 2020). This part of 322 O3 can notably aggravate near-ground O3 pollution in the PRD. In contrast, in summer, updrafts dominated the PRD at various 323 heights in both scenarios. Besides the PRD, most of the regions near the coast were characterised by updrafts above the land 324 as well as downdrafts offshore, further indicating the ubiquity of the summer monsoon cell. By comparing the two scenarios 325 in summer, we found that typhoons resulted in more areas being influenced by updrafts. The areas with high O3 levels did not 326 coincide with the downdraft-affected areas, and therefore, O3 transported from the upper air may play a less significant role in 327 the typhoon-induced O3 pollution in summer. 328

O3 production conditions: comparison of clouds 334
Clouds efficiently reflect solar radiation (Liou, 1976), and therefore, they have a notable impact on the local formation of O3. 335 Figure 6 displays the cross sections of mean ERA-Interim cloud liquid water contents (CLWC) at 14:00 LT of all O3 pollution 336 days corresponding to the typhoon-induced and no-typhoon scenarios of two seasons, which were also made along the 113.2°E 337 longitude line, from 26.0°N to 20.0°N (Fig. S4). The comparison of cloud liquid water contentCLWC in the cross sections 338 (Fig. 6, derived from the ERA-Interim datasets) suggests that typhoons generally resulted in fewer clouds in autumn but more 339 clouds in summer, which agrees well with the comparison of cloud covers in Table 2. The presence of fewer clouds on the 340 typhoon-induced days in autumn can be attributed to two reasons: the influence of dry air masses (indicated by lower RH in 341 Table 2 and Fig. S6) and/or the hindrance of cloud formation by downdrafts. In summer, the strengthened updrafts above the 342 land caused by typhoons favoured cloud formation, which is demonstrated by higher cloud liquid water contentCLWC at the 343 heights of 2-5 km 500-850 hPa and increases in medium and high cloud covers. In areas above the PRD below 850 hPa, 344 downdrafts led to slight decrease of clouds in the typhoon-induced scenario in summer, which is also indicated by reduced low 345 cloud cover. As a consequence of varied cloud covers in each scenario, on average, net surface solar radiation increased by 346 15 13% and decreased by 7% on the typhoon-induced days in autumn and summer, respectively (Table 2), which promoted and 347 hindered O3 production in the PRD during these two seasons, respectively.

O3 accumulation conditions: comparison of APRTs 359
The longer APRTs are, the more likely that O3 produced by local emissions accumulates within the targeted region and notably 360 contributes to near-ground O3 pollution. In order to study the effect of typhoons on O3 accumulation, we calculated APRTs in 361 the PRD in the four typhoon-induced and no-typhoon scenarios (Fig. 7) for the further comparisons. On the typhoon-induced 362 days in autumn, APRTs were typically 5-10 hours (mean = 9.5 hours)shorter than those on the no-typhoon days in autumn 363

Meteorological conditions on the close typhoon-induced days 379
On the close typhoon-induced days in the two seasons, stronger north winds prevailed and total cloud cover was higher than 380 that on the no-typhoon days (Table 2), suggesting better conditions for the transport of O3 but less favourable conditions for 381 O3 production. As displayed in Fig. S7, the APRT values were significantly lower on the close typhoon-induced days (mean 382 = 6.6 hours, 12.9 hours in autumn and summer, respectively) than on the no-typhoon days, making it even harder for locally 383 sourced O3 to accumulate within the PRD. Therefore, close typhoons are concluded to promote the transport of O3 from the 384 outside and to reduce the contributions of O3 produced from local emissions in a more notable way. In addition, close typhoons 385 led to stronger downdrafts in autumn and updrafts in summer than other scenarios in the same season (Fig. S8). It should be 386 noted that the structure of the summer monsoon cell near the PRD was destroyed in the close typhoon-induced scenario in 387 summer, indicating the stronger influence of typhoons on regional wind fields. The dominant role of O3 transport during O3 388 pollution days in this special scenario agrees well with the reported episode-based analyses (Lam et al., 2005;Li, 2013). 389

Comparisons of O3 processes and sources 390
The comparisons of meteorological conditions served as qualitative evidence to determine the general influence of typhoons 391 on O3 transport, production and accumulation in autumn and summer. Based on the comparison between the CMAQ modelling 392 results on typical O3 pollution days in October 2015 and July 2016, more quantitative evidence can be presented. Figure 8  393 displays modelled mean O3 MDA8 concentrations and wind fields (at 14:00 LT) in the four scenarios on the typhoon-induced 394 and no-typhoon O3 pollution days of two seasons. Large standard-exceedance (> 160 μg/m 3 ) areas were distributed in the PRD 395 on most days, and the typhoon-induced days of both seasons generally featured higher O3 levels. The distinct wind fields for 396 these scenarios, which were consistent with those in the longer timespan (Fig. S5) The PA tool in CMAQ was used to quantify the contributions of transport and chemical process to the O3 variations on O3 411 pollution days under various scenarios in October 2015 and July 2016. As is shown in Fig. 9, the daytime (9:00-17:00 LT) O3 412 PA results within the PRD in all scenarios share similar characteristics. Dry deposition dominated O3 removal near the surface, 413 and it also led to high gradients of O3 concentrations that promote downward O3 diffusion. Within the PBL (about 0-1 km in 414 height), O3 was mainly contributed by horizontal transport and chemical process, and vertical convection led to the drop of O3 415 concentrations. However, differences existed between the O3 PA results in four the typhoon-induced and no-typhoon scenarios, 416 indicating the impact of typhoons on the transport and production of O3. In both months, typhoons led to notably higher 417 contribution of horizontal transport to O3, especially in the lower and middle part of the PBL. Within the PBL, on average, it 418 increased from -0.9 ppb/h, -0.8 ppb/h to 1.2 ppb/h, 2.0 ppb/h under typhoon influence in autumn and summer, respectively. 419 The comparison of the contribution of chemical process (in absolute rates) suggests that they had opposite effects in the two  The contributions of various sources to O3 within the PRD are determined by the combined impact of O3 transport, production 432 and accumulation. The results for the mean daytime (9:00-17:00 LT) O3 SA near the ground (about 0-80 m in height) in four 433 scenarioson typhoon-induced and no-typhoon O3 pollution days are illustrated in Fig. 10. For polluted regions within the PRD, 434 stronger O3 production under typhoons did not lead to a higher proportion of local contributions to O3 pollution in October 435 2015it even decreased from 22% (on the no-typhoon days) to 17% (on the typhoon-induced days). The contributions of 436 EC-China emissions and BCON, in contrast, increased slightly from 37%, 41% to 40%, 43%, respectively. The distinction of 437 the O3 SA results is more apparent for the summer scenarios, that is, typhoons resulted in growing contributions from O3 438 transported from other regions (from 40% to 59%) but decreased local contributions (from 60% to 41%) in July 2016. More 439 favourable O3 accumulation conditions (indicated by higher APRTs on the representative typhoon-induced O3 pollution days 440 in summer (Fig. S9) Furthermore, owing to the variations of wind fields, the comparison results of O3 SA in different parts of the PRD may differ 449 from the regional ones. For instance, while the comparisons of O3 SA in the Xijiao and Modiesha site (located in the northeast 450 and central part of the PRD, respectively) agree well with those in the polluted regions of the PRD, higher contributions of 451 PRD emissions for O3 can be found in the Duanfen site (located in the southwest part of the PRD) on the typhoon-induced 452 days of two months in comparison to these on the corresponding no-typhoon days (Fig. 10). Since the site was located in the 453 downwind region in the typhoon-induced scenario in October 2015 (Fig. 8a), enhanced O3 production led by typhoons from 454 the massive emissions of O3 precursors in the central PRD (Zheng et al., 2009) contributed to higher local contributions for O3 455 pollution here (as the distribution of local contributions in percentage to daytime O3 shown in Fig. S10, the highest local 456 contribution in the PRD occurred in areas near the Duanfen site and almost reached 40% in this scenario, which was even 457 higher than that in the corresponding no-typhoon scenario (33%)). In the no-typhoon scenario in July 2016, the site was located 458 in the upwind regions under the prevailing of southwest winds, limiting the contributions of local emissions for O3 at the site 459 (Fig. 8d). Thus, higher local contributions can also be found in the typhoon-induced scenario in this month. 460

Discussion and conclusions 461
The significance of typhoons on O3 pollution in the PRD calls for thorough evaluations of the different causes of O3 pollution 462 with the appearance of typhoons in the Northwest Pacific. In this study, we revealed the different impacts of typhoons on O3 463 transport, production and accumulation in the PRD (as summarised in Fig. 11) through systematic comparisons of 464 meteorological conditions, the contributions of various O3 processes and sources in the typhoon-induced and no-typhoon 465 scenarios. We found that typhoons tended to promote O3 transport towards the PRD, but failed to provide more favourable O3 466 production and accumulation conditions simultaneously, which limited the contribution of local emissions to O3 pollution. 467 Furthermore, there were also differences between the influence of typhoons on O3 pollution in autumn and summer. More 468 favourable transport conditions occurred in the typhoon-induced scenario in autumn, which was characterised by higher wind 469 speeds and the increased influence of downdrafts. In summer, the mixed types of air masses in the typhoon-induced scenario 470 were likely to bring more O3 into the PRD than the clean marine air masses in the no-typhoon scenario, also suggesting 471 enhanced O3 transport under the influence of typhoons. Generally, typhoons led to cloudless conditions, stronger solar radiation, 472 and thus more rapid O3 production in autumn, but shorter APRTs (5-10 hours) suggest that locally sourced O3 was hard to 473 accumulate within the PRD. As a result, the contributions in percentage of local emissions to O3 pollution decreased (slightly 474 by ~5% for the polluted regions of the PRD in October 2015). In contrast, in summer, intensified updrafts associated with 475 typhoons strengthened cloud formation, weakened solar radiation, and thus restrained local O3 production. Longer APRTs (> 476 20 hour) under typhoon influence were far from sufficient to maintain high contributions of local emissions for O3 pollution 477 (which decreased by ~20% for the polluted regions of the PRD in July 2016). However, due to the variations of wind fields 478 under different scenarios, the changes of local and transport contributions for O3 led by typhoons were different in the 479 southwest part of the PRD, that is, higher contribution from emissions within the PRD and reduced transport contribution 480 23 occurred in the typhoon-induced scenarios in both seasons. As for the close typhoon-induced scenario, O3 transport was further 481 strengthened, but meteorological conditions in the PRD became less favourable for both the production and accumulation of 482

486
The East Asian monsoon, changing with seasons, has a pronounced impact on local meteorological conditions as well as the 487 characteristics of O3 pollution in East China (He et al., 2008). The seasonal behaviour of the East Asian monsoon is likely to 488 result in the seasonally varied effect of typhoons on O3 pollution in the PRD. In October, the summer monsoon has almost 489 finished its retraction and the winter monsoon is beginning (Ding, 1994). Thus, there are not many obstacles to the southward 490 movement of typhoon periphery and the transport of O3 towards the PRD by the continental and coastal air masses. Large 491 downdraft-influenced areas in Central and South China occur in this scenario, and high O3 levels and low RH in these areas 492 indicate the potentially important role of directly downward O3 transport. In July, the summer monsoon reaches its strongest 493 (Ding, 1994). The interaction between typhoon periphery and the summer monsoon results in stagnation and enhanced updrafts 494 above the land areas of the PRD and its surroundings. Only when typhoon is close enough to the PRD is the stagnation 495 terminated and the structure of the summer monsoon cell broken. This also explains why some summertime typhoon-induced 496 O3 episodes in the PRD can be typically divided into two periods, as stagnation leads to the accumulation of locally produced 497 O3 in the first phase and strong northerly winds strengthen O3 transport before the landing of typhoons (Lam et al., 2005;Li, 498 2013). It should be noted that updrafts, rather than downdrafts, prevailed on the typhoon-induced O3 pollution days in summer. 499 High levels of O3 did not necessarily occur in the regions dominated by downdrafts in this scenario, suggesting a less notable 500 connection between downdrafts and summertime O3 pollution in the PRD. Further investigations are required to trace the 501 detailed process of downward O3 transport, including the stratosphere-troposphere exchange (Stohl et al., 2003), in each 502 scenario, and quantify their contributions to near-ground O3 pollution. 503 504 24 Some limitations remain in this study. We chose O3 pollution days as individual samples, ignoring the influence of O3 pollution 505 on the previous days. Thus, more detailed full-episode analyses are required. Moreover, owing to the small sampling size, the 506 influence of typhoons on O3 pollution in the PRD is still not fully understood, including, for instance, the detailed connections 507 between the features of typhoons (intensity, position) and O3 pollution. However, the comparisons of meteorological conditions, 508 O3 processes and sources in different scenarios and seasons demonstrate the complex causes of typhoon-induced O3 pollution 509 in the PRDtyphoons tend to enhance O3 transport into the PRD in both seasons, but their impacts on the production and 510 accumulation of O3 are completely different. As a result, emissions within (outside of) the PRD are likely to contribute less 511 (more) on the typhoon-induced O3 pollution days than on the no-typhoon days, and more attention should be paid to controlling 512 anthropogenic emissions of O3 precursors on a larger scale under typhoon influence. In order to effectively alleviate O3 513 pollution and to reduce the population exposure in the PRD, more attention should be paid to controlling anthropogenic 514 emissions of O3 precursors on a larger scale, rather than focusing on local emission, under typhoon influence. For air quality 515 management, it is suggested to comprehensively evaluate the efficiency of fractional local and non-local emission reductions 516 to reduce O3 levels in the PRD in different scenarios (Thunis et al., 2019;Thunis et al., 2020). This study also suggests that a 517 thorough evaluation of O3 transport, production and accumulation conditions can be applied to understand the causes of 518 regional O3 pollution not only in the PRD, but also in other regions. The results will help find efficient strategies to alleviate 519 regional O3 pollution as well as to reduce its adverse effects.