In almost all cities of the YRD, PM_2.5 shows a downward trend from 2015 through 2021 (Supplementary Figure S2), and the proportion of cities with PM_2.5 exceeding the standard (ES) of air quality (annual mean PM_2.5 larger than 35 μg m⁻³) descends from 95.1% to 51.2%. The annual mean of PM_2.5 concentration was highest (66 μg m⁻³) in Hefei in 2015 and in Fuyang (49 μg m⁻³) in 2021. Meanwhile, MDA8 O₃ shows an upward trend in 2015–2019 and then a downward trend in 2020–2021, and the proportion of cities with MDA8 O₃ ES of air quality (MDA8 O₃ 90Per larger than 160 μg m⁻³) in total cities ascents from 41.5% to 80.5% and then descends to 68.3%. The 90Per of MDA8 O₃ was the highest (196 μg m⁻³) in Yangzhou in 2015 and in Changzhou (195 μg m⁻³) in 2021. On the other hand, in terms of spatial distribution pattern, PM_2.5 generally reduced from northwest to southeast, while O₃ reduced from northeast to southwest.

Figure 2 depicts the inter-annual variations of PM_2.5 and O₃ concentrations in the YRD. PM_2.5 decreased by 24.1%, 34.1%, 40.1%, and 39.1% between 2015 and 2021 in HBC, WJC, JZS, and ZSC city clusters, respectively, especially in the JZS, with the most notable reduction. Obviously, PM_2.5 in the HBC city cluster is much higher than others (Figure 2A), and most cities in this cluster suffer from severe particle pollution because of large coal consumption as they are located in coal-producing areas. In contrast, PM_2.5 in the ZSC city cluster is far lower than that in other city clusters due to natural vegetation coverage (Figure 2A). In particular, Lishui, Wenzhou, and Taizhou have already met the NAAQS. PM_2.5 of the HBC city cluster increased slightly before 2017 (Figure 2A). In fact, most cities in Anhui province showed an increase of PM_2.5 at the same time, especially in Fuyang, Huaibei, and Suzhou. It is worth noting that Xuzhou, situated at the junction of Jiangsu, Shandong, Henan, and Anhui provinces, is vulnerable to non-local pollutants transported from remote sources and thus has a relatively higher PM_2.5 level than surrounding cities. On the other hand, Xuzhou is the only southern city that is allowed to provide centralized heating in winter, which is bound to coal-fired emission pollution (Wang et al., 2020).

MDA8 O₃ shows a tipping point in 2019, with a rising trend before 2019 and a falling trend afterward (Figure 2B). The increasing rates were 22.2%, 66.9%, 1.7%, and 3% in the HBC, WJC, JZS, and ZSC city clusters, respectively, from 2015 through 2021. The MDA8 O₃ in the JZS city cluster exceeded the NAAQS and kept fluctuating at a high level (Figure 2B), indicating that the urban group has been suffering from serious O₃ pollution for a long time. By contrast, the rising trend of MDA8 O₃ in the ZSC city cluster was insignificant, and all cities met the NAAQS (Figure 2B). The most notable increase of MDA8 O₃ occurred in the WJC city cluster, especially in 2016, with an increase of 43 μg m⁻³ (Figure 2B), which is mainly attributed to anthropogenic emissions. On the whole, the ZSC city cluster has the best air quality in the YRD, whereas others have varying severity of combined PM_2.5 and O₃ pollution.

Figure 3 presents inter-monthly variations of mean PM_2.5 and MDA8 O₃ concentrations. PM_2.5 and MDA8 O₃ show contrary long-term trends on a monthly-scale. Previous studies have found that high particulate matter loadings usually match low O₃ concentrations in the cold season (Sun L. et al., 2019), and low particulate matter loadings match high O₃ concentrations in the warm season (Li et al., 2016). PM_2.5 generally exhibits high levels in winter and low levels in summer due to fossil fuel combustions in the cold season and turbulent vertical mixing (Ding et al., 2013; Yue et al., 2015; Zhang et al., 2018), usually reaching peaks in December or January and troughs in July, with a single-peak-valley pattern. The increased emissions during the winter heating period in North China can exacerbate particulate matter pollution in the Yangtze River Delta through the long-range transport of pollutants (Zhang H. L. et al., 2015; Zhao et al., 2015). The monthly PM_2.5 of the ZSC city cluster fluctuates slightly around 40 μg m⁻³, far less in magnitude compared with other city clusters (Figure 3A). Meanwhile, PM_2.5 in these city clusters had almost no difference in the first 2 years (2015-2016), but an obvious difference in the following 5 years (2017-2021), especially in the months before and after the peak. Except for July and August, the median PM_2.5 of all city clusters is higher than 35 μg m⁻³, indicating that most cities in the YRD suffer from PM_2.5 pollution to some extent.

MDA8 O₃ continues to rise with double-peak-valley fluctuations, with the peaks in May or June and September, and the troughs in July and December (Figure 3B). The closer to the peak month, the greater the difference in concentrations among four city clusters. High temperature, low relative humidity and intense solar radiation enhance O₃ formation in summer, whereas the meteorological conditions unfavorable to photochemistry and increasing NO titration suppress O₃ yield (Yang et al., 2021).

Figure 4 shows the ratios of major pollutants from 2015 through 2021 relative to 2015 as the reference in Huaibei, Hefei, Wuxi, and Jinhua cities. PM_2.5 exhibited a consistent decline in Hefei, Wuxi, and Jinhua. However, it rose dramatically before 2017 and then turned to a continuous decline in Huaibei (Figure 4A). In all four cities, MDA8 O₃ showed an increase before 2019 and subsequently decreased, a time point of COVID-19 occurrence (the beginning of December 2019). Hefei has the greatest increase from 2015 through 2019, with an increment of almost 80% (Figure 4B), far higher than the growth of the other three cities. NO₂ showed long-term changes similar to O₃. On the whole, SO₂ reduced year by year at an average range of 53.9%–73.7%, and CO gradually decreased by 22.6%–41.2%. The reasonable explanation is the emission reduction of primary pollutants and related precursors by the implementation of the air cleaning plan since 2013, including efficient control methods such as tightening industrial emissions, modernizing industrial boilers, retiring outdated industrial capacity, and encouraging clean fuels in the residential sector (Lang et al., 2017; Shao et al., 2018; Zhai et al., 2019; Zhang T. et al., 2019). Since the outbreak of COVID-19, human activities were reduced due to disease control regulations that lasted throughout 2020 and 2021 (Wang and Zhang 2020).

According to the Class II of NAAQS, we defined the single particulate matter pollution (uni-PM_2.5) as the day with only daily PM_2.5 exceeding 75 μg m⁻³, the single O₃ pollution (uni-O₃) as the day with only MDA8 O₃ exceeding 160 μg m⁻³, and the synergic pollution of particulate matter and O₃ (bi-PM_2.5-O₃) as both PM_2.5 and MDA8 O₃ above the aforementioned standards in a day (i.e., DHP pollution). Figures 5A–C shows the cumulative days of uni-PM_2.5, uni-O₃ and bi-PM_2.5-O₃ in the YRD from 2015 through 2021. Overall, from a spatial perspective, the total number of uni-PM_2.5 days decreased gradually from northwest to southeast in the YRD region, with the maximum number of days found in the north, including northern Anhui and northern Jiangsu provinces (Figure 5A). The uni-O₃ increased form northeast to southwest, with the maximum number of days in the central YRD, including southern Jiangsu, Shanghai, and northern Zhejiang (Figure 5B). The bi-PM_2.5-O₃ days mainly occurred in northern Anhui and Jiangsu provinces, especially on the line from Bozhou to Shanghai (Figure 5C). From seasonal perspective, the uni-PM_2.5 days mostly occurred in winter (November-February). However, the uni-O₃ days mainly appear in warm seasons (April -September), with two peaks in May, June and September (Supplementary Figure S3). The bi-PM_2.5-O₃ days mainly occur from late spring to early summer and late autumn, similar to O₃ pollution, with more distinct and obvious two peaks in April and October (Supplementary Figures S3, S4). Serious bi-PM_2.5-O₃ pollution always occurred in the cities in central and southern Jiangsu province in April, whereas in the cities in northern Anhui province, bi-PM_2.5-O₃ pollution occurred in October (Supplementary Figure S4). The uni-PM_2.5 days significantly decreased year by year in every city, especially in the HBC and WJC city clusters (Supplementary Figure S5A). Except for the ZSC city cluster, the uni-O₃ days increased in the other three city clusters, peaked in 2019, and then decreased until 2021 (Supplementary Figure S5B). Among them, the WJC city cluster has the most rapid growth, indicating an increasing severity of O₃ pollution in recent years. The bi-PM_2.5-O₃ days have prominent inter-annual changes and appear notably in 2015 and 2018 (Supplementary Figure S5C).

Figures 5D–F show the frequency statistics of pollution intensity of uni-PM_2.5, uni-O₃, and bi-PM_2.5-O₃ in YRD from 2015 through 2021. PM_2.5 concentrations were higher during uni-PM_2.5 than bi-PM_2.5-O₃. However, opposite to particulate matter, MDA8 O₃ concentrations were lower during uni-O₃ than bi-PM_2.5-O₃, which is similar to the results of Qin et al. (2021) and Awang et al. (2018). Figures 5G–I present the cumulative polluted days and the percentages of total ES days in city clusters from 2015 through 2021. The HBC city cluster has the most serious uni-PM_2.5, with a total of 496 polluted days (19.4%). The JZS and HBC city clusters exhibit the most severe uni-O₃, with percentage of 13.2% (338 days) and 13.7% (325 days). Meanwhile, the HBC and JZC city clusters have the worst bi-PM_2.5-O₃ pollution, with a total of 28 days (1.1%) and 25 days (1%). Qin et al. (2021) reported that the highest frequency of bi-PM_2.5-O₃ pollution appeared in Shanghai and the lowest in Anhui in the YRD. This inconsistency with our results may be related to the discrepancy in the pollution region divisions concerned.

Figure 9A shows the changes in annual mean PM_2.5 concentrations from 2015 through 2021 in four typical cities. A significant decline of annual mean PM_2.5 was observed in Huaibei, Hefei, Wuxi, and Jinhua cities, with a reduction of 12.21 μg m⁻³, 29.09 μg m⁻³, 28.48 μg m⁻³, and 23.61 μg m⁻³ between 2015 and 2021, respectively. Among them, the contributions of non-meteorological changes to PM_2.5 decrease were estimated to be 12.01 μg m⁻³, 28.85 μg m⁻³, 27.82 μg m⁻³, and 23.15 μg m⁻³, respectively, accounting for more than 90% of the total decrease. Zhang X. et al., 2019 reported that the improvement of air quality on PM_2.5 is primarily due to anthropogenic emission reductions of SO₂, NO_x, BC, OC, and primary particles. Therefore, implementing pollution control measures is crucial to lowering PM_2.5 loadings and reducing pollution (Zhang Q. et al., 2019; Chen et al., 2020; Li K. et al., 2020). The PM_2.5 of Huaibei decreased the most because of a significant increase in 2017 due to massive anthropogenic emissions (Supplementary Figure S6A). Compared with 2015, the PM_2.5 of Huaibei, Hefei, Wuxi, and Jinhua cities decreased by 0.2 μg m⁻³, 0.24 μg m⁻³, 0.66 μg m⁻³, and 0.46 μg m⁻³ due to changes in meteorological conditions, accounting for less than 5% of the total reduction (Figure 9A). Weather and climatic changes are conducive to PM_2.5 reduction, but do not lead to a substantial improvement in air quality (Zhang T. et al., 2019). However, in specific years, for example, Huaibei in 2016, Wuxi in 2018, and Hefei in 2019, the impact of meteorological conditions on PM_2.5 outweighed that of anthropogenic emissions (Supplementary Figure S6A). Therefore, the inter-annual variability in meteorology must be considered in designing future control strategies to improve air quality (Ding et al., 2019).

Figure 9B shows the annual mean MDA8 O₃ from 2015 through 2021 for four typical cities. The average MDA8 O₃ increased significantly by 5 μg m⁻³, 36.66 μg m⁻³, and 15.3 μg m⁻³ in Huaibei, Hefei, and Wuxi cities, respectively, but it decreased by 1.64 μg m⁻³ in Jinhua. The non-meteorological change contributed to the increase of MDA8 O₃ by 7.85 μg m⁻³, 36.42 μg m⁻³, and 14.16 μg m⁻³ in Huaibei, Hefei, and Wuxi cities, accounting for 73%, 99%, and 93% of total changes, respectively. MDA8 O₃ in Jinhua decreases by 6.43 μg m⁻³ due to non-meteorological changes (Figure 9B). O₃ pollution was possibly dominated by the NO_x-limited mechanism due to good vegetation coverage in local and surrounding regions. Previous studies have suggested that one reason for the O₃ increase is that low PM_2.5 loadings reduce sunlight scattering and absorption in the atmosphere, increasing ultraviolet radiation arriving at the ground and leading to high O₃ concentrations (Dickerson et al., 1997; Li et al., 2011; Tao et al., 2014). In addition, slowing aerosol sinks of hydrogen peroxide radicals can promote O₃ formation (Li et al., 2019). Furthermore, increasing VOCs emissions and reducing NO_x titration over urban areas can result in high O₃ concentrations (Fu et al., 2019; Sun W. W. et al., 2019). As shown in Figure 9B, the meteorological-driven changes in MDA8 O₃ are estimated to be −2.85 µg m⁻³, 0.24 µg m⁻³, 1.14 μg m⁻³ and 4.79 μg m⁻³ in Huaibei, Hefei, Wuxi and Jinhua cities, respectively. Meteorological conditions have a greater effect on O₃ than on PM_2.5, varying with city and time (Supplementary Figure S6).

PM_2.5 has a positive correlation with O₃ at high temperatures and a negative correlation at low temperatures (Zhang et al., 2018; Chen et al., 2019; Yang et al., 2021). In general, the synergic pollution of particulate matter and O₃ occurs in April and October (Supplementary Figures S3, S4). Figure 10A shows the annual and monthly trends of PM_2.5 concentrations from 2015 through 2021. The mean PM_2.5 trend observed in Huaibei, Hefei, Wuxi, and Jinhua cities were −1.7 μg m⁻³yr⁻¹, −4.6 μg m⁻³yr⁻¹, −4.2 μg m⁻³yr⁻¹, and −3.8 μg m⁻³yr⁻¹, with meteorological factor contributions of 0.05 μg m⁻³yr⁻¹, 0.05 μg m⁻³yr⁻¹, 0.11 μg m⁻³yr⁻¹, and −0.01 μg m⁻³yr⁻¹, respectively. The changes in meteorological conditions caused an increase or decrease of particulate matter, and their relative contributions for April and October are estimated to be 21% and −21% in Huaibei, 11% and −12% in Hefei, 20% and 3% in Wuxi, and −5% and 21% in Jinhua. We further identified the most important meteorological factors that affect the long-term PM_2.5 trend. For example, in Huaibei, the 2-m relative humidity (RH₂, −1.35% yr⁻¹) was responsible for 49% of total meteorological contributions in April, and the planetary boundary layer height (PBLH, −6.8 m yr⁻¹) was responsible for 34% in October. As for Hefei, the primary meteorological factor to drive the long-term PM_2.5 trend was the daytime PBLH (−12.29 m yr⁻¹), which accounted for 45% of meteorological contributions in October, while sea level pressure (SLP, 0.77 hPa yr⁻¹) was the primary factor in April. As for Wuxi, the most important meteorological factor was SLP (0.68 hPa yr⁻¹) accounting for 61% of meteorological contributions in April, while daytime total cloud cover (TCC, −0.03% yr⁻¹) accounted for 31% of meteorological contributions in October. For Jinhua, the precipitation changes (−0.03 mm yr⁻¹) could explain 43% of the meteorological-driven PM_2.5 trend in April, and the daytime PBLH (15.07 m yr⁻¹) could explain 56% of the change in October.

Supplementary Figures S7A,B present PM_2.5 changes relative to the previous year (increment) for April and October. SLP and RH play prominent roles in the PM_2.5 increment in April when meteorological impacts outweigh non-meteorological effects, whereas PBLH is prominent in October. According to Chen et al. (2020), the most significant changes in favorable climatic circumstances for improving PM_2.5 air quality are lowering RH₂ and deepening PBLH. The deep boundary layer can enhance the diffusion of pollutants through turbulent transport and vertical mixing, which reduces PM_2.5 pollution (Liu et al., 2018; Su et al., 2018; Miao et al., 2019). Particularly in periods of heavy winter pollution, the formation of secondary particles is inhibited by a reduction in water vapor, which ultimately results in a reduction in PM_2.5 (Song et al., 2018). On the other hand, high RH promotes aerosol hygroscopic growth and hastens gaseous pollutant transformation into secondary aerosol components (Cheng et al., 2015; Qiao et al., 2016).

Figure 10B displays the growth rates of yearly and monthly MDA8 O₃ from 2015 through 2021. The mean growth trends of MDA8 O₃ were +2.42 μg m⁻³yr⁻¹, +4.87 μg m⁻³yr⁻¹, +2.7 μg m⁻³yr⁻¹, and +0.83 μg m⁻³yr⁻¹ in Huaibei, Hefei, Wuxi, and Jinhua cities, respectively, in which the contributions from meteorology were −0.3 μg m⁻³yr⁻¹ (−12%), +0.26 μg m⁻³yr⁻¹ (5%), +0.34 μg m⁻³yr⁻¹ (12%), and +0.43 μg m⁻³yr⁻¹ (52%). The non-meteorological contributions are positive for all cities. The fact that the non-meteorological contribution in Hefei was significantly more than the meteorological contribution suggested that anthropogenic activity was directly responsible for worsening O₃ pollution. The meteorology contributions to MDA8 O₃ growth trends in April and October are estimated to be 65% and −4% for Huaibei, 7% and 35% for Hefei, 41% and 68% for Wuxi, and 64% and 82% for Jinhua. Further investigation was conducted to identify the most important meteorological factors to O₃ changes. The daytime surface solar radiation (SSRD) played the most important role in meteorologically induced MDA8 O₃ changes (−0.2 W m⁻² yr⁻¹), accounting for 46% of the change in April in Huaibei, while the SLP (0.41 hPa yr⁻¹) accounted for 37% in October. The primary meteorological factor was total precipitation (−0.02 mm yr⁻¹), accounting for 44% of meteorological contributions for Hefei in April, and TCC (−0.01% yr⁻¹) 26% of meteorological contributions in October. As for Wuxi, the daytime SSRD (0.21 W m⁻² yr⁻¹) could explain 39% and 43% of the total meteorological contributions for April and October, respectively. As for Jinhua, the most significant meteorological factor was precipitation (−0.03 mm yr⁻¹), attributing to 52% in April, and daytime SSRD (0.13 W m⁻² yr⁻¹) for 42% in October.

Supplementary Figures S7C, D show the meteorological driven MDA8 O₃ changes relative to the previous year for April and October. Solar radiation and cloud cover are the major meteorological factors that impact O₃ concentrations in April and October. Previous studies have shown a direct positive correlation between temperature and O₃, that is, higher temperature accelerates biological emissions of precursors and chemical reaction rate and, in turn, promotes O₃ production (Aw and Kleeman 2003; Gupta and Mohan 2015). Similarly, intense solar radiation accelerates chemical reactions and raises O₃ levels (Chang et al., 2019). In addition to the reduced downward ultraviolet radiation on the ground, through aqueous phase chemistry and photochemistry, clouds can diminish O₃ by improving oxidant elimination and lowering tropospheric oxidation capacity (Lelieveld and Crutzen 1990). Low RH₂ increases O₃ because certain complicated chemical reactions are inhibited at high humidity levels. Furthermore, low RH₂ is always accompanied by less cloudiness to speed up the photochemical synthesis of O₃ (Camalier et al., 2007). O₃ can also be significantly impacted by 500 hPa winds, for example, in the case of Huaibei in October 2020 and 2021, because wind fields have the potential to significantly affect O₃ and its precursors through transportation (Lu et al., 2019; Liu and Wang 2020).