In recent decades, particulate pollution featured by high concentrations of PM_2.5 (aerosol particle with aerodynamic diameter less than 2.5 µm) has become a serious environmental problem in China, exerting great effects on visibility impairment, urban air quality, and human health (Pui et al., 2014; Li et al., 2015; Li G. et al., 2016; Chen Z. et al., 2020). For example, in January 2013, a hazardous dense haze covered 1.4 million km² of China and affected more than 800 million people (Xu et al., 2013). As the most densely populated region in China, the Eastern China usually experiences the severest particulate pollution in the past decades (Zhang et al., 2015; Wang and Chen, 2016; Zhang et al., 2020). Therefore, understanding the mechanisms responsible for the occurrence and variation of particulate pollution in Eastern China is of great importance on social and scientific development.

As indicated by previous studies, the high-PM_2.5 events in China are usually associated with local meteorological conditions such as the changes of East Asian winter monsoon (EAWM). For example, low-level southerly anomalies associated with the weakened East Asian winter monsoon (EAWM), higher humidity, enhanced low-level static stability, and deficient precipitation are usually conducive to the formation of haze pollution (Zhang et al., 2014; Li Q. et al., 2016; Gong et al., 2018; Cheng et al., 2021). The EAWM variability is closely tied to the changes of large-scale atmospheric circulation, such as atmospheric teleconnections over mid- and high latitudes of Northern Hemisphere (e.g., Gong et al., 2019a; 2019b; 2019c). Apart from the direct atmospheric circulation factors, the atmospheric external forcings also have great impacts on PM_2.5 pollution in Eastern China. It is well known that the El Niño–Southern Oscillation (ENSO) is the most prominent atmosphere-ocean coupled climate phenomena in the tropics on the interannual timescales. It has the strongest impacts on Eastern China in winter than in other seasons (e.g., Gong et al., 2014; 2015). Previous studies have shown that ENSO may have different impacts on the interannual variations of haze pollution over the Beijing–Tianjin–Hebei (BTH) in early winter (November–December) and late winter (January–February) (Zhao et al., 2022). ENSO can induce the southerly wind anomaly, which is responsible for more haze days over North China by generating an anomalous anticyclone over the North eastern Asia (He et al., 2019; Yu et al., 2020). In addition, He et al. (2019) found that ENSO had negative influence on the winter haze days on southern China by inducing the deficient (excessive) precipitation during La Niña (El Niño) events. Other factors such as the Arctic Sea Ice (ASI), the Pacific Decadal Oscillation (PDO), and the Eurasian snowpack may also be conducive to the variations of the PM_2.5 pollution in China (Zhao et al., 2016; Zou et al., 2017; Yin et al., 2019).

However, the above studies mostly concentrated on the haze pollution in BTH, which experienced the most serious haze pollution in China due to the high anthropogenic emissions associated with rapid industrialization and urbanization (Zhang et al., 2017; Dang and Liao, 2019; Wang et al., 2019). There is little attention paid to air pollution over the whole of Eastern China. In addition, most previous works are conducted based on the atmospheric visibility data. For example, some scholars studied the air pollution by defining a winter haze days (WHD) index which is derived from observations of visibility (Wang and Chen, 2016; Mao et al., 2018; He et al., 2019). Other studies calculated the dry extinction coefficient (DEC) based on the observational visibility data to objectively describe haze pollution (Guo et al., 2017; Chen S. et al., 2020; Zhao et al., 2022). Nevertheless, the researches directly based on PM_2.5 concentration data are very limited due to the deficiency of long-term PM_2.5 observation data. In this study, the new published reconstructed high-quality PM_2.5 data are used to study the spatiotemporal variations of PM_2.5 in December over Eastern China and the possible causes.

The rest of this analysis is structured as follows. The data and methods utilized in the study are described in Section 2. Section 3 demonstrates the spatiotemporal characteristics of PM_2.5 in December over Eastern China. The possible mechanisms associated with the variations of PM_2.5 in Eastern China are also investigated. Finally, the summary and discussions are provided in Section 4.

Figure 1A presents the spatial distribution of PM_2.5 climatology in Eastern China in December during 2000–2020. There are evident regional differences of the climatology, and the PM_2.5 concentration was the highest in North China Plain (NCP) with more than 100 μg/m³ concentration. Figure 1B shows the time series of the area-mean PM_2.5 concentration in Eastern China. It is clear that the PM_2.5 concentration is featured by remarkable interannual variations, reaching a peak (70.9 μg/m³) in 2007, and decreased significantly since 2013. Such a decrease of PM_2.5 concentration may attribute to the Clean Air Action and the Blue Sky Protection Campaigns in China (Zhang et al., 2019). Figure 2 shows the first two EOF modes of the PM_2.5 concentration in Eastern China. These two modes can be significantly separated from each other via the North Test. The first EOF pattern, explaining 59.05% of the total variance, indicates the consistent spatial characteristics of PM_2.5 in the whole of Eastern China. The highest PM_2.5 is located in the NCP, indicating that the PM_2.5 variability in this region is the largest. Figure 2B shows the normalized principal component (PC1) time series corresponding to EOF1, and the correlation coefficient of PC1 with the area-averaged PM_2.5 time series (Figure 1B) is 0.98, exceeding the 99% confidence level. Therefore, EOF1 mainly shows a consistent spatial variability of PM_2.5 in Eastern China, and reflects the consistent anthropogenic emission in the whole region. The second EOF pattern and its time series are shown in Figures 2C, D, and the percentage contributions of PC2 to the total variance is 13.82%. EOF2 presents a North-South dipole pattern of PM_2.5 in the Eastern China (Figure 2C), and it reflects the opposite variability characteristics of PM_2.5 in the North and South of Eastern China, which may be related to different impacts of large-scale atmospheric circulations on the North Eastern and South Eastern China.

In order to study the influence of atmospheric circulation on PM_2.5 in Eastern China, we primarily focus on the EOF2 pattern with the North-South dipole distribution in the following analysis. Figure 3 presents the regression patterns of atmospheric circulation onto the normalized PC2. Significant anomalies of local meteorological conditions associated with the positive phase of PC2 can be observed in the Eastern China. The positive PM_2.5 concentrations in the North China are associated with the weakened surface wind speed, the decreased boundary layer height (BLH), and the increased inversion of air temperature (Figures 3A–C). The weak wind speed and the decreased BLH suppress the horizontal and vertical dispersion of the pollutants (Han et al., 2017; Wang et al., 2018). Temperature inversion in the lower troposphere usually increases the atmospheric stability and suppresses the vertical diffusion of pollutant, which finally results in more-than-normal PM_2.5 pollution accumulated in the North Eastern China. In contrast, in the South Eastern China, the increased precipitation is conducive to PM_2.5 dispersion by wet deposition (Figure 3D). Although BLH decreased slightly and temperature inversion increased over South Eastern China (Figures 3B, C), these meteorological variables are unfavorable for haze pollution over South Eastern China compared with precipitation (He et al., 2019). Hence, these results indicate that the local meteorological conditions favorable for the variations of PM_2.5 pollution are completely different in the South Eastern China and in the North Eastern China.

The spatial patterns of atmospheric circulation anomalies related to the PC2 show an equivalent barotropic structure from the lower troposphere to the upper troposphere (Figures 4, 5). The 850-hPa wind presents a strong anticyclonic anomaly over the North Eastern China, accompanied with a prominent southerly anomaly over the East China (Figure 4). The southerly anomalies reduce the climatological mean northerly (Supplementary Figure S9), and result in the decreased surface wind speed (Figure 3A) and suppress the horizontal dispersion of the pollutants, thus causing more serious PM_2.5 pollution in the North Eastern China. Figure 5A shows the results of 250-hPa geopotential height and T-N wave flux regressed onto the normalized PC2 index during 2000–2020. An atmospheric wave train is evidently observed extending from the Atlantic to the Eurasia. When the wave train arrives the Eastern Europe, it splits into two branches. The north wave train passes through the Ural Mountains arriving at the North Eastern China, while the South wave train propagates through the northern Indian toward the Eastern China. These two wave trains converge in Eastern China, and are in favor of the formation of an anticyclone in the North and a weak cyclone in the South of Eastern China, respectively, consistent with the North-South dipole pattern of the PM_2.5 concentrations (Figure 2C).

In order to explore the formation mechanism of the North wave train and the South wave train, the simultaneous sea surface temperature (SST) are further regressed onto the normalized PC2 index during 2000–2020 (Figure 5B). It is clear that there are obvious SST anomalies (SSTAs) in the central and eastern equatorial Pacific, presenting the most significantly positive SSTAs similar to ENSO. It indicates that the ENSO may have the possibility to affect the dipole distribution of PM_2.5 concentration in Eastern China. In order to investigate the possible effects of ENSO, the Niño-3.4 index is calculated as the area-mean SSTAs over 5°S-5°N, 120°-170°W. The correlation coefficient between the Niño-3.4 index and the PC2 was 0.52 above the 95% significance level. It indicates that ENSO can explain about 27.0% of the total variance of the PC2. Figure 6 displays the geopotential height and wave activity fluxes at 250-hPa obtained by regression upon the Niño-3.4 index. It should be noted that the regression results of the South wave train are similar with that regressed by the normalized PC2 (Figure 5A), but the North wave train is weaker compared with PC2 regressed ones. The results indicate that ENSO contribute greatly to the formation of the South wave train, and can also partly lead to the formation of the North wave train. Actually, according to previous studies, ENSO may induce a wave train that splits into two branches in Eurasia or cause a strong wave train to spread from mid-high latitudes to eastern China (Sun et al., 2019; An et al., 2022). Ma et al. (2022) pointed out that ENSO can induce significant precipitation anomalies in the eastern Indian Ocean/western Pacific through the double Walker circulation, which in turn affects the atmospheric circulations and results in a wave train similar to the South wave train in the upper troposphere. However, ENSO may indirectly have an effect on the North wave train because ENSO can affect the SST over the remote North Atlantic through the effects of atmospheric bridge, which in turn spreads its influences on the East Asian climate. Hence, to confirm the effects of ENSO on the PM_2.5 in Eastern China, Figure 7 shows the local meteorological elements associated with the Niño-3.4 index, which displays a large resemblance to those in Figure 3. For example, the decreased wind speed and the BLH and the increased inversion temperature emerge in the North of Eastern China. The increased precipitation anomalies occur in the South of Eastern China. The results suggest that ENSO can affect the dipole distribution of PM_2.5 in Eastern China by affecting the atmospheric wave train. The North wave train may also be affected by other atmospheric external forcing signals other than ENSO because of the weaker North wave train related to ENSO.

In order to study the influence of other atmospheric external forcing signals rather than ENSO on the North wave train, we use the partial regression method to eliminate the ENSO signal from the PC2, the SST and the SIC fields. Figure 8A displays anomalies of simultaneous SST obtained by regression upon the normalized PC2 after removing the ENSO signals. It is clear that there are still significant cold SSTAs in the Labrador Sea of the North Atlantic. Hence, we further define the Labrador Sea SST index (LSI) as the spatially weighted SST anomalies over the North Atlantic (50°–62°N, 60°–47°W). Figure 9A shows the regression patterns of the geopotential height and wave activity fluxes at 250-hPa upon the LSI. A North wave train that originates from the Atlantic and propagates eastward toward the Eastern China can be clearly found when the Labrador SST was warm. Similarly; Figure 8B shows the regression patterns of November SIC upon the normalized PC2 after eliminating ENSO signals. It is found that there is a significant sea ice reduction anomaly near the Kara Sea. As a result, the Kara Sea ice index (KSI) is then defined as the spatially weighted SIC over 78°–85°N and 90°–114°E. The 250-hPa geopotential height and T-N wave flux associated with the decreased KSI (Figure 9B) present a similar North wave train propagating toward North of the Eastern China. These results indicate that besides ENSO, the Labrador Sea SST and preceding Kara Sea ice can also affect the formation of the North wave train and may contribute to the variations of PM_2.5 in Eastern China. This is consistent with the previous studies (Ma et al., 2021).

The above analysis shows that the ENSO, the Labrador Sea SST and the Kara Sea ice all can affect the dipole pattern of PM_2.5 in Eastern China. To comprehensively characterize the effect of these atmospheric external forcing signals on the dipole structure of the PM_2.5 concentrations in Eastern China, we define a reconstructed index (REC), REC= (r1* Niño-3.4 + r2*LSI + r3*KSI)/(|r1|+|r2|+|r3|), where r1, r2, and r3 are the correlation coefficients of Niño-3.4, LSI, and KSI with the PC2 (Niño-3.4, LSI, and KSI are independent of each other). Figure 10A shows that the correlation coefficient between the REC and PC2 can reach 0.72 exceeding the 99% significance test; Figures 10B, C show the anomalous patterns of the sea level pressure, 850-hPa wind, 250-hPa geopotential height and T-N wave flux regressed upon the normalized REC index during 2000–2020. During the positive phase of the REC index, an anticyclonic circulation and the southerly anomaly are significant in the North China (Figure 10B). In addition, the North and South wave trains regressed by the REC index (Figure 10C) are also similar to those onto the PC2 index (Figure 4A; Figure 5A). These atmospheric circulations facilitate the PM_2.5 anomalies in Eastern China to present a North-South dipoles pattern. Hence, the reconstruction index defined in the study can better explain the EOF2 pattern of the PM_2.5 in Eastern China, and may provide the prediction potential for the PM_2.5 in Eastern China in December.

In this study, we used the latest PM_2.5 data provided by Tsinghua University to study the temporal-spatial distribution and mechanism of PM_2.5 over Eastern China in December. Our study found that during 2000–2020, the PM_2.5 climatology in Eastern China displays significant regional differences, with the maximum values located in the North China Plain. The area-mean PM_2.5 index in Eastern China presents obvious interannual variations, and also experiences a significantly decadal decline after 2013, which may be closely related to national emission reduction policy in recent years. Then, an EOF method is used to investigate the leading temporal and spatial modes of the PM_2.5 in Eastern China. Two significant and separated EOF modes are obtained. The EOF1 shows a consistent variability in the whole of the Eastern China, and explains about 59.05% of the total variance. The correlation coefficient between the PC1 and the original area-mean PM_2.5 index is 0.98, indicating that the first pattern mainly reveals the emission trend of PM_2.5 in Eastern China. The EOF2 presents a North-South dipole pattern in Eastern China, which reflects the opposite influence of atmospheric circulations on PM_2.5 in the North and South of Eastern China.

To explore the formation mechanism of the North-South dipole pattern in Eastern China, the atmospheric variables, sea surface temperature, and sea ice are regressed upon the normalized PC2 index. Results show that the increase of PM_2.5 concentrations in the North of Eastern China is related to the decrease of surface wind speed, the decrease of boundary layer height and the increase of inversion temperature. In contrast, the decrease of PM_2.5 concentrations in the South of Eastern China is associated with the increase of precipitation. There are two atmospheric wave trains in the upper troposphere affecting the North-South dipole pattern of PM_2.5 concentrations in Eastern China. The North wave train induces an anticyclone formed in the North China, and the South wave train induces a cyclone in the South China. The South wave train is mainly affected by the precipitation anomaly in the eastern Indian Ocean/western Pacific through a double Walker circulation induced by ENSO, which is helpful to the formation of the South wave train (Ma et al., 2022). The North wave train is jointly affected by ENSO, the sea surface temperature over the Labrador Sea and the preceding sea ice over the Kara Sea. In order to comprehensively characterize the influence of atmospheric external forcing signals on the South and North wave trains, we use the Niño-3.4, LSI, KSI index to define a reconstructed index, The correlation coefficient between the REC and PC2 can reach 0.72, exceeding the 99% confidence level. In addition, the regression patterns of the atmospheric circulations and PM_2.5 concentrations onto the normalized REC index present a large similarity to those regressed onto the PC2 index. We perform a parallel analysis using atmospheric circulation data from JRA-55 and SST data from HadISST in the period 2000–2020. The results are basically consistent with those based on ERA5 and ERSST data (Supplementary Figures S1S8). Hence, it suggests that the variations of the dipole pattern of PM_2.5 in Eastern China can be attributed to these atmospheric external forcing signals, which may provide some prediction potential. However, there is no particularly in-depth analysis of the sources of each external signal, which deserves a further study in the future.