Figure 3a shows the time series of autumn MDA8 ozone concentration over the SEC from 2014 to 2024 (All Days, black line), with the blue and grey lines representing the average for days with TC (TC Days) and without TC (Non-TC Days), respectively. The interannual variation of the mean MDA8 ozone concentration reveals a fluctuating but overall increasing trend, with peak values occurring in 2019 and 2022. Throughout this period, ozone levels on TC Days consistently exceeded those on All Days, particularly in 2022 when the difference reached 8 ppb. In contrast, Non-TC Days exhibited relatively lower ozone concentration without a significant linear trend over the 11 years period. These results demonstrate that TC activity is closely associated with elevated autumn ozone concentration in SEC.

After classifying TCs within 105–140° E and 10–40° N by track type, clear differences emerge in the interannual variability of SEC ozone among the three TC types (Fig. 3a). Ozone on westward TC days (yellow dashed line) show the largest year-to-year fluctuations, with relatively low values occur in 2014 and 2018 but much higher values in 2015, 2019, and 2022, indicating strong case-to-case variability in the ozone response to westward TCs. In contrast, ozone on landfalling TC days (green dashed line) is generally lower and varies only weakly from year to year, remaining near 43–46 ppb in most years. In several years, no statistics are available because no landfalling TCs occurred. This suggests that, although landfalling TCs can also be associated with elevated ozone episodes, their effect on regionally averaged ozone is comparatively limited. By comparison, ozone on northward TC Days (red dashed line) stays at a high level in most years and is overall higher than that for the other two track types, especially in 2019, 2022, and 2024, with the highest value occurring in 2022 (about 63.66 ppb). This indicates that northward TCs show the strongest association with autumn high-ozone conditions over the SEC, and that their interannual variability agrees better with the years of elevated regional ozone.

To directly illustrate how the three TC track types affect ozone over SEC during their movement, Fig. 3b–d show the TC positions for each TC type together with the corresponding ozone levels over SEC. The marker sizes and colors represent the intensity of TCs and the MDA8 ozone levels in SEC corresponding to the positions of TCs, respectively. Results show that westward TCs strengthen as they approach the southern side of the SEC, accompanied by varying degrees of ozone concentration enhancement. When these TCs make landfall between 12–18° N subsequently, the ozone concentration across the entire SEC remains at a relatively high level. However, landfall over Hainan (southern area of SEC) induces a clear ozone reduction in SEC (Fig. 3b). For landfalling TCs, both intensity and ozone concentration in SEC initially increase then decrease during their approach, though overall ozone levels remain relatively low among the three TC types (Fig. 3c). During the period of the northward TC influence (Fig. 3d), ozone concentrations in SEC frequently exceeded 50 ppb, significantly higher than that of the other TC types. These findings demonstrate that TC intensity also plays an important role in influencing the ozone concentration of SEC, in addition to track type and location.

To further investigate the influence of TC intensity on ozone in SEC, Fig. 4 presents the ozone concentration across different TC intensity categories for each TC type. For westward TCs (Fig. 4a), both mean and median ozone concentration are obviously higher when TCs reach STY or SSTY intensity, suggesting that approaching but not yet landfalling TCs are more conducive to ozone formation and accumulation in SEC. Landfalling TCs (Fig. 4b) show no clear ozone concentration trend with increasing intensity, with the mean ozone concentration at STY being even lower than that at TY. In contrast, northward TCs (Fig. 4c) maintain higher ozone concentration than the other two types. Particularly when TCs reach TY or higher intensity, the mean, median, and 85th percentile of ozone concentration all reach their highest levels. The findings above collectively identify intense northward TCs as a key contributor to elevated autumn ozone levels in SEC.

In addition, considering that the influence of northward TCs on SEC may vary substantially with TC location (Hu et al., 2023), we further divided the activity domain of northward TCs into six subregions at 10° latitude–longitude intervals and calculated, for each subregion, the number of TC influence days and the corresponding mean ozone concentration, in order to identify the most representative area affected by northward TCs. According to the table in the lower-left corner of Fig. 3d, although the two northern subregions exhibit greater ozone enhancement, their sample sizes are much smaller. By contrast, the 120–130° E and 20–30° N subregion has the largest number of TC influence days (64 d) and a relatively high mean ozone concentration (55 ppb), indicating the strongest combined influence in terms of both frequency and ozone level. As shown in Fig. 4d, the mean ozone concentration in this key subregion exceeds 50 ppb at nearly all intensities and reaches a maximum at TY. Unlike the broader-domain result shown in Fig. 4c, ozone decreases at STY and SSTY in this key subregion. This suggests that, when stronger TCs occur within this subregion, the accompanying meteorological changes may shift SEC ozone from an increasing to a decreasing tendency. On this basis, the following section focuses on how northward TC intensity affects ozone over SEC when the TC is located within the key subregion of 120–130° E and 20–30° N.

To further clarify how circulation changes associated with northward TCs within the key subregion at different intensities modulate local meteorological conditions and thereby affect ozone over SEC, we examined the daytime (08:00–18:00) average characteristics of meteorological variables at each intensity (Fig. 5) together with the corresponding atmospheric circulation anomalies relative to the autumn climatology during 2014–2024 (Fig. 6). The variables analyzed in Fig. 5 include total cloud cover, surface solar radiation downwards, 1000 hPa relative humidity, total precipitation, 2 m temperature, boundary layer height, 850 hPa vertical velocity, and 10 m wind speed. In Fig. 6, shading denotes 500 hPa geopotential height anomalies, wind arrows indicate 850 hPa wind anomalies, and the dark red dashed lines mark the mean position of the WPSH. It should be noted that, because TCs usually continue to develop and intensify before reaching the key subregion, the number of days in the TD, TS, and STS is relatively small. Accordingly, the fluctuations shown for these categories may partly reflect sampling variability rather than robust physical differences associated with TC intensity.

From TD to TS (Fig. 6a–b), positive 500 hPa geopotential height anomalies dominate over SEC and the regions to its north, while the anomalous winds at 850 hPa remain relatively weak. This indicates that the mid-tropospheric circulation is still mainly controlled by a relatively stable weak anticyclonic background. It is worth noting that the WPSH splits into a double-center structure under the intrusion of the northward TC from TS, and this feature also persists in other periods (Fig. 6b–f). Under this circulation configuration, solar radiation and 2 m temperature increase slightly, but total cloud cover (0.63) and total precipitation (0.26 mm h⁻¹) remain relatively high (Fig. 5), and ozone therefore changes little (Fig. 4d). At STS, weak negative 500 hPa geopotential height anomalies emerge over SEC, and the anomalous cyclonic circulation to its east becomes more distinct at 850 hPa, suggesting a strengthened influence of TC-induced disturbances. Although total cloud cover and 1000 hPa relative humidity decrease to 0.53 and 67.77 %, respectively, surface solar radiation and 2 m temperature also drop markedly to 1.27 MJ m⁻² and 22.54 °C, while total precipitation remains relatively high at 0.26 mm h⁻¹. As a result, ozone decreases to a relatively low level of 47.5 ppb.

When TC intensifies to TY, a pronounced center of negative 500 hPa geopotential height anomalies develops over the coastal ocean east of SEC, accompanied by a well-defined anomalous cyclonic circulation at 850 hPa. SEC is located on its western flank and is influenced by a more evident anomalous northerly flow, indicating that the study region comes under stronger control of the TC peripheral circulation (Fig. 6d). Under these circulation conditions, the 850 hPa vertical velocity changes from negative (-0.02 Pa s⁻¹) to positive (0.01 Pa s⁻¹), favoring suppressed convection and reduced cloud and rainfall activity. Correspondingly, total cloud cover, 1000 hPa relative humidity, and total precipitation decrease substantially from 0.53, 67.77 %, and 0.26 mm h⁻¹ at STS to 0.34, 62.69 %, and 0.10 mm h⁻¹, respectively. At the same time, solar radiation, 2 m temperature, and boundary layer height increase to 1.66 MJ m⁻², 25.64 °C, and 864.53 m, whereas 10 m wind speed remains around 4 m s⁻¹. Overall, SEC experiences a meteorological environment during TY characterized by stronger subsidence, fewer clouds, stronger solar radiation, higher boundary layer height, lower relative humidity, and weaker precipitation, all of which are highly favorable for ozone enhancement. Consequently, ozone rises sharply to 67 ppb.

During STY and SSTY, the WPSH retreats eastward and weakens further. Meanwhile, the area of negative 500 hPa geopotential height anomalies over and east of SEC expands relative to TY, and the anomalous cyclonic circulation at 850 hPa becomes more complete. The area affected by anomalous northerly winds over the SEC also broadens, indicating that the mid-tropospheric circulation remains strongly modulated by TC disturbances. Under such circulation pattern, the 850 hPa vertical velocity turns back to weak updraft (-0.01 and -0.02 Pa s⁻¹), surface solar radiation downwards shows a modest decrease (1.59 and 1.56 MJ m⁻²), and total cloud cover (0.52 and 0.54), 1000 hPa relative humidity (67.54 % and 68.66 %), total precipitation (0.18 and 0.2 mm h⁻¹), and 10 m wind speed (4.03 and 4.2 m s⁻¹) all increase slightly relative to TY. This suggests that the favorable meteorological configuration established at TY is no longer maintained as the TC continues to intensify. Therefore, ozone over SEC stays at a relatively high level but declines compared with TY.

The above results suggest that northward TCs within the key subregion, when intensifying to TY, can jointly promote a peak in SEC ozone concentration through fewer clouds, stronger solar radiation, higher boundary layer height, lower relative humidity, weaker precipitation, weaker surface winds, and suppressed upward motion. Although TC intensity continues to increase during STY and SSTY, the associated meteorological conditions such as solar radiation, total cloud cover, relative humidity, and wind speeds do not evolve further toward a state more favorable for ozone production and accumulation. Consequently, ozone concentration over SEC shows a slight decline rather than a further increase.

Figure 7a to j present the TC moving tracks, the daily 14:00 LT position of the WPSH (grey shading), and the observed MDA8 ozone concentration across eastern China during two severe ozone pollution episodes (from 28 September to 2 October 2019 and from 8–15 September 2022). Before 28 September 2019 (Fig. 7a) and 8 September 2022 (Fig. 7f), a persistent WPSH influenced over southern China (not shown), which caused a subsidence that suppressed the vertical dispersion of surface air pollutants. Enhanced solar radiation under the clear sky promoted efficient reactions of accumulated ozone precursors, resulting in widespread high ozone levels across eastern China. As the TCs developed and intensified rapidly over the western Pacific, they crossed the ridge of high pressure all the way northward to the southeast of Taiwan, resulting in a break in the subtropical system (Fig. 7b and g). During the subsequent northward movement of TCs, SEC (except Fujian Province near the TCs centre) maintained high ozone levels (Fig. 7c and h). When the TCs made landfall along the coast of ZJ on 1 October 2019 (Fig. 7d) and 14 September 2022 (Fig. 7i), precipitation and strong winds significantly reduced ozone concentrations in this area. Although the overall ozone levels in SEC decreased (particularly in ZJ) as the TCs generally moved away, PRD sustained high values (Fig. 7e and j).

Figure 7k and l further present the observed ozone time series for the entire d02 region (red line), the PRD (blue line), and the ZJ (black line) during two ozone pollution episodes. For the entire SEC, the afternoon ozone concentration approached or exceeded 75 ppb in the early periods (before 30 September 2019 and before 13 September 2022), and then decreased significantly. The PRD region experienced afternoon ozone concentration exceeding 75 ppb throughout the TC period. Notably, ozone concentration in ZJ showed a trend of being relatively high in the early period and decreasing significantly later (1–2 October 2019 and 13–14 September 2022).

The WRF-CMAQ model was utilized to simulate the two TC-related ozone pollution episodes, and ozone in the d02 domain was evaluated (Fig. 8). It can be seen that the model captures the diurnal variation of ozone concentration in two episodes, reproducing the characteristic pattern of a peak in the afternoon followed by a gradual decline at nighttime. While slightly underestimating daytime maxima, the simulations achieved good performance metrics with correlation coefficients (R) and index of agreement (IOA) as high as 0.95, normalized mean biases (NMB) of -0.33 % and -13.89 %, and root mean square errors (RMSE) of 7.46 and 7.79 ppb. These results demonstrate that the model performance is within an acceptable range, giving confidence for conducting further analyses.

Section 3.2.1 reveals that the two ozone pollution episodes shared similar TC moving tracks and consequently exerted comparable impacts on the ozone variations in SEC. Here, we selected three paired dates with similar TC locations in two episodes: 29 September 2019 and 10 September 2022, 30 September 2019 and 12 September 2022, as well as 1 October 2019 and 14 September 2022. We then categorized them into three periods according to the latitude where TC was located (20, 25, and 30° N) to further explore the impact of TC position within the key subregion on ozone concentration in different regions of the SEC.

Figure 9 displays the spatial distribution of near-surface ozone and 10 m wind fields at 14:00 for the CTL. Ozone concentration in SEC along the peripheral northerly airflow region of TC remained at a relatively high level, particularly in PRD and offshore areas. As TC moved northward and made landfall, ozone levels in ZJ decreased from about 100 ppb during the 20° N (Fig. 9a) and 25° N (Fig. 9b) periods to about 40 ppb at 30° N (Fig. 9c).

To investigate how TC located at different positions within the key subregion influences ozone variations over SEC at different intensities of TC, we conducted a comparative analysis of ozone differences in IT_TC (Fig. 10) and NO_TC (Fig. 11) sensitivity experiments, together with the corresponding meteorological fields (solar radiation reaching surface, 2 m temperature, and 2 m relative humidity; Figs. S1–S6 in the Supplement). When the intensified TC (IT_TC) is located within 20–25° N before landfall, northerly winds gradually strengthen (Fig. 10a and b), while solar radiation changes little (Fig. S1a and b), 2 m temperature decreases slightly overall (Fig. S2a and b), and 2 m relative humidity generally increases (Fig. S3a and b). Under this background, ozone concentration over most of SEC exhibits weak positive anomalies (Fig. 10a and b). That is to say, the TC enhancement did not lead to a significant increase in ozone concentration of SEC. In addition, the strong northerly wind also transported the locally high-concentration ozone southward, reducing ozone levels in upstream areas such as the vicinity of ZJ and some parts of the PRD, while increasing ozone concentration in downwind areas like offshore PRD areas (Fig. 10a and b). As mentioned in Sect. 3.1, when TC intensity exceeds TY level, the associated meteorological changes no longer continue to favor ozone production and accumulation. Consequently, for the TC Muifa in 2022 with a maximum intensity of STY, ozone concentration in coastal cities along SEC decreased by more than 10 ppb under its stronger influence on 12 September (bottom panel in Fig. 10b). After the enhanced TC made landfall, stronger winds and enhanced near-surface moisture (Fig. S3c) alleviated the high-temperature conditions (Fig. S2c) and decreased ozone concentration by more than 10 ppb across western SEC compared to the CTL (Fig. 10c), with a greater range of decline during the 2022 Muifa TC period (bottom panel in Fig. 10c).

After removing the influence of TC, the northerly wind weakened and the southward wind component enhanced (Fig. 11). More warm and moist air from the ocean increased the relative humidity (Fig. S6), which, combined with weak solar radiation (Fig. S4), collectively led to a reduction in ozone concentration of more than 10 ppb across most regions of the SEC. This indicates that, under the combined effects of removing the TC-related dynamical influence and imposing a weaker non-TC radiative regime, the situation of high ozone concentration in most parts of the SEC would be alleviated. Notably, without the enhanced near-surface moisture (Fig. S6c) and strong winds brought by the TC landfall, solar radiation (Fig. S4b and c) and 2 m temperature (Fig. S5b and c) both increase markedly, leading to ozone concentration in ZJ and its surrounding areas on the east side of the SEC increased by more than 10 ppb (Fig. 11b and c).

Further insights can be seen from Fig. 12, which compares atmospheric circulation and ozone vertical distribution during two ozone pollution episodes when TCs are in similar positions. During TC intensification (panels in the third row), stronger northerly winds transport more ozone within the boundary layer from ZJ (A) to PRD (B). Consequently, ozone concentration along the line AB, particularly over PRD and the offshore sea area, increased compared to the CTL (Fig. 12a and d, b and e). As the intensified TC moved to 25° N, ozone concentration over ZJ began to decline significantly (Fig. 12b and e). This reduction became more pronounced upon TC landfall near 30° N (Fig. 12c and f), where stronger winds effectively mitigated the high ozone across a broader region of ZJ. At the same time, decreased 2 m temperature and enhanced 2 m relative humidity near the surface (Figs. S2 and S3) indicate a cooler and moister environment after landfall, which further contributed to the ozone reduction. In NO_TC (panels in the fourth row), reduced solar radiation (Fig. S4) weakened photochemical ozone production, leading to a general decrease in ozone within the boundary layer. Meanwhile, the reduction in wind speeds weakens horizontal transport at all levels along the line AB, causing ozone to accumulate over PRD rather than being blown outside the sea. Subsequently, due to the lack of wind and rainfall brought by the TC landfall, ozone concentration over ZJ increased relative to the CTL experiment (Fig. 12c and f).

In summary, an intensified TC initially enhances northerly winds over the SEC, leading to a slight increase in ozone concentration with downwind transport. Subsequently, as the TC makes landfall, improved diffusion conditions and precipitation lead to a more pronounced ozone reduction in ZJ and surrounding areas. Conversely, when there is no TC, strengthened southerly winds bring more warm and humid marine air, coupled with reduced solar radiation, which causes the ozone concentration in the boundary layer of the SEC to decline by more than 10 ppb overall. Later, the lack of TC-induced rainfall and strong winds contributes to elevated ozone concentrations in the ZJ region.

On the other hand, the high temperature and intense solar radiation associated with the TC periphery can induce more biogenic VOC (BVOCs) emissions from forest areas in southern China (Wang et al., 2022; Kou et al., 2023), which are important precursors to ozone formation. Since isoprene (ISOP) is the most abundant BVOC species in the forest areas of southern China (Zheng et al., 2010; Li et al., 2023a; Wang et al., 2024b), we employ ISOP as a representative indicator of BVOC emissions to investigate the impact of solar radiation differences on BVOCs emissions between conditions with and without TC. Figure 13a and b present the variability rate of ISOP caused by solar radiation changes before and after removing TC during two ozone pollution episodes, showing that when solar radiation changes by 37 %, the ISOP concentration in SEC exhibits a corresponding linear reduction of 37 %. This led to an overall decline of approximately 1 ppb in ozone concentration (averaged from 12:00 to 18:00 during the ozone pollution episodes), with a drop of about 3 ppb in the PRD and the central part of SEC, followed by the ZJ of around 2 ppb (Fig. 13c and d). In other words, changes in BVOCs emissions due to altered solar radiation contributed approximately 1–3 ppb to the ozone production. This range of BVOCs emissions contribution is lower than Wang et al. (2022) in PRD between 2014–2018 (from an average of 16.9 % in summer to 26.6 % during TC periods) or Xu et al. (2023) in Beijing-Tianjin-Hebei during TC In-Fa (2021, 10 ppb), which may be related to differences in the research season or the types of TC selected.

To further elucidate the contribution of TCs with varying intensities (different TC experiments) to ozone inside the boundary layer in different SEC regions (PRD and ZJ) at distinct latitudinal positions (TC movement across latitudes), Table 2 summarizes the daytime (08:00–18:00) average contributions of photochemical (CHEM) and transport (HTRA and VTRA) processes results during two ozone pollution episodes. When the TC was positioned at 20° N, CHEM constituted the primary ozone source in both regions (2.13 ppb h⁻¹ in PRD and 1.64 ppb h⁻¹ in ZJ). HTRA mainly showed negative contributions, and VTRA exhibited regional variations. This indicates that even with prevailing northerly winds at the TC periphery driving southward ozone transport, photochemical processes remained dominant when TCs approached the SEC. Upon TC intensification (IT_TC), the contributions of individual processes to ozone in ZJ showed minimal changes (little change in solar radiation and 2 m temperature), suggesting limited influence of enhanced TCs at lower latitudes on the northern SEC. In PRD, the positive contribution of CHEM decreased slightly (an increase in 2 m relative humidity and a slight decrease in 2 m temperature), while the negative contribution of HTRA weakened, demonstrating that intensified northerly winds generally enhanced ozone transport to the southern SEC (PRD). For NO_TC, the reduced positive CHEM contribution (weaker solar radiation) and the significantly weakened negative HTRA contribution (reduced northerly winds) further confirmed that both diminished solar radiation and decreased northerly winds transport collectively led to ozone reduction across the SEC.

When the TC moved to 25° N, the contributions of various processes exhibited regional divergence with changes in TC intensity. Compared to CTL, both regions in IT_TC showed reduced CHEM contributions (a slight decrease in 2 m temperature), with ZJ experiencing a more pronounced decrease. The change in HTRA is similar to that when the TC is at 20° N. This indicates that although TC enhancement still facilitated outward horizontal transport, the increase in relative humidity (Fig. S3b) simultaneously suppressed photochemical production of ozone. In NO_TC, the lack of TC-induced uplifting compensation resulted in stronger subsidence over ZJ (VTRA: 1.1 ppb h⁻¹), whereas PRD was dominated by localized updraft, yielding a more negative VTRA (-1.02 ppb h⁻¹).

As the TC made landfall at 30° N, PRD showed minimal variations in individual process contributions between CTL and IT_TC (CHEM: 1.9 and 1.94 ppb h⁻¹; HTRA: -0.23 and -0.22 ppb h⁻¹; VTRA: -1.38 and -1.49 ppb h⁻¹) due to weaker TC circulation influence. In contrast, ZJ exhibited significantly enhanced negative CHEM (lower 2 m temperature, higher 2 m relative humidity, and reduced solar radiation) contribution in IT_TC resulting from TC landfall-induced precipitation. Meanwhile, intensified convergence airflows caused HTRA in ZJ to shift from negative to positive (0.86 ppb h⁻¹). Under NO_TC, with the disappearance of TC-induced disturbances, the HTRA contribution in PRD transitioned from negative to positive, while CHEM in ZJ rebounded to -0.04 ppb h⁻¹ (higher solar radiation and 2 m temperature, and lower 2 m relative humidity).

These results indicate that, when the TC is at 20° N, photochemical production dominates ozone formation, with TC enhancement primarily promoting southward transport of ozone. TC removal leads to concurrent radiation reduction and transport weakening, collectively decreasing regional ozone levels. As the TC moves to 25° N, while the enhanced TC favors horizontal transport, the accompanying relative humidity increase suppresses the photochemical production of ozone. Without TC, strengthened subsidence occurs over ZJ, while ozone in the boundary layer of PRD diffuses outward to a greater extent. During TC landfall at 30° N, individual process contributions in PRD remain relatively stable as the TC intensifies, whereas ZJ shows intensified chemical consumption due to precipitation and strong winds. The photochemical production in ZJ roughly recovers when TC influences are removed.