This study utilized daily and hourly ozone data from Beijing for the period 2013 to 2023, obtained from the China National Environmental Monitoring Centre (CNEMC; https://www.cnemc.cn/en/, last access: 6 February 2025). Ozone concentrations were measured using Thermo-Fisher 49C ozone analyzers, which employ ultraviolet photometry and report concentration in standard mass units (µg m⁻³). Data was collected from 23 national air quality monitoring stations strategically distributed across urban and suburban areas to ensure representative spatial coverage. City-level hourly ozone levels were calculated as the arithmetic means of measurements from all operational stations, with city MDA8 derived from MDA8 of stations. All measurements comply with the ambient air quality standards (GB3095-2012) established by the Ministry of Ecology and Environment of China.

According to the standards set by the Ministry of Ecology and Environment of China, days on which the MDA8 exceeded 160 µg m⁻³ were classified as ozone pollution days. This research analyzed the temporal trends in the first and last pollution days of each year to determine the duration of the Ozone Season, defined as the interval between these two dates. Additionally, the ozone concentration changes on representative pollution days over the past decade were examined to characterize shifts in ozone production and loss regimes in Beijing.

Over the past decade, the onset of the first annual ozone pollution event in Beijing has demonstrably shifted to an earlier time in the year. Figure 2 illustrates the variations in the Ozone Season from 2013 to 2023. Here, FPD denotes the first ozone pollution day of the year, marking the beginning of the Ozone Season, while LPD represents the last ozone pollution day, indicating the season's conclusion. The secondary axis displays the duration of the Ozone Season. Analysis shows that this event typically occurs between March and May. In 2023, the earliest date was recorded on 30 March, whereas in 2021, the latest occurrence was on 16 May. April exhibited the highest frequency of first pollution days, with seven occurrences, accounting for 66 % of the total. May follows with three occurrences, and March with one. This trend reveals that the date of the first pollution event advanced from 5 May 2013 to 30 March 2023, indicating an average shift of approximately 3.5 d yr⁻¹ over the 11-year period.

In contrast, the date of the last pollution event (LPD) in Beijing has been delayed over the same period. The final pollution occurrences are recorded between September and October, with the earliest date on 11 September 2013 and the latest on 6 October 2015. Unlike the first pollution days, the timing of the last pollution days exhibited substantially less variability, with 88 % of occurrences in September. The last pollution date was postponed from 11 September 2013 to 27 September 2023, representing a delay of 16 d and an average trend of 1.5 d yr⁻¹, which is slower than the trend observed for the first pollution event.

Furthermore, the duration of the Ozone Season in Beijing has increased over the past decade. Defined as the interval between the first and last pollution days, the season ranged from 116 to 176 d, with an average length of 155 d. The length of the Ozone Season varied considerably from year to year, with 2021 being the shortest and 2022 the longest. The earlier occurrence of the first pollution event and the later occurrence of the last pollution event have collectively extended the Ozone Season by an average of 5.0 d yr⁻¹, representing approximately 3.2 % of the average season length.

To assess the robustness of Ozone Season variations, we applied the Mann–Kendall (MK) test to the interannual trends of the first pollution day (FPD), last pollution day (LPD), and Ozone Season length. None of these metrics exhibited a statistically significant monotonic trend at the α=0.05 level. This is not entirely unexpected, as ozone photochemical production is highly sensitive to meteorological conditions, including temperature, solar radiation, and synoptic-scale circulation patterns (Li et al., 2019; Lu et al., 2020; Wang et al., 2025). Such climate-driven variability can introduce significant noise that may mask underlying long-term trends, especially in a relatively short (11 year) time series (Weatherhead et al., 1998). We identified 2017 and 2021 as anomalous years with markedly shorter Ozone Seasons, likely attributable to such meteorological anomalies. To test the underlying trend with reduced noise, we excluded these two outlier years. After this adjustment, the MK test Z values for FPD and Ozone Season length increased to 1.80 and 1.89, respectively, approaching the critical value of 1.96 for significance at the 95 % confidence level. This suggests a marginally significant tendency toward an earlier FPD and a prolonged season length once meteorological noise is reduced, whereas the LPD still exhibited a relatively weak Z value. Thus, after filtering out climate-driven anomalies, the results tentatively support the inference that the Ozone Season onset has shifted earlier and its duration has extended, although these trends remain sensitive to interannual meteorological variability.

Currently, international research has predominantly focused on the characteristics of ozone concentration(Chen et al., 2020; Yin et al., 2017; Cao et al., 2024), with insufficient studies addressing the long-term changes in the duration of the Ozone Season. As the Ozone Season represents a period of heightened ozone pollution, its extension-driven by a combination of advancing onset and delayed conclusion, poses additional health risks to the population, necessitating greater attention to the impacts of ozone over extended periods. The prolongation of the Ozone Season is likely linked to climate change and alterations in pollutant emission patterns (Schnell et al., 2016; Doherty et al., 2013; Sicard, 2021). Therefore, our findings highlight an urgent need for adaptive pollution control strategies that account for this lengthening high-risk period.

To examine the distribution characteristics of MDA8 concentrations across different years, box plots were generated for the period from 2013 to 2023. Figure 3 illustrates the minimum, 25th percentile, mean, median (50th percentile), 75th percentile, and maximum values of MDA8 for each year. The mean MDA8 concentrations ranged from 85 to 111 µg m⁻³, with the lowest recorded in 2013 and the highest in 2023. This demonstrates considerable interannual variability and an overall upward trend, with an average increase of approximately 2.4 µg m⁻³ yr⁻¹. Notably, the anomalous year of 2021 exhibited a mean concentration of 87 µg m⁻³, representing the second lowest value during this period.

Over the past decade, the lower percentile concentrations of MDA8 in Beijing have also increased. Specifically, the minimum MDA8 concentrations rose from 4 µg m⁻³ in 2013 to 14 µg m⁻³ in 2023, reflecting a growth of 225 % with an average increase of 1.0 µg m⁻³ yr⁻¹. Similarly, the 25th and 50th percentile MDA8 concentrations increased by 24 and 31 µg m⁻³, respectively, with growth rates of 2.4 and 3.1 µg m⁻³ yr⁻¹. This rise in lower percentile concentrations suggests an increase in background ozone levels and an enhancement of atmospheric oxidizing capacity. The minimum and 25th percentile ozone concentrations passed the Mann-Kendall test (α=0.05), exhibiting a statistically significant increasing trend.

In contrast, high percentile MDA8 concentrations have exhibited no significant changes Over the study period. The maximum MDA8 concentration recorded was 311 µg m⁻³ in 2014, while the minimum was 256 µg m⁻³ in 2021. Fluctuations in the maximum and 75th percentile MDA8 concentrations demonstrated weak trends over time, likely attributable to substantial interannual meteorological variability. Notably, the 90th percentile MDA8 concentration exhibited a declining trend of approximately -2.5 µg m⁻³ yr⁻¹. The correlation of changes in high-percentile ozone concentrations Over the study period was relatively weak, possibly due to the dominant influence of interannual meteorological fluctuations on extreme ozone values. This results in less pronounced long-term trends compared to lower concentration percentiles.

Recent studies have indicated that low percentiles (below the 25th percentile) of ozone concentrations have increased globally across all regions, whereas high percentiles (above the 75th percentile) have decreased in North America and Europe, but increased in the Southern Hemisphere and Japan (Christiansen et al., 2022). Yan's research (Yan et al., 2018) also identified an increase in the 5th percentile of ozone concentrations in Europe, alongside a rise in the 95th percentile. Chen's study (Chen et al., 2020) reported an upward trend in the 10th and 25th percentile low concentrations in Beijing since 2013, consistent with these global findings. However, changes in high percentile concentrations in Beijing were found to be insignificant. As many studies have suggested, the observed rise in low percentile concentrations across various global regions may be linked to decreased emissions of ozone precursors, along with contributions from stratospheric or upper-tropospheric ozone transport. The overall increase in mean concentrations may indicate a worsening trend in ozone pollution, potentially connected to broader global climate changes that enhance ozone long-range transport and create more favorable conditions for ozone production under elevated temperatures.

The maximum daily 8 h average (MDA8) identifies the time window with the highest ozone concentration on a given day. For sensitive populations, this window warrants particular attention for health protection measures. We introduce the term ozone exposure window time (T_MDA8) to denote this period and examine its interannual variability as well as the differences between the warm and cold seasons.

Figure 4 reveals significant differences in ozone exposure window (T_MDA8) between the cold and warm seasons in Beijing. From November to January, the ozone exposure window ranges from 11:00 to 19:00. As temperatures rise, this exposure window shifts to later hours, extending from 14:00 to 21:00 during April to September. This indicates that populations are exposed to elevated ozone levels even during nighttime in the warm season. This shift is consistent with the extended photochemical production period during longer daylight hours and the delayed NO titration minimum in the warm season.

Over the study period, a noticeable trend of delayed ozone exposure times was observed during the cold season (October to March). The starting time (t1) of ozone exposure shifted from predominantly 10:00 and 11:00 in 2013 to a new window of 12:00 to 13:00 by 2023, indicating a postponement in the occurrence of peak ozone during the cold season, as shown in Fig. 5. A similar trend is observed in the warm season. However, this delay is less pronounced compared to that in the cold season, as shown in Fig. 6a. The starting time (t1) was 11:00–12:00 in 2013, but shifted to 12:00–13:00 during 2014 and 2023. This difference between seasons may be attributable to the more concentrated timing of ozone exposure events during the warm season, which reduces interannual variability in the peak onset time compared to the cold season.

This postponement of ozone exposure times suggested a delay in the occurrence of high ozone levels over the study period, which may extend the impact on human health into the nighttime hours. Currently, there is a lack of long-term studies on ozone exposure times in China. The variations in ozone exposure time in Beijing are influenced by changes in local ozone production characteristics, highlighting the importance of understanding these dynamics for evaluating the effectiveness of pollution control policies in addressing urban ozone levels.

Significant differences in ozone exposure times were observed between pollution days (MDA8 > 160 µg m⁻³) and non-pollution days. Because the number of pollution days per year is substantially lower than that of non-pollution days, the frequency distribution was normalized to obtain relative frequencies (expressed as a percentage), as shown in Fig. 7. On pollution days, the starting time (t1) of ozone exposure window ranged from 11:00 to 16:00, with a peak occurring between 13:00 and 14:00. In contrast, on non-pollution days, the starting time (t1) spanned from 01:00 to 16:00, with the majority occurring between 10:00 and 15:00. This indicates that ozone exposure time on pollution days is more temporally concentrated, primarily occurring during the afternoon to the evening. This time frame is critical for implementing protective measures for sensitive populations. This temporal concentration on pollution days reflects the dominance of intense photochemical production under favorable meteorological conditions, whereas the broader distribution on non-pollution days may result from weaker photochemical activity and greater influence of non-local transport processes.

Over the study period, a noticeable delay in the starting time (t1) of ozone exposure on non-pollution days was observed, shifting from predominantly 10:00–12:00 in 2013 to 11:00–14:00 in 2023. In contrast, changes in t1 on pollution days are relatively minor, as shown in Fig. 6b. This difference may be attributed to the dominant influence of meteorological factors, particularly temperature, on ozone production during pollution days. Because pollution days predominantly occur in the warm season, the timing of high temperatures tends to be more consistent, resulting in a more constrained diurnal window for ozone production. On non-pollution days, which are distributed across both warm and cold seasons, weaker photochemical forcing and greater meteorological variability allow the ozone exposure window to shift more readily in response to changing synoptic conditions.

Based on hourly ozone data with meteorological influences removed during the Ozone Season, typical diurnal variation patterns of ozone from 2013 to 2023 were analyzed. Each year's diurnal ozone profile exhibited a clear unimodal pattern. Ozone concentrations decreased from 01:00, reaching a minimum between 06:00 and 07:00, then rose to a peak between 16:00 and 18:00, followed by a subsequent decline (Fig. 8a and b).

The analysis revealed a decreasing trend in peak ozone concentrations, accompanied by a delay in their occurrence. In contrast, trough concentrations exhibited an increasing trend, with trough times shifting earlier over the same period. From 2013 to 2023, peak ozone concentrations during the Ozone Season ranged from 137.8 to 184.5 µg m⁻³, with a declining trend in diurnal peak values at a rate of 3.6 µg m⁻³ yr⁻¹. Peak values typically occurred between 16:00 and 18:00, with a noticeable delay over the study period. Between 2013 and 2020 (excluding 2017), peak concentrations were consistently recorded at 16:00, whereas from 2021 to 2023, they shifted to 17:00–18:00. Trough values ranged from 23 to 42 µg m⁻³, exhibiting an increasing trend at a rate of 1.7 µg m⁻³ yr⁻¹, with trough times occurring between 06:00 and 07:00. Notably, the trough time shifted to 06:00 from 2017 to 2023 (excluding 2021).

A notable observation was that daytime ozone concentrations decreased while nocturnal concentrations increased over the study period. The interannual variation curves of hourly ozone concentrations (Fig. 8c) indicated significant declining trends during the high-value period from 13:00 to 15:00 (R=-0.54 to -0.80), with the most pronounced average decrease occurring at 16:00, at a rate of approximately 3.7 µg m⁻³ yr⁻¹. This trend remained consistent, except for anomalous values recorded in 2021. Conversely, increasing trends in ozone concentrations were observed during other periods, particularly during the low-value period from 22:00 to 09:00 (R=0.78 to 0.94), with an average increase at 08:00 of approximately 2.2 µg m⁻³ yr⁻¹. This opposing trend indicates a reduction in the diurnal amplitude of ozone, decreasing from 150 µg m⁻³ in 2013 to 104 µg m⁻³ in 2023, at an average rate of 5.3 µg m⁻³ yr⁻¹. The flattening of the diurnal ozone profile suggests a weakening of both daytime ozone production and nocturnal ozone loss.

Application of the MK test to the interannual evolution of hourly ozone concentrations, revealed a statistically significant increasing trend during 01:00–10:00 local time and a significant decreasing trend during 14:00–19:00 (both p<0.05). These opposing phase shifts in the diurnal cycle may reflect changing photochemical regimes and evolving precursor emission patterns. These results align well with the findings presented above, further substantiating the conclusion that daytime peak ozone concentrations have declined, whereas nocturnal ozone concentrations have exhibited an upward trend.

Similar phenomena have been observed in the United States during the summer (He et al., 2020), where increases in morning and evening ozone concentrations are linked to reductions in anthropogenic NO_x emissions, which weaken ozone titration. Chen (Chen et al., 2020) also noted a decrease in diurnal ozone amplitude in Beijing in recent years, particularly at lower concentration ranges. This study posits that substantial reductions in NO_x emissions Over the study period have led to decreased nocturnal NO concentrations, weakening nighttime ozone loss and contributing to elevated nocturnal ozone levels. Conversely, the decline in daytime ozone peaks is likely associated with coordinated reductions in volatile organic compounds (VOCs) emissions in Beijing (Kang et al., 2021; Zhan et al., 2021; Wang et al., 2022). Since 2017, stricter VOCs emission controls have been implemented in Beijing and surrounding cities, leading to a continuous decrease in ambient VOCs concentrations (Wang et al., 2022). This decrease inhibits daytime ozone production and contributes to the observed decline in peak daytime concentrations.

By calculating the difference in ozone concentration between consecutive hours, we derived the hourly ozone change rate of (Vi), expressed in µg m⁻³ h⁻¹. A positive Vi indicates a net ozone production, while a negative Vi reflects a net ozone loss. A value of zero signifies a balance between production and loss. The hourly ozone change rate represents the total net ozone tendency. Following Kaser and other studies (Kaser et al., 2017; He et al., 2023; Song et al., 2024), this metric is understood as the integrated outcome of all physical and chemical processes contributing to the local ozone budget, including entrainment driven by boundary-layer growth, horizontal advection, dry deposition, and photochemistry. As noted in previous studies (Gu et al., 2024; Liu et al., 2021), long-term averaging of diurnal ozone variations effectively suppresses synoptic-scale advection effects. This interpretation assumes that long-term averaging effectively filters out transport effects, such that the diurnal cycle of Vi primarily reflects local photochemical production and loss rather than advective contributions. Accordingly, we employ the hourly ozone change rate in this study to characterize net ozone production and loss. 5Vi=Pi-Pi-1ti-ti-1

We plotted heat maps and time series of Vi for each year (Fig. 9a, b). The absolute value of Vi was lowest at approximately 07:00, suggesting that, on average, the increase in ozone concentration begins around this time. The primary net production of ozone occurred between 07:00 and 16:00, after which Vi became negative after 17:00, indicating the onset of net loss. This led to a decline in ozone concentration until approximately 06:00 the following day.

Over the study period, the maximum ozone production rate exhibited a decreasing trend, accompanied by an advancement in its timing. In contrast, the maximum loss rate diminished without significant changes in its timing. The maximum value of Vi (Vmax), representing the peak production rate, occurred during the hour preceding and following the most rapid production period. Prior to 2016, the peak production rate was concentrated between 11:00 and 13:00. However, after 2016, it gradually shifted earlier to 10:00–12:00, indicating an advancement in peak production timing. From 2013 to 2015, Vmax increased slightly from 24.7 to 29.1 µg m⁻³ h⁻¹, but subsequently declined to 18.2 µg m⁻³ h⁻¹ in 2023 (Fig. 9c). This reflects a decreasing trend in maximum ozone production rates after 2015, at a rate of 1.0 µg m⁻³ yr⁻¹.

The minimum value of Vi (Vmin) indicates the maximum net loss rate, ranging from -25.5 to -13.5 µg m⁻³ h⁻¹ and exhibiting a smaller magnitude relative to the maximum production rate. Vmin typically occurred between 20:00 and 22:00, with 9 out of 11 years recording occurrences at 21:00. Over the study period, a clear upward trend in Vmin was observed, increasing at a rate of 1.2 µg m⁻³ h⁻¹ yr⁻¹, suggesting a weakening of nocturnal ozone loss.

On average, ozone production rates exhibited a downward trend, resulting in a flattened diurnal profile. The interval between the morning and evening transition points of positive to negative Vi represents the duration of ozone production (T), which averaged 9 to 11 h, with the shortest duration of 9 h recorded in 2015. By subtracting the trough concentration from the peak concentration and dividing by the production duration (T), we calculated the average production speed (Vp=(Vmax-Vmin)/T) of ozone. Over the study period, Vp ranged from 9.4 to 17.1 µg m⁻³ h⁻¹, demonstrating a decreasing trend at a rate of approximately 0.69 µg m⁻³ h⁻¹ yr⁻¹. This indicates that both the maximum production and loss rates of ozone in Beijing have slowed, with the rate of change in loss exceeding that of production, resulting in a reduced diurnal variation.

The observed changes in ozone production and loss rates support the earlier conclusions about diurnal variation. This amplitude of diurnal ozone variation is diminishing, concomitant with a weakening of both ozone production and loss rates. Over the study period, reductions in anthropogenic emissions have altered the characteristics of ozone photochemistry. Specifically, coordinated decreases in VOCs and NO_x have led to lower diurnal mean ozone production rates. Additionally, the reduction in nocturnal NO concentrations has diminished the titration sink for ozone during nighttime hours.

To further investigate the variation in ozone pollution in Beijing over the study period, we analyzed pollution days (MDA8 > 160 µg m⁻³) for each year (Table 2). From 2013 to 2023, the number of pollution days ranged from 24 to 74, exhibiting substantial annual variability. Elevated numbers of pollution days were observed from 2014 to 2019, peaking in 2019, followed by a marked decline since 2020. The duration of the Ozone Season varied from 115 to 181 d. A weak positive correlation was found between the duration of the Ozone Season and the number of pollution days, suggesting that a longer Ozone Season may be accompanied by more pollution days.

The peak and trough concentrations on pollution days exhibited trends similar to those observed during the Ozone Season, characterized by declining peaks and elevated troughs. A comparison of diurnal variations between pollution days and the Ozone Season revealed that the timing of peaks and troughs was consistent, with peaks primarily occurring around 16:00 and troughs between 06:00 and 07:00. On pollution days, a discernible trend of decreasing peak values and increasing trough values was observed, with peak concentrations declining at a rate of approximately 2.5 µg m⁻³ yr⁻¹, which is 30.9 % lower than the peak reduction rate during the Ozone Season. In contrast, the trough increase rate of approximately 2.4 µg m⁻³ yr⁻¹ was 37.3 % higher than the corresponding rate during the Ozone Season.

The similarity in diurnal variation patterns suggests that both ozone pollution days and typical days in Beijing are characterized by weakening ozone production and loss rates. The slower decline rate in peak concentrations and faster increase in trough concentrations on pollution days imply that favorable meteorological conditions lead to elevated precursors and NO concentrations in the atmosphere during these events, thereby sustaining ozone production capacity, while nocturnal conditions facilitate enhanced loss.

Over the study period, the concentration differences between pollution days and the Ozone Season have widened, with the ozone production capabilities on pollution days remaining relatively robust compared to those during the Ozone Season. Figure 10 presents a heat map illustrating the hourly concentration differences between pollution days and the Ozone Season. Hourly ozone concentrations on pollution days consistently exceeded those in the Ozone Season, with peak values surpassing those of the ozone season by 43 to 57 µg m⁻³. In contrast, the differences in trough values were less pronounced, ranging from 1 to 11 µg m⁻³.

Notably, the hourly concentration differences between pollution days and the Ozone Season exhibited an increasing trend over the study period, with peak differences increasing at a rate of approximately 1.0 µg m⁻³ yr⁻¹ and trough differences increasing at approximately 0.65 µg m⁻³ yr⁻¹. The correlation coefficients for these trends were 0.65 and 0.68, respectively. Additionally, nocturnal concentration differences became considerably more pronounced, rising from 19.2 to 34.2 µg m⁻³, at a rate of approximately 1.3 µg m⁻³ yr⁻¹.

The widening of hourly concentration differences between pollution days and the Ozone Season suggests that, although overall ozone production capabilities are declining, the rate of reduction on pollution days is comparatively slower. This indicates that pollution days retain a significant capacity for ozone production.

The Multi-resolution Emission Inventory for Climate and Air Pollution Research (MEIC), developed by Qiang Zhang and colleagues at Tsinghua University, is a high-resolution anthropogenic emission database for China (Li et al., 2017; Zheng et al., 2018, 2021). It provides long-term estimates of major air pollutants and greenhouse gases with spatial resolutions down to 1 km, covering approximately 800 anthropogenic source categories across power, industry, residential, and transportation sectors (Zheng et al., 2018, 2021). The inventory has been extensively validated through satellite observations and widely applied in atmospheric modeling and policy assessment studies across China (Zheng et al., 2018). Based on NO_x and VOCs emission data from MEIC, coupled with annual mean NO₂ and MDA8 ozone observations in Beijing, we analyzed the influence of pollutant emissions on local ozone formation characteristics.

To maintain consistency with the preceding analysis, we separately calculated and examined the multi-year variations in pollutant emissions and ambient concentrations for the warm and cold seasons, with both expressed as arithmetic means for the respective periods. Figure 11a and b present the results for the warm and cold seasons, respectively.

Overall, NO_x and VOCs emissions exhibited a declining trend over the study period in both seasons, consistent with the nationwide emission reductions driven by China's Clean Air Actions (Zheng et al., 2018). However, a notable rebound in warm-season emissions was observed in 2021. The VOCs / NO_x emission ratio ranged from 1.6 to 2.2 during the warm season, showing marked increases in 2016 and 2020–2021, followed by a rapid decline during 2022–2023. In the cold season, this ratio ranged from 1.5 to 1.9 and remained relatively stable across most years, except for 2022. The long-term reductions in NO_x and VOCs emissions reflect Beijing's sustained regulatory efforts targeting ozone precursors. The rapid decrease in the warm-season VOCs / NO_x emission ratio since 2022 may be attributed to Beijing's intensified VOCs control measures and the coordinated mitigation of NO_x and VOCs emissions from industrial and mobile sources.

Over the study period, ambient NO₂ concentrations in both seasons showed consistent downward trends, closely tracking the NO_x emission trends (R=0.95 for the warm season and R=0.97 for the cold season). This strong coupling between NO₂ observations and NO_x emissions corroborates the reliability of the MEIC inventory in capturing emission changes (Zheng et al., 2018). However, warm-season MDA8 ozone showed no significant correlation with either VOCs emissions or the VOCs / NO_x ratio. In the cold season, MDA8 ozone exhibited a weak negative correlation with VOCs emissions (R=-0.69), but no clear relationship with the VOCs / NO_x ratio. These contrasting behaviors likely arise because ambient NO₂ is closely linked to local NO_x emissions through relatively linear chemistry, whereas ozone photochemistry is highly nonlinear and strongly modulated by meteorological conditions (Li et al., 2019; Lu et al., 2020; Wang et al., 2021). This decoupling between precursor emission reductions and ozone concentration trends underscores the dominant role of meteorological variability in driving interannual ozone fluctuations. Ozone pollution control is considerably more challenging than the mitigation of primary pollutants, necessitating region-specific, regime-targeted strategies that account for both precursor ratios and synoptic conditions (Qiu et al., 2025).