Quasi-weekly oscillation of regional PM_2.5 transport over China driven by the synoptic-scale disturbance of the East Asian winter monsoon circulation

Regional PM_2.5 transport is an important cause of atmospheric environment change. However, the variations in regional PM_2.5 transport on a synoptic scale with meteorological drivers have been incomprehensively understood. Therefore, this study is targeted at the quasi-weekly oscillation (QWO) of regional PM_2.5 transport over central and eastern China (CEC) with the influence of synoptic-scale disturbance of the East Asian winter monsoon (EAWM) circulation. By constructing the data of daily PM_2.5 transport flux in CEC in the winters of 2015–2019, we utilize the extended empirical orthogonal function (EEOF) decomposition and other statistical methods to extract the moving spatial distribution of regional PM_2.5 transport over CEC, recognizing the QWO in regional PM_2.5 transport with the spatial–temporal variations over CEC. The source–receptor relationship in regional transport of PM_2.5 is identified with the 2 d lag effect of the North China Plain, as the upwind source region, on the PM_2.5 pollution change in the Twain-Hu Basin, as the downwind receptor region in central China. The QWO of regional PM_2.5 transport over CEC is regulated by the synoptic-scale disturbance of the EAWM circulation with the periodic activities of the Siberian high. These findings provide new insights into the understanding of regional PM_2.5 transport with the source–receptor relationship and the meteorological mechanism in atmospheric environment change.

1 Introduction

PM_2.5 pollution has attracted worldwide attention due to its adverse impact on the environment and human health (Fan et al., 2016; Geng et al., 2021; Lin et al., 2018). PM_2.5 pollution in the cold season has become a major atmospheric environmental problem in China (An et al., 2019; Huang et al., 2020b). High-concentration PM_2.5 tends to occur with extensive spatiotemporal coverage (Tao et al., 2016; Zhang et al., 2019), and synthetic physical–chemical processes cause such heavy PM_2.5 pollution events (Ding et al., 2017; Quan et al., 2020), including emissions (Liu et al., 2016; Zheng et al., 2018a), chemical formation (Huang et al., 2014; Nie et al., 2014), atmospheric boundary layer processes (Huang et al., 2018; Zhong et al., 2018), and localized circulation (Miao et al., 2015; Shu et al., 2021; Zheng et al., 2018b), as well as weather and climate (Cai et al., 2017; Wu et al., 2016). The interactions among these physical and chemical processes make it more challenging to comprehend severe haze formation, which serves as one of the major difficulties in forecasting and controlling atmospheric environment change and heavy air pollution (Zhang et al., 2012; Zhang et al., 2019).

PM_2.5 has with complex spatiotemporal changes on multiple scales (Georgoulias and Kourtidis, 2012; Wu et al., 2021). PM_2.5 oscillates periodically at multiple timescales, and the periodic oscillation of atmospheric circulation is the leading cause of the cyclical variations in PM_2.5 (Chen et al., 2020; Dong et al., 2021; Fu et al., 2020; Perrone et al., 2018). To be specific, the 1 d periodic change or diurnal variation in near-surface PM_2.5 concentrations is mainly attributed to atmospheric boundary layer processes and localized circulation (Miao et al., 2019); the periodic change of around 7 d may be controlled by the fluctuation of the longwave trough in middle and high latitudes (Guo et al., 2014); the oscillating cycle of about 14 d is closely related to the quasi-biweekly oscillation of the synoptic circulation (Gao et al., 2020; Zhao et al., 2019); and the 30–60 d intra-seasonal oscillation is mainly caused by the impact of monsoon circulation change (Xu et al., 2014; Zhang et al., 2019). Comprehensively revealing the interaction between PM_2.5 and meteorology at different timescales is essential for solving air pollution problems more effectively (Bäumer and Vogel, 2007; Wang et al., 2020). Previous studies, which mainly focused on the multiscale periodic variation in atmospheric pollutants in a certain region or local area, have not yet investigated the PM_2.5 trans-regional and periodic oscillation in the large area of central and eastern China (CEC).

The East Asian winter monsoon (EAWM) is one of the most active atmospheric circulation systems in the cold season over the Northern Hemisphere (Ding et al., 2017; Wu and Wang, 2002), which is also a critical leading factor for the variation in wintertime air pollution in CEC (Chin, 2012; Li et al., 2016). Being a major circulation system of the EAWM, the Siberian high dominates the cold seasons, acting as a particular driver of cold airflows, thus having an important impact on the wintertime atmospheric environment in CEC (An et al., 2019; Shen et al., 2021, 2022; Wu et al., 2016). The rapid southward advance of cold air with the strong Siberian high can effectively drive the regional transport of air pollutants with less accumulation across CEC, while the weak Siberian high with the slow southward movement of cold air can be particularly favorable for the transport of air pollutants from the northern source regions to southern receptor region over CEC (Hou et al., 2020; Zhang et al., 2016). When the position of the Siberian high is more eastern than normal, the transport of air pollutants from northern China to the south is weakened, and the aggravation of pollution is enhanced in northern China (Jia et al., 2015). Regional pollutant transport driven by the southward movement of a cold front with the Siberian high would exacerbate the air quality in the corresponding receptor regions (Kang et al., 2019; Hu et al., 2021; Shen et al., 2022). The characteristics of atmospheric circulation anomalies favoring heavy haze pollution in China have changed in recent years, and the leading formation mechanism of severe haze has been shifting from local accumulation to regional transport processes in eastern China (Yang et al., 2021b). Therefore, studying the influence of the EAWM circulation system on regional pollutant transport over CEC is an important issue in atmospheric environment changes (Bai et al., 2021, 2022; Ge et al., 2018; Merrill and Kim, 2004; Tan et al., 2021; Yang, et al., 2021a).

Previous studies have primarily focused on the relationship between atmospheric intraseasonal oscillations in the mid-latitudes to high latitudes of the Eurasian region and persistent PM_2.5 pollution (An et al., 2022; Gao et al., 2020; Li et al., 2021; Liu et al., 2022; Wu et al., 2023; Yang et al., 2024b). PM_2.5 concentration anomalies in north China exhibit significant lifetimes of 10–30 d, with anticyclonic anomalies and related meteorological conditions (e.g., surface air temperature, boundary layer height) in northeast Asia influencing local PM_2.5 accumulation and hygroscopic growth (An et al., 2022; Yang et al., 2024b). These studies have investigated the quasi-biweekly life cycle of persistent PM_2.5 pollution events in north China through phase synthesis methods (Gao et al., 2020; Wu et al., 2023; Yang et al., 2024b). However, there remains a lack of systematic studies on the synoptic-scale oscillation of regional PM_2.5 transport.

The “harbor” effect on the eastern lee of the Tibetan Plateau's large topography on the westerlies is possibly an important factor influencing the regional distribution of PM_2.5 pollution in CEC with weak horizontal winds and sinking motion in the lower troposphere, which exacerbates the environmental impacts of local air pollutant emissions, establishing a “susceptibility zone” in this region (Xu et al., 2016; Zhu et al., 2018). Anticyclones and cyclones alternatively affect the region on a timescale of 3–7 d, resulting in periodic air pollution in cities (Guo et al., 2014). Thus, the weather system in CEC is basically characterized by periodic changes and cold air in winter, with the EAWM oscillating in quasi-weekly periods (Wu and Wang, 2002; Wu et al., 2016). However, the influence of the synoptic-scale disturbance of the EAWM on regional PM_2.5 transport over CEC is not yet clear. Responding to this problem, this study aims to reveal from a new perspective the quasi-weekly oscillation (QWO) of regional PM_2.5 transport over CEC affected by the EAWM and its underlying mechanism with the synoptic-scale oscillation of the EAWM circulation. This study could deepen the understanding of regional PM_2.5 transport, its source–receptor relationship, and its meteorological mechanism in the atmospheric environment changes and provide scientific evidence for air pollution forecast, early warning, and coordinated control.

2 Data and methods

2.1 Environmental and meteorological data

The daily dataset of PM_2.5 concentrations selected for this study was from the China National Environmental Monitoring Center (https://quotsoft.net/air/, last access: 27 January 2025), including daily PM_2.5 concentrations from 1079 air quality monitoring stations in CEC during the winters (December–February) of 2015–2019.

Meteorological data were selected from the NCEP/NCAR global reanalysis daily data (https://psl.noaa.gov/data/gridded/tables/daily.html, last access: 27 January 2025) with a grid resolution of 2.5°×2.5° for the large-scale circulation analysis. It is composed of the daily sea level pressure (SLP), air temperature at 1000 hPa, and the U and V components of wind at 1000 hPa during the winters of 2015–2019.

In addition, the ERA5-Land high-resolution reanalysis hourly dataset (https://cds.climate.copernicus.eu/cdsapp#!/dataset/reanalysis-era5-land?tab=form, last access: 27 January 2025) with a spatial resolution of 0.1°×0.1° was selected for the calculation of transport flux (TF) of PM_2.5 in CEC. The U and V components of the 10 m wind over CEC were obtained at 00:00, 06:00, 12:00, and 18:00 UTC daily during the winter (December–February) of 2015–2019. In order to match the resolution of PM_2.5 daily data, the ERA5-Land high-resolution 10 m wind was processed into daily average data.

2.2 PM_2.5 TF and its divergence

In order to quantitatively characterize the horizontal transport direction and intensity of PM_2.5 as well as convergence or divergence during regional PM_2.5 transport, we introduced the concepts of PM_2.5 TF and divergence of PM_2.5 TF. Generally, there are two types of TF: horizontal and vertical. This study only addresses the near-surface horizontal PM_2.5 TF. The horizontal PM_2.5 TF is defined as the PM_2.5 mass passing through the unit area in unit time (unit: µg m⁻² s⁻¹), expressed as the product of wind vector and PM_2.5 concentration (Liu et al., 2019; Ma et al., 2021), and its vector points to the same direction as the horizontal wind. The zonal component (F_u) and meridional component (F_v) of the PM_2.5 TF vector (TFV) and the TF magnitude (TFM) are calculated as follows:

$$\begin{array}{}\text{(1)}& {\displaystyle }{F}_u=C\cdot u\text{(2)}& {\displaystyle }{F}_v=C\cdot v\text{(3)}& {\displaystyle }\mathrm{TFV}=F_{u}i+F_{v}j\text{(4)}& {\displaystyle }\mathrm{TFM}=\sqrt{F_u^{\mathrm{2}}+F_v^{\mathrm{2}}},\end{array}$$

where C is the surface PM_2.5 concentration, and u and v are the zonal and meridional components of the 10 m wind speed, respectively.

Firstly, the U and V components of ERA5-Land high-resolution 10 m wind are interpolated to 1079 stations of environmental measurements in CEC for calculations of near-surface PM_2.5 TF in this study. Then, the daily PM_2.5 TF of the 1079 stations for the winters from 2015 to 2019 is calculated according to the calculation by Formulas (1)–(4).

The divergence of PM_2.5 TF can be an indicator for the PM_2.5 budget. When positive divergence occurs, the air pollutants were net outflow from the domain region, and vice versa (Wang et al., 2021). The divergence of horizontal PM_2.5 TF near the surface is calculated as follows (Wang et al., 2021):

$$\begin{array}{}\text{(5)}& D={\displaystyle \frac{\partial F_u}{\partial x}}+{\displaystyle \frac{\partial F_v}{\partial y}},\end{array}$$

where D is the horizontal PM_2.5 TF divergence (unit: µg m⁻³ s⁻¹). If D is positive (negative), it indicates divergence (convergence) of PM_2.5 TF.

In the i and j grids, the expression of Formula (5) for the differential calculation with d the grid spacing is

$$\begin{array}{}\text{(6)}& D={\displaystyle \frac{F{u}_{i+\mathrm{1},j}-F{u}_{i-\mathrm{1},j}+F{v}_{i,j+\mathrm{1}}-F{v}_{i,j-\mathrm{1}}}{\mathrm{2}d}}.\end{array}$$

When calculating the horizontal divergence of the transport PM_2.5 flux, it is necessary to interpolate the station data of zonal and meridional components (F_u, F_v) of PM_2.5 TFV to grid spacing with 0.25 by 0.25° in longitude and latitude in CEC and then calculate the divergence of PM_2.5 TF at each grid point according to Formula (6).

2.3 Butterworth filter

Atmospheric motion encompasses a variety of temporal and spatial scales. The sequences of meteorological variables often contain complex periodic components and exhibit multiple-timescale variations, including daily, weekly, seasonal, and interannual variations. Numerous observations have found the QWO with periods of less than 10 d across various meteorological elements in the EAWM system (Compo et al., 1999; Murakami, 1979; Wu and Wang, 2002). Synoptic-scale atmospheric variations are closely related to atmospheric longwave adjustments, with QWO periods of 4–7 d observed in cold-air activities of the EAWM (Bai et al., 2022; Wu and Wang, 2002). The synoptic-scale disturbance regulates the generation, transport, and removal of PM_2.5 in air pollution, which is a key mechanism behind the 4–7 d periodic changes in PM_2.5 in CEC during the periods of the EAWM (Guo et al., 2014; Liu et al., 2018; Quan et al., 2014, 2020). Based on the research objectives, identifying the desired periodic components from the original observational sequences is referred to as sequence filtering. In this study, we employed a Butterworth filter to extract the QWO from observational data.

The Butterworth filter is commonly used to separate atmospheric periodic variations across specific frequency bands. Due to its smooth amplitude response, linear phase characteristics, and ease of implementation, the Butterworth filter has been widely applied in climate and meteorological studies (Gouirand et al., 2012; Yang et al., 2024a). The Butterworth filter can be configured as a low-pass, high-pass, or band-pass filter, depending on the specific requirements. A band-pass filter only allows signals within a defined frequency range to pass through with attenuating signals outside the defined frequency range. It is often employed to extract and analyze signals within specific frequency bands, such as particular weather patterns and climate cycles. In this study, to investigate the QWO (8 d) of regional PM_2.5 transport over CEC under the influence of the EAWM circulation on the synoptic scale, we applied Butterworth band-pass filtering to the daily TFM of PM_2.5 change and daily SLP anomalies during the winters of 2015–2019 for identifying the quasi-weekly (6–9 d) synoptic-scale component of regional transport of PM_2.5 over CEC.

2.4 Extended empirical orthogonal function (EEOF)

The empirical orthogonal function (EOF) analysis is a widely applied climate statistical method in atmospheric and oceanographic scientific studies (Kim and Ha, 2015; Li et al., 2019; Schepanski et al., 2016), also used to investigate the variability of atmospheric aerosols at different spatiotemporal scales (Bai et al., 2022; Feng et al., 2020). The mathematical process of EOF analysis is to decompose the variable field X_m×n, which consists of observations at n times at m spatial points, into a linear combination of p spatial eigenvectors (modes) with corresponding time-weighting coefficients:

$$\begin{array}{}\text{(7)}& X_{m\times n}=V_{m\times p}{T}_{p\times n},\end{array}$$

where V is the spatial eigenvector (load), and T represents the time coefficient. The main information of the variable field X_m×n is represented by several eigenvectors. Since the method has been widely applied, the detailed calculation steps of EOF decomposition are omitted here, and our focus is on how to construct the observation matrix.

Firstly, we decompose the daily PM_2.5 TFM anomalies of 1079 stations in CEC during the winters of 2015–2019 using the EOF method. Thus, the following observation matrix can be obtained:

$$\begin{array}{}\text{(8)}& X=\left[\begin{array}{l}{x}_{\mathrm{11}}\mathrm{\cdots }{x}_{\mathrm{1}n}\\ \mathrm{⋮}\\ x_{m\mathrm{1}}\mathrm{\cdots }{x}_{mn}\end{array}\right],\end{array}$$

where X represents the PM_2.5 TFM anomalies, m represents the spatial points for 1079 stations, and n represents the observation times of 450 d. Then, the variable field X is decomposed into the sum of the product of space and time functions according to Formula (7).

EOF decomposition of PM_2.5 TFV anomalies can be performed by employing the complex matrix; hence the following observation matrix is constructed:

$$\begin{array}{}\text{(9)}& X=\left[\begin{array}{l}{u}_{\mathrm{11}}\mathrm{\cdots }{u}_{\mathrm{1}n}\\ \mathrm{⋮}\\ u_{m\mathrm{1}}\mathrm{\cdots }{u}_{mn}\\ v_{\mathrm{11}}\mathrm{\cdots }{v}_{\mathrm{1}n}\\ \mathrm{⋮}\\ v_{m\mathrm{1}}\mathrm{\cdots }{v}_{mn}\end{array}\right],\end{array}$$

where X is the PM_2.5 TFV anomalies, and u and v refer to the zonal and meridional components of TFV anomalies.

With EOF analysis we can get the spatial distribution structure, which is in a fixed time pattern of climate variables, but we cannot get a temporally moving spatial distribution structure. The EEOF is an extension of the EOF to analyze the autocorrelations of the variable field over time. By selecting a lag time, the original observational matrix is expanded into multiple continuous time matrices, diagnosing the temporal changes in the spatial structure of variable fields. This method has widespread applications in the analysis and prediction of marine and atmospheric motions (Dey et al., 2018; Qian et al., 2019; Wang et al., 2019).

Figure 1 Spatial pattern of the (a) EOF1 and (b) EOF2 loads in the daily change of PM_2.5 TFV anomalies (vectors, unitless) and TFM anomalies (color contours, unitless) over CEC in the winters of 2015–2019. The red and blue boxes indicate the NCP and THB, respectively. The grid cells in white represent “missing values”.

In this study, we utilized the EEOF analysis to reveal the evolution of PM_2.5 TF to reveal the spatiotemporal variations in regional PM_2.5 transport. On the basis of Formula (8), a new extension matrix of the PM_2.5 TFM is constructed. Due to the study on the synoptic scale, five lag times are selected, and each lag time is 1 d in length. The constructed observation matrix is as follows:

$$\begin{array}{}\text{(10)}& X=\left[\begin{array}{l}{x}_{\mathrm{1},\mathrm{1}}\mathrm{\cdots }{x}_{\mathrm{1},n-\mathrm{5}}\\ \mathrm{⋮}\\ x_{m,\mathrm{1}}\mathrm{\cdots }{x}_{m,n-\mathrm{5}}\\ x_{\mathrm{1},\mathrm{2}}\mathrm{\cdots }{x}_{\mathrm{1},n-\mathrm{4}}\\ \mathrm{⋮}\\ x_{m,\mathrm{2}}\mathrm{\cdots }{x}_{m,n-\mathrm{4}}\\ x_{\mathrm{1},\mathrm{3}}\mathrm{\cdots }{x}_{\mathrm{1},n-\mathrm{3}}\\ \mathrm{⋮}\\ x_{m,\mathrm{3}}\mathrm{\cdots }{x}_{m,n-\mathrm{3}}\\ x_{\mathrm{1},\mathrm{4}}\mathrm{\cdots }{x}_{\mathrm{1},n-\mathrm{2}}\\ \mathrm{⋮}\\ x_{m,\mathrm{4}}\mathrm{\cdots }{x}_{m,n-\mathrm{2}}\\ x_{\mathrm{1},\mathrm{5}}\mathrm{\cdots }{x}_{\mathrm{1},n-\mathrm{1}}\\ \mathrm{⋮}\\ x_{m,\mathrm{5}}\mathrm{\cdots }{x}_{m,n-\mathrm{1}}\\ x_{\mathrm{1},\mathrm{6}}\mathrm{\cdots }{x}_{\mathrm{1},n}\\ \mathrm{⋮}\\ x_{m,\mathrm{6}}\mathrm{\cdots }{x}_{m,n}\end{array}\right].\end{array}$$

Seen from Formula (10), the new extended matrix is composed of $X_{\mathrm{6}m,n-\mathrm{5}}$, where X is the PM_2.5 TFM anomalies, m is the spatial points of observation station, and n is the observation times of 450 d. When EEOF decomposition is performed on PM_2.5 TFV, the complex matrix is still used for the extension, and the same lag scheme is adopted to construct a new extended matrix of PM_2.5 TFV based on Formula (9). After constructing the initial data matrix, the EEOF decomposition method is in line with the classical EOF decomposition method.

Additionally, existing studies have utilized wavelet analysis, power spectrum analysis, and band-pass filtering methods to extract intraseasonal oscillation sequences of regional PM_2.5 concentrations (An et al., 2022; Gao et al., 2020; Li et al., 2021; Liu et al., 2022; Wu et al., 2023; Yang et al., 2024b). Such approaches may serve as alternative methods to EEOF analysis for establishing the quasi-weekly life cycle of regional PM_2.5 transport.

3 Results and discussion

3.1 QWO of regional PM_2.5 transport over CEC

The EOF decomposition is carried out on the daily anomalies of the PM_2.5 TFM and TFV in the winters of 2015–2019 over CEC. The first two EOFs explain 26.6 % and 14.2 % (29.1 % and 11.8 %) of the total anomalous variations in the PM_2.5 TFM (TFV), which is very helpful for better characterizing regional PM_2.5 transport variations.

Two principal modes govern the variations in PM_2.5 TF anomalies over CEC: the first leading monopole mode (EOF1) and the second meridional dipole mode (EOF2) (Fig. 1). EOF1 indicates the enhanced PM_2.5 TF over CEC (Fig. 1a). The large value center of TF mainly occurs in central China, and the transport vector direction is abnormally from the north to south. The horizontal PM_2.5 transport is unusually strong in central China, affected by the EAWM, presenting a typical channel for regional PM_2.5 transport over CEC (Yang et al., 2021a). The dipole mode of PM_2.5 TF anomalies displays a south–north out-of-phase pattern, with the flux large value centers located in the North China Plain (NCP) and the Twain-Hu Basin (THB), respectively, and the vector directions opposite (Fig. 1b). This mode indicates that the air pollutants from the NCP in the upwind region are transported to THB in the downwind region driven by the prevailing northerlies of the EAWM (Hu et al., 2021; Shen et al., 2022), and the PM_2.5 flux in NCP decreases, while that in THB increases in the regional PM_2.5 transport process.

Figure 2 (a–d) The first four phases (days) of the QWO (8 d) during the regional PM_2.5 transport over CEC and (e–h) the first four phases (days) of the next cycle. The loads of PM_2.5 TFM anomalies (color contours, unitless) for EEOF2 and TFV anomalies (vectors, unitless) for EEOF1 with lag times of (a) 0 d, (b) 1 d, (c) 2 d, and (d) 3 d and loads of TFM anomalies (color contours, unitless) for EEOF3 and TFV anomalies (vectors, unitless) for EEOF2 with lag times of (e) 2 d, (f) 3 d, (j) 4 d, and (h) 5 d over CEC in the winters of 2015–2019.

Through EOF decomposition, the PM_2.5 TF could be understood from the perspective of a fixed time pattern of climate, but the temporal changes in the moving spatial structure of PM_2.5 TF over CEC failed to be obtained. However, EEOF decomposition can be used to analyze the continuous structural evolution of the main modes of regional PM_2.5 TF over CEC.

EEOF decomposition was carried out for the daily variations in PM_2.5 TFM anomalies and TFV anomalies, respectively, over CEC during the winters of 2015–2019. Figure 2 and Fig. S1 in the Supplement show the spatial distribution of different lag times for the main modes of EEOFs, which account for about 20 % of the total variation. According to the analysis, the PM_2.5 TFM anomalies for EEOF2 and EEOF3, as well as TFV anomalies for EEOF1 and EEOF2, all show the structural evolutions in the different phases of regional PM_2.5 transport in one cycle. As can be seen, Figs. 2a–d, S1a–d, and 2e–h describe the evolution of the first and second four phases in a cycle and the first four phases in the next cycle, respectively (one phase represents 1 d).

Figure 2a–d illustrate the positive anomalies of PM_2.5 TF shifting from the NCP to THB in the first four phases under the effect of the EAWM, causing the upwind PM_2.5 TF to decrease and the downwind PM_2.5 TF to increase, which is in line with the spatial pattern of the EOF modes in Fig. 1. The last four phases show the out-of-phase pattern of the first half cycle (Fig. S1a–d). It is noted that when anomalies of PM_2.5 TFV in the NCP turn to the northerly direction (Figs. S1d and 2e), it is a strong signal initiating the regional PM_2.5 transport. Then, the transport is repeated in the next periodic cycle (Fig. 2e–h). Therefore, the regional PM_2.5 transport over CEC shows a quasi-weekly (8 d) oscillation pattern.

Figure 3 Spatial distributions of the composited (a–d) PM_2.5 TFM (color contours, unit: µg m⁻² s⁻¹) and TFV (vectors, unit: µg m⁻² s⁻¹), (e–h) anomalies of PM_2.5 concentrations (color contours, unit: µg m⁻³) and 10 m wind vectors (unit: m s⁻¹), and (i–l) divergence of PM_2.5 flux (color contours, unit: 10⁻³ µg m⁻³ s⁻¹) in the first four phases of the QWO during the 23 typical events of regional PM_2.5 transport over CEC.

To further study the variations in regional PM_2.5 transport over CEC, we have screened out 23 typical events with greater than 1.5 times standard deviations based on the standardized time coefficient of the EEOF and then used the 8 consecutive days of each event as the eight phases of the QWO in the composite analysis on the 23 typical events of regional PM_2.5 transport over CEC.

Figure 3 shows the composited PM_2.5 TF, divergence of PM_2.5 TF, and PM_2.5 concentration anomalies in the first four phases of the QWO. The high fluxes of PM_2.5 transport from north to south persist for 3–4 d over CEC and decline in the THB (Fig. 3a–d). The regional PM_2.5 transport lifetime corresponding to synoptic systems is about 3–5 d (Huang et al., 2020a). Abnormal northerly winds drive the heavy PM_2.5 pollution from the upwind NCP to the downwind regions, aggravating PM_2.5 pollution in the downwind THB (Fig. 3e–h). In the context of the QWO, the average PM_2.5 TFM in NCP decreases from approximately 400 µg m⁻² s⁻¹ in the first and second phases to 200 and 100 µg m⁻² s⁻¹ in the third and fourth phases, respectively (Fig. S2a). Correspondingly, the PM_2.5 concentration anomalies decline from around 100 µg m⁻³ to approximately −50 µg m⁻³ (Fig. S2c). In the downwind THB, the average PM_2.5 TFM increases from about 200 µg m⁻² s⁻¹ in the first phase to approximately 300 µg m⁻² s⁻¹ in the second and third phases (Fig. S2b), with PM_2.5 concentration anomalies also rising to around 50 µg m⁻³ (Fig. S2d).

It is noteworthy that the regions' PM_2.5 TF convergence zone (negative value of divergence) matches the positive anomaly centers of PM_2.5 concentrations spatially, which is confirmed with a significantly negative correlation of the PM_2.5 concentrations with divergence of PM_2.5 TF in the 23 typical events (Fig. S3). The PM_2.5 transport is accompanied by flux convergence, which is beneficial to the PM_2.5 accumulation. In addition, the PM_2.5 TF in the upwind NCP changes from convergence to divergence, and the divergence of the PM_2.5 TF in the downwind THB alters to convergence in the meantime (Fig. 3i–l), indicating that the PM_2.5 over THB is transported from the upwind NCP.

3.2 Source–receptor relationship in regional PM_2.5 transport from the NCP to THB

The regional pollutant transport governed by emissions and meteorology leads to a complex source–receptor relationship of air pollution changes (Yu et al., 2020). Band-pass filtering is performed on the daily PM_2.5 TFM anomalies at a quasi-weekly (6–9 d) synoptic scale in the winters of 2015–2019. In Fig. 4a, we composite the filter components of the PM_2.5 TFM in the eight phases of QWO during the 23 typical events of regional PM_2.5 transport over the NCP and THB, respectively. The PM_2.5 TF exhibits an obvious QWO on the synoptic scale (Fig. 4a). The PM_2.5 TF over the NCP continues to decline in the first four phases, while that of THB first rises and then falls in the last four phases, and the PM_2.5 TF over the NCP increases continuously, while that of THB falls first and then rises. We can see that the QWO of PM_2.5 TF over THB lags behind the NCP by two phases (Fig. 4a). The high TFM of PM_2.5 from the NCP in the first phase spread to THB, resulting in the peak of PM_2.5 TF over THB in the third phase.

In addition, the distribution of the differences in PM_2.5 TF and the vectors between phase 3 and phase 1 of the QWO and the PM_2.5 TF decrease and increase from phase 1 to phase 3, respectively, over the upwind NCP and the downwind THB, which is in accordance with the spatial pattern of the EOF mode (Figs. 1b and 4b), indicating the source–receptor relationship in regional PM_2.5 transport from the NCP to THB over CEC.

Figure 4 (a) The eight phases (P1–P8) of QWO during the 23 typical events of regional PM_2.5 transport over the NCP and THB with composited 6–9 d band-pass filtering of the PM_2.5 TFM and (b) spatial distribution of the differences in the PM_2.5 TFM (color contours, unit: µg m⁻² s⁻¹) and TFV (vectors, unit: µg m⁻² s⁻¹) between the third phase and the first phase of QWO. The red and black boxes represent the NCP and THB.

The statistical analysis based on long-term observation also shows that there is a significant 2 d lag relationship of positive correlation between the NCP and THB in PM_2.5 TF in the QWO (Fig. 5a). This shows that the air pollutants are transported from the upwind NCP to the downwind THB in 2 d, confirming a quasi-2 d lag in the regional PM_2.5 transport from the NCP to THB (Hu et al., 2021; Shen et al., 2021). Additionally, in the long-term change of air pollution, the divergences of PM_2.5 TF in the NCP are significantly negatively correlated to that of THB (Fig. 5b); that is, the PM_2.5 TF convergence in the downwind THB fits well with the PM_2.5 TF divergence in the upwind NCP. It can be reflected that the changes in the synoptic scale of the EAWM atmospheric circulation cause the regional PM_2.5 transport to build the source–receptor relationship of atmospheric pollutants between the NCP and THB.

Figure 5 (a) Scatter plot of 6–9 d filtering components of the PM_2.5 TFM (10⁻³ µg m⁻² s⁻¹) over THB in 2 d lag and the NCP during the winters of 2015–2019 and (b) scatter plot of PM_2.5 TF divergences (10⁻³µg m⁻³ s⁻¹) between THB and the NCP; the PM_2.5 TF divergences are averaged over the value interval of 0.1.

Driven by prevailing winds of the EAWM, the THB became the main receptor for regional transport of air pollutants over CEC (Bai et al., 2022; Shen et al., 2021). During 2015–2019, approximately 65.2 % of the total PM_2.5 heavy pollution events in the THB were triggered by regional transport of air pollutants over CEC (Hu et al., 2022; Shen et al., 2021). Such PM_2.5 transport from upstream source regions in CEC contributes 51 %–85.7 % of the PM_2.5 pollution over the THB receptor region (Hu et al., 2021; Lu et al., 2017; Shen et al., 2022; Yu et al., 2020), revealing the dominance of regional transport of air pollutants from CEC to the THB with the meteorological drivers. Our research emphasizes the QWO of regional PM_2.5 transport over CEC with the driver of the synoptic-scale disturbances of the EAWM circulation, confirming the source–receptor relationships with their 2 d lagging effects in the regional PM_2.5 transport between the upstream NCP source region and the THB receptor region.

3.3 Effect of synoptic-scale disturbance of the EAWM circulation on QWO of regional PM_2.5 transport over CEC

Meteorological change is the essential factor in regulating the occurrence and development of PM_2.5 pollution on synoptic scales. To investigate the QWO of the EAWM circulation in the synoptic-scale disturbance, this study performs the 6–9 d band-pass filtering of the daily SLP anomalies (denoted as SLP_QWO) in East Asia during the winters of 2015–2019. The SLP and SLP_QWO fields (Figs. 6 and 7) as well as PM_2.5 concentrations and 10 m winds (Fig. S4) in the eight phases of QWO during the 23 typical events were composited, respectively. The QWO of regional PM_2.5 transport is connected with the “weekly cycle” synoptic process of PM_2.5 transport and accumulation over CEC (Fig. S4), and it is powered mainly by the Siberian high circulation with the synoptic-scale disturbance of the EAWM circulation (Figs. 6 and 7).

Figure 6 The composited differences between the current day and the previous day of SLP (black contour lines, unit: hPa), 1000 hPa air temperature (color contours, unit: °C), and wind vectors (unit: m s⁻¹) in the first four phases (a–d) of QWO during the 23 typical events.

The condition of uniform pressure in the front of the Siberian high could favor the PM_2.5 accumulation over the NCP for triggering regional PM_2.5 transport over CEC (Fig. 7a). The regional heavy pollution of PM_2.5>150 µg m⁻³ lasts for 1–2 d (Fig. S4a and b). With the development of the Siberian high, the extension of the high pressure guides the cold air to advance southward (Park et al., 2014). As the result of the increasing air pressure gradients, the strong northerly winds in the EAWM circulation system deliver high-level PM_2.5 air mass from the NCP to THB (Figs. 7a–d, S4a–d). In addition, the cold and high air pressure system with the abnormal northerly airflows moves from the Siberia–Mongolia region to CEC in the first four phases (Fig. 6), providing beneficial synoptic circulation patterns for regional PM_2.5 transport. Thus, the periodic extension of the Siberian high with the associated strong cold-air intrusion is an important driver in the regional PM_2.5 transport over CEC.

Notably, we can see that in the first four phases, the SLP_QWO positive anomalies occur, develop, and expand southward from the Siberia–Mongolia region to CEC (Fig. 7a–d). The synoptic-scale disturbance with the extension of the Siberian high and the southward movement of cold air could drive the regional PM_2.5 transport over CEC (Fig. 7a–d). The situation of the last four phases is opposite to the SLP_QWO negative anomalies in Siberia–Mongolia region, inhibiting the Siberian high and cold-air intrusion (Fig. 7e–h). The low and uniform pressure is beneficial to the accumulation of PM_2.5. Therefore, the periodic changes in the synoptic-scale disturbance of the EAWM circulation cause the QWO of regional PM_2.5 transport over CEC.

Figure 7 Composited SLP (black contour lines, unit: hPa) and its synoptic-scale filter component SLP_QWO (color contours, unit: hPa) in the eight phases (a–h) of QWO during the 23 typical events. Corr represents the spatial correlation coefficients between SLP_QWO and the load of SLP_QWO decomposed by the EEOF in Fig. S4.

In addition, EEOF decomposition is carried out on the SLP_QWO field in the winters of 2015–2019 to recognize the periodic activities in the synoptic scale of the EAWM circulation. The cold-air activity of the EAWM presents QWO (Wu and Wang, 2002). The positive (negative) synoptic-scale disturbance occurs in the Siberia–Mongolia region and then spreads to CEC along the northwest–southeast path, contributing to the 8 d cycle of QWO (Fig. S5). Notably, the spatial correlation coefficients between the load of SLP_QWO decomposed by the EEOF (Fig. S5) and the SLP_QWO composited during 23 typical events (Fig. 7) are highly positively correlated in the eight phases. Therefore, the QWO in the synoptic-scale activities of the Siberian high is an important factor for driving the QWO of regional PM_2.5 transport over CEC.

4 Conclusions

Exploring the periodical oscillations of PM_2.5 pollution over CEC and the meteorological effect is crucial for understanding the change in the atmospheric environment and improving regional air quality forecasts. In this study that constructed a dataset of the daily PM_2.5 TF, the EEOF and statistical methods are used to identify the QWO of regional PM_2.5 transport with the spatiotemporal variations over CEC in winters from 2015 to 2019. The source–receptor relationship is recognized between NCP and THB with the QWO of regional PM_2.5 transport over CEC with the typical EAWM climate. Furthermore, the driving effect of synoptic-scale disturbance of the EAWM circulation on the QWO of regional PM_2.5 transport over China is revealed.

The variations in PM_2.5 TF over CEC are dominated by the first leading monopole mode and the second meridional dipole mode. The monopole mode indicates the high PM_2.5 flux along the channel of regional PM_2.5 transport from the NCP to THB under the EAWM circulation, and the dipole mode exhibits a south–north out-of-phase pattern, with two centers existing, respectively, in the upwind NCP and the downwind THB in regional transport of PM_2.5 over CEC. In terms of the long-term changes in air pollution of 2015–2019, the regional PM_2.5 transport over CEC is confirmed by the QWO, verifying a source–receptor relationship for the regional PM_2.5 transport from the NCP to THB in 2 d. Such changes are incurred by the QWO in the activities of the Siberian high, and this synoptic-scale disturbance of the EAWM circulation is generated in the Siberia–Mongolia region and then develops, moving into CEC, regulating the QWO of regional PM_2.5 transport.

The EEOF analysis with the temporal lag of the spatial fields is able to better characterize the spatial and temporal evolution of perturbations, especially propagating waves in the atmosphere (Weare and Nasstrom, 1982; Qian et al., 2019; Yang et al., 2024b). Due to its technical advantages, the EEOF method is commonly employed to extract atmospheric oscillation patterns to reveal the impacts and mechanisms of atmospheric fluctuations and monsoon circulation on regional weather, climate, and atmospheric environments (Dey et al., 2018; Qian et al., 2019; Yang et al., 2024b). In this study, we employed the EEOF method to identify regional PM_2.5 transport modes on a synoptic scale, by constructing the PM_2.5 transport flux vector (TFV) and the transport flux magnitude (TFM) with the product of near-surface PM_2.5 concentrations and wind components at 1079 stations across China during the winters of 2015–2019. We performed EEOF analysis on the PM_2.5 TFV and TFM, resulting in the spatial structure of PM_2.5 transport flux under the temporal disturbances at the synoptic scale and revealing the connection between synoptic-scale disturbances in the EAWM and QWO in regional PM_2.5 transport in CEC. Our study focuses on the driving effects of synoptic-scale disturbances associated with cold-air activity with the anomalous northerly winds in EAWM on QWO of regional PM_2.5 transport over CEC, exacerbating PM_2.5 pollution in the downwind THB. Differently from the studies on stagnant meteorological conditions associated with PM_2.5 accumulations (Gao et al., 2020; Wu et al., 2023; Yang et al., 2024b), this study provides new insights into the understanding of regional PM_2.5 transport with the source–receptor relationship and the meteorological mechanism in atmospheric environment change.

Based on the five-winter (2015–2019) observations of PM_2.5 concentrations and the corresponding meteorological reanalysis data, this study with the climate statistical and diagnostic methods investigates the QWO of regional PM_2.5 transport in China with the influence of synoptic-scale disturbance of the EAWM circulation, providing new insight into the understanding of regional air pollutant transport with meteorological drivers in atmospheric environment changes. Besides the EEOF method used in this study, the alternative methods of wavelet analysis, power spectrum analysis, and band-pass filtering could be used in further study. Future studies with utilizing long-term observations of air pollutants and meteorology over CEC could more comprehensively understand the variations in the regional transport of particles and the gaseous precursors with their contributions to air pollution, through the integration of artificial intelligence and physical–chemical process analyses.