Air quality deterioration episode associated with typhoon over the complex topographic environment in central Taiwan

Abstract. Air pollution is typically at its lowest in Taiwan during summer. The mean concentrations of PM10, PM2.5, and daytime ozone (08:00–17:00 LST) during summer (June–August) over central Taiwan are 35–40 µg/m3, 18–22 µg/m3, and 30–42 ppb, respectively, between 2004 and 2019. Sampling analysis revealed that the contribution of organic carbon (OC) in PM2.5 could exceed 30 % in urban and inland mountain sites during July in 2017 and 2018. Frequent episodes of air quality deterioration occur over the western plains of Taiwan when an easterly typhoon circulation interacts with the complex topographic structure of the island. We explored an episode of air quality deterioration that was associated with a typhoon between 15 and 17 July 2018, using the Weather Research Forecasting with Chemistry (WRF-Chem) model. The results indicated that the continual formation of low-pressure systems or typhoons in the area between Taiwan and Luzon island in the Philippines provided a strong easterly ambient flow, which lasted for an extended period between 15 and 17 July. The interaction between the easterly flow and Taiwan’s Central Mountain Range (CMR) resulted in stable weather conditions and weak wind speed in western Taiwan during the study period. Numerical modeling also indicated that a lee side vortex easily formation and the wind direction could be changed from southwesterly to northwesterly over central Taiwan because of the interaction between the typhoon circulation and the CMR. The northwesterly wind coupled with a sea breeze was conducive to the transport of air pollutants, from the coastal upstream industrial and urban areas to the inland area. The dynamic process for the wind direction changed given a reasonable explanation why the observed SO42− became the major contributor to PM2.5 during the episode. SO42− contribution proportions (%) to PM2.5 at the coastal, urban, and mountain sites were 9.4 µg/m3 (30.5 %), 12.1 µg/m3 (29.9 %), and 11.6 µg/m3 (29.7 %), respectively. Moreover, the variation of the boundary layer height had a strong effect on the concentration level of both PM2.5 and ozone. The combination of the lee vortex and land-sea breeze, as well as the boundary layer development, were the key mechanisms in air pollutants accumulation and transport. As typhoons frequently occur around Taiwan during summer and fall, and their effect on the island’s air quality merits further research attention.


northwesterly wind coupled with a sea breeze was conducive to the transport of air 20 pollutants, from the coastal upstream industrial and urban areas to the inland area. The 21 dynamic process for the wind direction changed given a reasonable explanation why the 22 observed SO4 2became the major contributor to PM2.5 during the episode. SO4 2-23 contribution proportions (%) to PM2.5 at the coastal, urban, and mountain sites were 9.4 24 µg/m3 (30.5%), 12.1 µg/m3 (29.9%), and 11.6 µg/m3 (29.7%), respectively. Moreover, 25 the variation of the boundary layer height had a strong effect on the concentration level 26 of both PM2.5 and ozone. The combination of the lee vortex and land-sea breeze, as well 27 as the boundary layer development, were the key mechanisms in air pollutants 28 accumulation and transport. As typhoons frequently occur around Taiwan

Introduction: 39
Tropical cyclones (also known as typhoons) are a frequent occurrence in East Asia 40 during summer and fall. Typhoons significantly affect not only meteorological parameters 41 but also air quality. That is because air pollution is strongly related to atmospheric 42 conditions, and typhoon circulation typically alters atmospheric stability and air pollutant 43 diffusion in specific locations. For example, researchers revealed that ozone episodes in 44 Hong Kong and southeastern China are strongly related to the passage of typhoons as 45 they approach the area (Lee et al., 2002;Ding et al., 2004;Huang et al., 2005 and46 Yang et al., 2012;Zhang et al., 2013;Zhang et al., 2014;Wei et al., 2016;Yan et al. 2016;47 Luo et al. 2018;Deng et al. 2019;Hung et al., 2021). The stagnant meteorological 48 conditions associated with strong subsidence and stable stratification in the boundary 49 layer results in pollutant accumulation before typhoons make landfall. Huang et al. (2005) 50 reported that approximately 30% of total ozone in Hong Kong was due to local chemical 51 production in the lower atmospheric boundary layer, and approximately 70% was 52 contributed by long-range transport from southern China (i.e., the Pearl River Delta). 53 According to the dynamic process perspective, Chow et al. (2018) reported frequent high-54 O3 days when typhoons were located between Hong Kong and Taiwan (Fig. 1a)  Taiwan also experiences air quality deterioration as typhoons approach (Feng et al. 58 2007;Chang et al., 2011, Cheng et al., 2014Hsu and Cheng, 2019). However, not all 59 typhoons are associated with poorer air quality in Taiwan. The effect of typhoons on air 60 quality is highly related to the location of the typhoon and its circulation's interaction 61 with Taiwan's Central Mountain Range (CMR; Fig. 1b). Thus, the mechanism of the 62 formation of poor air quality may differ between Taiwan and Hong Kong. Air quality 63 deterioration frequently occurs over the western plains of Taiwan when typhoons pass 64 between Taiwan and Luzon island in the Philippines; the distance of the typhoons from 65 Taiwan is typically several hundred kilometers but may even be greater than 1000 66 kilometers. Under such conditions, the weather is typically stable, with clear skies, strong 67 solar intensity, and weak wind speeds over Taiwan's western plains because of the 68 interactions of the typhoon's easterly circulations with the CMR. Furthermore, such 69 typhoons are usually associated with a Pacific high-pressure system during summer; thus, 70 monitoring stations. To elucidate the spatial distribution of air pollutants, we classified 116 the observed stations over central Taiwan into "coast," "urban," and "mountain." Each 117 of these categories represents the mean concentration of the numbers derived from 118 stations of the same type. The coast category included two stations: Shalu (SL) and 119 Xianxi (XX; Fig. 1c). The urban category included five stations: Fengyuan (FY), Xitun 120 (XT), Zhongming (ZM), Changhua (CH), and Dali (DL; Fig. 1c). The mountain 121 category included three stations: Nantou (NT), Zhushan (ZS), and Puli (PL), which 122 were located nearby or in basins surrounded by high mountains (Fig. 1c). Two stations 123 on small islands were also considered in the analysis. One was in Kinmen (KM), which 124 is located close to Xiamen city in southeast China, and the other was Magong (MG) 125 station located in the Taiwan Strait (Fig. 1a). 126 To explore the air pollution episodes during summer, we recorded data in central 127 Taiwan   respectively. The remaining two sampling sites, SL and CSM, were located in a coastal 134 suburban and urban area (Fig. 1c), respectively. The sampling period of each sample 135 was 12 h; daytime samples were collected from 08:00 to 19:00 LST, whereas nighttime 136 sampling was conducted from 20:00 LST to 07:00 LST. We determined mass 137 concentrations of the aerosols using a gravimetric measurement of the samples 138 collected on polytetrafluoroethylene membrane filters (Chou et al. 2008 MG TEPA station (Fig. 1a). 141 During summer, the land-sea breeze easily combines with mountains' up/down 142 slope wind during daytime/nighttime. As the sea breeze develops, air flows are typically 143 transported from coastal areas and pass over the Taichung metropolitan region (Fig. 1c) 144 coupled with mountain slope flow to the inland area. The Taichung metropolis is a large 145 urban environment comprising residential, industrial, and agricultural lands (Cheng et 146 al., 2009). In particular, Taichung Power Plant (TPP, Fig. 1c), which is coal-fired, and 147 the Taichung Harbor Industrial (THI, Fig. 1c) zone are both located on the coast and 148 are responsible for substantial emissions in central Taiwan. Thus, severe emission 149 sources contribute to and affect the air quality in the Taichung metropolitan area under 150 favorable weather conditions. For detailed information on the instruments used in the 151 sampling analysis, please refer to Lee et al. (2019). Meteorological parameters, 152 including wind speed and wind direction, temperature, and relative humidity were 153 acquired from a meteorological station in the same location where data were collected 154 for this study. 155 156

Model configurations 157
In this study, we used the Weather Research and Forecasting model (WRF) 158 coupled with the WRF-Chem version 3.9 to study the air pollutants transport during 159 the episode. We obtained the meteorological initial and boundary conditions for 160 were well simulated, we employed the four-dimensional data assimilation scheme 167 according to the NCEP-GFS data. Transport processes included advection by winds, 168 convection by clouds, and diffusion by turbulent mixing. Removal processes included 169 gravitational settling, surface deposition, and wet deposition (scavenging in 170 convective updrafts and rainout or washout in large-scale precipitation). The kinetic 171 preprocessor (KPP) interface was used in both the chemistry scheme of the Regional 172 Atmospheric Chemistry Mechanism (Stockwell et al., 1990). The secondary organic 173 aerosol formation module, the Modal Aerosol Dynamics Model for Europe (MADE) 174 (Ackermann et al., 1998)/Volatility Basis Set (VBS) (Ahmadov et al., 2012) was 175 employed in the WRF-Chem model.  The daytime ozone peaked at 56 ppb and 48 ppb in October and April, respectively (Fig.  187 2c). For PM10 and PM2.5, the peak concentrations were 70-75 µg/m 3 and 40-45 µg/m 3 188 over the western plains in March (Fig. 2a, b). Regarding the characteristics of ozone 189 distribution, the concentration at the mountain site was typically higher than that in 190 urban areas and the coast. For PM10 and PM2.5, the mountain site also typically had 191 higher concentrations than did the urban and coastal areas, except during summer (Fig. 192 2a,b). The monsoon dominates the prevailing wind over East Asia. During summer, a 193 southwesterly wind prevails, whereas a northeasterly wind prevails during fall, winter, 194 and spring. The characteristics of the seasonal variations might be due to the summer 195 having a cleaner background and higher boundary layer height than those in other 196 seasons. As mentioned earlier, the major emission sources such as industry and traffic 197 are located in coastal and urban areas. The mean highest concentration of ozone 198 typically occurs over rural mountain areas during summer; thus, the dominant land-sea 199 breeze might play a critical role in the air quality in western Taiwan. 200 During summer (July only in this study) in 2017 and 2018, we conducted sampling 201 campaigns in central Taiwan. Table 1  The inland rural mountain site, ZS, clearly had the highest total PM2.5 concentration. 205 Organic carbon (OC) and SO4 2had the highest concentrations of the species in PM2.5, 206 and both increased from the coast to the inland mountain area (Table 1) PM2.5 were OC, SO4 2-, NO3 -, NH4 + , and elemental carbon (EC; Table 1). At the coastal 210 station SL, the concentrations of OC and SO4 2were comparable at 4.3 µg/m 3 and 4.5 211 µg/m 3 , accounting for 27.5% and 28.6% of PM2.5, respectively. At the city site CSM 212 and the inland rural mountain station ZS, OC had concentrations of 5.6 (33.1% of PM2.5) 213 and 6.6 µg/m 3 (30.9% of PM2.5), respectively. The results indicated that the contribution 214 of OC in PM2.5 could exceed 30% at the urban and inland mountain sites. The 215 concentration of OC increased from the coast (4.3 µg/m 3 ; 27.5% of PM2.5) to the 216 mountain station (6.6 µg/m 3 ; 30.9% of PM2.5), and the urban site had the highest 217 proportion (5.6 µg/m 3 ; 33.1% of PM2.5) in PM2.5 among these stations (Table 1). SO4 2-218 also exhibited an increased concentration from coastal areas to the inland mountain area, 219 but the changes were minor (4.5-4.8 µg/m 3 ). Notably, the proportion of SO4 2in PM2.5 220 decreased from the coast to the mountain area because the major sources, TPP and THI 221 ( Fig. 1c), are located on the coast. The other species, namely NO3 -, NH4 + , and EC, at 222 SL, CSM, and ZS had comparable concentrations between stations (1.0-1.4, 1.7-2.0, 223 and 1.1-1.4 µg/m3, respectively; Table 1). The inland rural station ZS was located in a 224 foothill valley of the CMR and surrounded by mountains. Thus, the high concentration 225 at ZS might be due to sea breeze transport. 226 In general, OC and SO4 2were the major species over western Taiwan, especially 227 in inland areas. These results suggest that local contribution, such as traffic, industry, 228 and even agricultural emissions, might play critical roles in the composition of PM2.5. 229 Furthermore, the spatial distributions of highest PM2.5 and daytime ozone concentration 230 were not always in urban areas; instead, concentrations accumulated in inland rural 231 areas ( Fig. 2 and Table 1). The roles that the land-sea breeze, boundary layer 232 development, and interaction of typhoon circulation with complex geographic 233 structures play in air quality require clarification. The mechanism of these complex

Weather condition and observation 238
To explore air quality deterioration processes and formation mechanisms, we 239 Taiwan to an easterly direction for an extended period between 15 and 17 July ( Fig. 3a-249 c). The easterly ambient flow was easily blocked by Taiwan's CMR, resulting in a lee 250 vortex formation associated with stable atmospheric conditions and weak wind speed 251 in western Taiwan. The mechanism of lee vortex formation on the lee side of a high 252 mountain has been described through a laboratory experiment (Hunt and Synder, 1980) 253 and numerical modeling (e.g., Smolarkiewicz and Rotunno, 1989). Li and Chen (1998)  H is the height of an obstacle), and the low-level airflow easily split off the northern 257 coast and moved around the island of Taiwan. The current study is an example of a low 258 Fr case (<0.5; assumed average wind speed, U = 10 ms -1 ; Brunt-Vaisala frequency, N 259 = 10 -2 s -1 ; and average mountain height, H = 2.5 km). Thus, we expected wind speeds to 260 be weak and atmospheric conditions to be more stable on the lee side of the CMR 261 compared with the windward side of eastern Taiwan. 262 Sounding data (Fig. 4) recorded at the CWB station in Penghu island (46734,263 close to MG in Fig. 1a) indicated a relatively weak wind speed (<5 m/s) in the low 264 boundary (below 850 hPa) during the study period from 15 to 17 July 2018 ( Fig.4a-c). 265 Above 700 hPa (3000 m), a strong easterly wind (>10 m/s) prevailed due to the typhoon 266 circulations. Furthermore, clear subsidence and multiple inversion layers were revealed 267 in the sounding between 16 and 17 July (Fig. 4b,c). On 17 July, the inversion layer was 268 even lower than 950 hPa (Fig. 4c); that is, only a few hundred meters over Penghu 269 island in the Taiwan Strait. The sounding data revealed stable atmospheric conditions, 270 high relative humidity, and weak wind speed on the leeside of the mountains over 271 western Taiwan. 272 The wind speed at MG was weaker than that at KM because MG is close to Taiwan and 283 was likely affected by the mountain blocking effect mentioned earlier. Because the wind 284 speed did not change considerably, the PM2.5 and O3 concentration levels did not 285 fluctuate obviously at MG during the study period. coastal, urban, and mountain sites. Peak PM2.5 at the coastal and urban sites was 293 observed around noon, whereas peak PM2.5 at the inland mountain site occurred at 18:00 294 LT on 17 July 2018 (Fig. 5b). The differences in the timing of the peak PM2.5 295 concentrations between the coastal and urban sites and the inland mountain site could 296 be attributed to the transport of the sea breeze. No clear diurnal variation in PM2.5 297 concentration was observed between the urban and mountain sites between 16 and 17 298 July. That is, even at night and in the early morning, the concentration remained as high 299 as 40 µg/m 3 (Fig. 5b) because atmospheric conditions were favorable for air pollutant 300 accumulation. The peak ozone concentration occurred around noon at the coast and 301 urban sites, whereas the peak at the mountain site occurred later at 16:00 LST (Fig. 5c). 302 We estimated that the concentrations of PM2.5 and ozone on the episode day on 17 July 303 (Fig. 5b,c) were three times higher than the mean concentration during summer (Fig. 2) 304 in central Taiwan. As mentioned earlier, the major emissions were generated by coastal 305 industry and the Taichung city metropolitan area, but the peak ozone concentration 306 occurred at the inland mountain station (120 ppb at PL) because of sea breeze transport 307 from upstream to downstream sites. 308 Spatial distribution of wind field and PM2.5 concentration (Fig. 6) (Fig. 6a-f). Over western Taiwan, a sea breeze developed 313 after 10:00 LST, and a strong onshore flow blew air pollutants to the inland area( Fig.6b-314 d). A high PM2.5 concentration (>50 µg/m 3 ) extended from the coast to the urban area 315 at noon (Fig. 6b-c), which was subsequently transported to the inland mountain area in 316 the afternoon and nighttime (Fig. 6d-f). The high PM2.5 concentration accumulated in 317 Maoli county (located north of Taichung city) at midnight owing to the convergence of 318 southerly and land breeze (Fig. 6f). Actually, the spatial variation of PM2.5 could also 319 be observed on the previous day (16 July; Fig. 5b), which contributed approximately 320 30 µg/m 3 in the early morning on 17 July in central Taiwan. 321 The location of the high-pollution ozone was also strongly associated with the 322 land-sea breeze during the daytime (Fig. 7 b-e). A high concentration of ozone was 323 observed at the urban station at noontime (Fig. 7c); the ozone was transported to the 324 inland mountain station, resulting in peak concentrations higher than 120 ppb between 325 16:00 and 18:00 LST (Fig. 7d-f). By 22:00 LST, the ozone concentration had declined 326 more rapidly in the city than in the mountain area because of the dilution effect ( Fig. 7  327 g-h). The detailed pollution process and mechanism are demonstrated and discussed in  The hourly comparison between observed (red solid) and simulated (blue dashed) 331 PM2.5 and ozone between 12 and 18 July 2018 are presented in Fig. 5b,c. In general, 332 our simulation reasonably captured the variation of PM2.5 and ozone in western Taiwan  333 and small island sites, MG and KM (Table 2). For PM2.5, the root mean square error 334 (RMSE) at all sites was less than 1.0 µg/m 3 , and the correlation between observed and 335 simulated values was 0.72 and 0.81 at the urban and mountain sites, respectively. 336 Regarding the mean bias of PM2.5, it was slightly overestimated at coastal and urban 337 sites and underestimated at the mountain site and sites on the two islands. In the ozone 338 simulation, the correlation between observed and simulated values was as high as 0.73-339 0.9, except for MG. The RMSE of ozone for all areas was less than 1.45 ppb. For the 340 mean bias of ozone, the maximum underestimation (−10 ppb) occurred at the coastal 341 site, and the maximum overestimation (13.8 ppb) occurred over the mountain area 342 because of the simulation of the spatial distribution difference. 343 (08:00-16:00 LST) on 17 July 2018. After 08:00 LST, the sea breeze gradually 353 developed and the onshore wind speed increased (Fig. 8a-c); thus, the high-354 concentration PM2.5 plume was transported from the coast to the inland mountain area. 355 Even though the area has high emissions, the PM2.5 concentration along the coastal area 356 of China was low because of the strong wind speed (Fig. 8a-c). As sea breeze developed 357 after 08:00 LST, and the vortex circulation was coupled with the onshore flow ( Fig. 8a-358 d). The lee vortex circulation was not clear because it combined with the sea breeze and 359 enhanced the air pollutant transport to the inland area during the daytime. However, the 360 lee vortex circulation was clearly formed in the area from 23.5 to 24.5 °N in the 361 afternoon until early morning on the next day because the land breeze interacted with 362 the mountain lee-side flows (Fig. 8e-f). After the lee vortex circulation formed, the 363 southerly flow in the western plain was enhanced (Fig. 8e-f) and 17 over central Taiwan (Figs. 5a,b, and 6f). Thus, the lee vortex formation was 369 adverse to the development of the offshore flow (land breeze) and prolonged the air 370 pollutant accumulation in western central Taiwan ( Fig. 6 and 8). These critical 371 processes explain why air pollutants tended to accumulate in central Taiwan during the 372 episode days. Notably, the wind speed was strong and the concentration of PM2.5 was 373 low in the Taiwan Strait close to coastal areas of China in the simulation (Fig. 8a-f) and 374 according to observations at KM (Fig. 5a). According to the spatial distribution, a strong 375 wind speed can limit the number of air pollutants transported southward from mainland 376 China to Taiwan (Fig. 8b-f). That is, the pollution type was locally dominated during 377 the event days. 378 Similar to the observed zone (Fig.7), the simulated ozone ( Fig.9) was also 379 dominated by circulations associated with the land-sea breeze and the interaction of the 380 easterly flow with the CMR. Most of the area had steady low concentrations in the early 381 morning on 17 July (Fig. 9a) because of the dilution effect of the ozone formation in 382 the nighttime and early morning ( Fig. 9a and h-i). A high concentration already existed 383 over the mountain area in Miaoli County (Fig.1b) in the early morning at 04:00 LST 384 western plains in the afternoon (Fig.9 c-f) on 17 July. The area of high ozone 388 concentration extended over the western plains when the sea breeze developed after 389 10:00 LST on 17 July (Fig. 9c). Following increases in wind speed, the high ozone 390 concentration extended to the inland area and was transported further south of Taichung 391 City ( Fig. 9 d-e). The peak ozone concentration at the inland rural site occurred at 16:00 392 LST, whereas it occurred in the city center at the urban site at 12-14:00 LST (Figs. 5c;393 7c,d;9d,e). Because the major emission sources were coastal industry and the urban 394 area, the high ozone concentration at the inland site was the result of ozone being 395 transported by the sea breeze. The simulated peak ozone concentration occurred 396 between 14:00 and 16:00 LST at the inland site because of the sea breeze coupled with 397 the mountain upslope wind (Fig.9 c-f). Moreover, the high-ozone plume was associated 398 with the lee vortex circulation over the Taiwan Strait and provided a southerly flow 399 component during the nighttime and early morning (Fig. 9a, and g-i). 400 As mentioned earlier, sounding data indicated multiple inversion layers on the 401 event days. To further investigate the boundary layer development and air pollutant 402 distribution in the vertical, a northwest-southeast cross-section AA' (Fig.10a) was 403 superimposed over the high concentration area, as illustrated in Fig. 10. In the early 404 morning at 05 LST (Fig.10b), a separate high-concentration plume was observed at 405 ground level and another remained at an elevation of 1000 m on 17 July. It is a typical 406 https://doi.org/10.5194/acp-2021-204 Preprint. Discussion started: 8 April 2021 c Author(s) 2021. CC BY 4.0 License. boundary layer structure due to ground surface radiation cooling under stable 407 atmospheric conditions during nighttime and early morning. These two layers' plume 408 coupled together due to boundary layer gradually developed in the morning after 0700 409 LST (Fig.10 b-d). Because the emissions increased during rush hour, the concentration 410 promptly increased as the PM2.5 plumes of these two layers coupled well in the vertical 411 below 1000 m at 10:00 LST (Fig. 10 d). The wind speed was weak at elevations below 412 1500 m but strong and offshore in a southeast-northwest direction above 2000 m due to 413 easterly tropical cyclone circulation. The high-PM2.5 plume (concentration > 50 µg/m 3 ) 414 was pushed by the sea breeze coupled with the upslope wind and accumulated in the 415 inland rural area during daytime (12:00-16:00 LST) (Fig. 10e-g). The highest 416 concentration was not at ground level but heights between 500 and 1000 m at noontime 417 ( Fig.10e) and 1000-1500 m in the afternoon (Fig. 10 f-g). The boundary layer structure 418 and the coupled between sea breeze and mountain upslope wind played important roles 419 for the PM2.5 concentration distribution in the vertical along the cross-section (Fig.10d-420 g). As offshore wind developed, which pushed the air pollutants from the mountain area 421 to the plain and coastal area (Fig. 10 g-i), and the elevation of the plume was 422 predominantly between 500 and 1500 m after 20:00 LST. The discussion above of the dilution effect in the early morning at 04:00 LST (Fig. 11a) on 17 July. However, 429 a high-ozone layer was observed between 500 and 1500 m because of the previous 430 day's contribution. After 08:00 LST, the mixing layer developed, and emissions from 431 traffic and industry also increased. Concurrently, both the onshore sea breeze over the 432 plain and the upslope wind over the mountain developed; thus, wind speed also 433 enhanced in the low boundary (Fig. 11b-e). The sea breeze and weak wind speed also 434 exacerbated the high-concentration ozone in the inland area during the daytime (Fig.  435 11c-f). At nighttime, the ozone concentration gradually decreased because of the 436 dilution effect below 500 m ( Fig. 11h-i). However, TEPA measurements revealed that 437 a layer with high ozone concentration remained between 1000 and 1500 m (Fig. 7g-h

Discussion: 448
The wind direction over Taiwan during summer is mostly southerly to 449 southwesterly (Table 1). However, the wind direction during the episode was westerly 450 to northwesterly ( Table 2). The wind direction changed because of the critical 451 interaction between typhoon circulations and the CMR. Moreover, the concentration of 452 PM2.5 and its composition during the episode also differed significantly from the 453 monthly mean, as revealed in Table 2. A substantial increase in daily mean PM2.5 was 454 observed at all sites, especially at the CSM site (urban), where concentration increased 455 from 16.9 to 40.5 µg/m 3 (Table 2). Furthermore, SO4 2became the dominant species in 456 PM2.5 from the coastal to the mountain area, ranging from 30.5 to 29.7% during the 457 episode. The SO4 2concentration during the episode (Table 2) was more than twice that 458 of the monthly mean (Table 1) in the Taichung area. This variation was due to the wind 459 direction changing from southwesterly to northwesterly, resulting in a contribution 460 increase from the upstream TPP and THI (Fig. 1c), which are the major sources in 461 central Taiwan. 26.5°C, respectively. Thus, the daily mean temperature during the episode period was 470 1-2 °C higher than is typical for days in July. In general, the mean wind speed on the 471 episode days at these three sites was weaker (<1 m/s) than the monthly mean (Tables 1  472 and 2). Such stable weather conditions, weak wind speed, and high air temperature were 473 conducive to the generation and formation of a secondary aerosol. This is exemplified 474 by the concentrations of other species, such as OC, NO3and NH4 + , being considerably 475 higher during the episode days (Table 2) compared with the monthly mean in Table 1. 476 Notably, EC increased to a lesser extent than did the other species. These results suggest 477 that secondary aerosol plays a critical role under such stable weather conditions and 478 wind direction. Because ambient wind changes during typhoon formation between 479 Taiwan and Luzon island in the Philippines are not uncommon, the air quality impacts 480 in such weather conditions merit further research. A detailed discussion of variations in 481 aerosol chemical composition transformation will be presented in a separate paper.

Summary: 483
The lowest air pollution levels in Taiwan  weather condition, concentrations of PM2.5 and ozone could be higher than 2 times of 493 those monthly mean. During the episode, SO4 2became the major contributor to PM2.5, 494 and its concentration and contribution proportion (%) in PM2.5 at coastal, urban, and 495 mountain sites were 9.4 µg/m 3 (30.5%), 12.1 µg/m 3 (29.9%), and 11.6 µg/m 3 (29.7%), 496 respectively. It is due to the northwesterly wind was conducive to the transport of SO2 497 and sulfate from the coastal upstream major emission sources (areas in TPP and THI) 498 to the inland area. 499 To explore the mechanism of air pollution formation, we conducted a detailed (1) First, typhoon circulations provided a strong easterly ambient flow. This easterly 514 flow interacted with the CMR, resulting in a lee vortex formation over western 515 Taiwan. (Fig.12, left panel) 516 (2) During the nighttime, the offshore wind that developed pushed the air pollutants (3) In the morning, this residual layer with polluted air mass combined with and 527 contributed to the ground surface air concentration level because the boundary 528 layer height increased. This also explains why the ozone and PM2.5 concentrations 529 dramatically increased after the boundary layer development during the daytime. 530 For this reason, the high-concentration ozone plume was located in a low-531 emission mountain area and the episode occurred at an earlier time than in the 532 plain area where the major emission sources are located. 533 During the daytime, the lee vortex flow coupled with a sea breeze and combined 534 with a mountain upslope wind; this resulted in the accumulation of air pollutants 535 in the inland mountain area. Furthermore, because of the mountain upslope flow, 536 the high PM2.5 and ozone concentrations were located not at ground level but at 537 heights between 500 and 1000 m. The peak concentration at the inland mountain 538 site occurred approximately 4-6 hours later than at the upstream coastal site