Introduction

The world is currently facing the double burden of climate change and air pollution. Studies have proven that air pollution is one of the key causes of global and regional climate crisis, threatening the sustainable development of socio-economic, human well-being and accelerating the pandemic with complex mechanisms (Fig. 1) (Åström, 2023; Liang et al. 2023a, b; Lei et al., 2022). The process of industrialization and urbanization has promoted a sharp increase in fossil energy consumption and aggravated global air pollution (Fig. S1). Around 4.2 million premature deaths worldwide are caused by air pollution each year (Gonzalez et al. 2022). Currently, China’s air pollution presents the characteristics of multiple pollution sources, superposition of multiple pollutants, and combination of urban and regional pollution (Table 1). Moreover, the top three CO2 emissions in 2020 are China, the United States and India, accounting for 30.93%, 13.86% and 7.19% of the global total carbon dioxide emissions respectively (Fig. S2). Coordinated reduction of pollutants and carbon emissions is the fundamental guarantee for China to reach carbon neutrality by 2060 (Dong et al. 2022; Li et al. 2023). In 2020, the proportion of severe and above pollution day (1.2%) has improved significantly compared to 2015 (2.8%) (Azmi et al. 2023; Zhang et al. 2022a, b, c). However, the concentration of PM2.5 improves slowly, and the reduction of NOx emission is not significant. Moreover, O3 pollution is becoming more and more prominent, and the coordinated reduction of PM2.5 and O3 has become a severe challenge.

Fig. 1
figure 1

Links between air pollution, energy, climate and health

Full size image
Table 1 Evolution of air pollution in China
Full size table

The regions with the worst air pollution in China are mainly concentrated in the Beijing-Tianjin-Hebei region (BTH), the Pearl River Delta (PRD), the Yangtze River Delta (YRD), the Fenwei Plain (FWP), Jingjinji (JJJ) and the Sichuan basin (SCB) (Fig. 2) (Feng et al. 2023). Since the location of the SCB is unique worldwide, it provides an excellent platform for complex terrain air pollution investigations. However, unlike other heavily polluted regions, air pollution in the SCB is not well understood. Based on the relevant regulations of the Emergency Plan for Heavy Pollution Weather in Sichuan Province (revised in 2018), with eight or more cities reaching moderate and above levels of pollution as the standard, it was found that there were 21 regional air pollution processes in SCB during 2015–2018 (Table S1). Relevant literature has confirmed that severe air pollution frequently occurs in highly industrialized and urbanized cities with complex terrains, such as Beijing, Lanzhou and Chengdu in China (Fan et al. 2019; Meng et al., 2020)Footnote 1.

Fig. 2
figure 2

The locations of the five urban agglomerations in China

Full size image

As a megacity, Chengdu has a more arduous task of preventing ecological and environmental risks, and the establishment of a long-term mechanism for normalized collaborative governance still faces challenges (Fan et al. 2019). Although numerous research have been done in different regions of SCB in recent years, it is still lack of detailed and in-depth studies on Chengdu (Hou et al. 2022; Hu and Wang 2021; Tan et al. 2023). Moreover, most of the reported studies explore the short-term impacts and typical pollution events, and are limited to the data of a single pollutant at a few sites. Only few studies have explored the spatial variation characteristics of pollutants at multiple sites over a long period of time. Therefore, the results obtained by these studies may not conclusive and could lack of spatial contrast. Nevertheless, there is a lack of detailed discussion of the variation characteristics of multiple meteorological parameters at multiple stations and their impact mechanisms with multiple pollutants in previous studies. In addition, the trajectory clustering model, Potential Source Contribution Function (PSCF) model and Concentration Weighted Trajectory (CWT) models are effective tools for exploring regional pollution transport (Berriban et al. 2022; Cao et al. 2023), but they are rarely used in SCB regions, especially in the Chengdu megacity. Extensive research has revealed that even for the same research subjects, there may be significant bias in the results due to differences in research parameters and methods (Liu et al. 2021). Therefore, it is difficult for some existing studies to accurately characterize the long-term pollution status of Chengdu megacity.

Through the different research methods and models, a more detailed analysis of air pollution in megacities from the perspectives of spatial distribution, meteorological influence factors, and regional transmission can accurately grasp the status of regional air pollution, and provide reference for traceability of pollution sources and more accurate trend prediction. Besides, this research can provide new methods and data references for exploring efficient air pollution control measures in complex topography, and also accumulate some experience in the regional joint prevention of megacities in similar situations in the world. Using the long-term monitoring data pollutants in Chengdu from 2014 to 2020, the spatial fluctuation characteristics of the six criteria pollutants (PM2.5, PM10, NO2, SO2, O3 and CO) were comprehensively investigated. The annual, seasonal and monthly spatiotemporal evolution of seven meteorological factors was then systematically elucidated. Furthermore, a comprehensive and detailed analysis of the effects of multi-meteorological parameters on the air pollutants also were explored. Finally, combined with the cluster analysis, PSCF model and the CWT method, the potential source areas of PM2.5 and O3 in 2020 were further quantitatively simulated from different season.

Methodology

Study area

Surrounded by mountains, the Chengdu megacity (30° 05' N-31° 26' N, 102° 54' E-104° 53' E) is located on the western edge of the SCB, with a total area of 14,605 km2 (larger than Shanghai, Tianjin megacities) (Chen et al. 2022a, b, c; Deng et al. 2019) (Fig. 3a). Compared with 2010, in 2020, its GDP, urbanization rate, population and car ownership have increased by 200%, 19.8%, 49.04% and 200%, respectively (Fig. 4). By the end of 2021, the resident population had reached 21.19 million, the urbanization rate had reached about 79.5%, the total value of regional power generation was 1.99 billion and the number of motor vehicles was about 6.26 million (National Bureau of Statistics of China 2022). The industry is mainly characterized by automobiles, coal-fired power plants, building materials, machinery, food, etc. (Zhang et al. 2018; Wang et al. 2018a, b). In addition, the terrain of this area is of high in the west and low in the east, with a height difference of 4,966 m, and the terrain is relatively closed (Ning et al. 2019; Yang et al. 2020). The altitude in this area is averagely 750 m, with the lowest altitude of 359 m. Due to the huge vertical height difference, unique landform types of plains (40.1%), hills (27.6%) and mountains (32.3%) are formed in this area.

Fig. 3
figure 3

(a) The location of Chengdu and (b) the topography and monitoring station of Chengdu

Full size image
Fig. 4
figure 4

Area and population and GDP in division of Chengdu in 2020 ( Tan et al. 2023; Zhang et al. 2022a, b, c; Zhu et al. 2020)

Full size image

Data collection

The six criteria pollutants in this study are mainly based on the data from eight state-controlled monitoring stations in Chengdu (Fig. 3b). Besides, seven surface meteorological parameters (Win, Temperature (TEM), Ground Surface Temperature (GST), Relative Humidity (RHU), Precipitation (PRE), Pressure (PRS), Sunshine Duration (SSD)) mainly come from 13 meteorological stations (Fig. 3b). Besides, the analysis data also refers to third-party sources, such as China air quality platform (http://www.aqistudy.cn/). Daily meteorological data mainly refer Chengdu Meteorological Monitoring Database (http://data.cma.cn/). Besides, the Geospatial Data Cloud (https://www.gscloud.cn), Sichuan and Chengdu Statistical Yearbook (http://m.chdstats.gov.cn/) also referred to obtaining geographic information data and socio-economic data. Also, meteorological data obtained from the National Center for Environmental Prediction of the United States (ftp://arlftp.arlhq.noaa.gov/pub) are used for backward trajectory model. In addition, the parameter settings in the backward trajectory model are shown in Table 2 (Bai et al. 2021a, b; Berriban et al. 2022).

Table 2 2014–2020 backward trajectory model parameters
Full size table

Analysis methods

Correlation analysis

As a non-parametric test method, the Spearman rank correlation coefficient is mainly used to present the relationship between variables (Gao et al. 2016; Tarasov et al. 2018). When there is no repeated data and one variable is a strictly monotonic function of another variable, the Spearman rank correlation coefficient is equal to + 1 or -1, which is called the complete Spearman rank correlation of variables (Liu et al. 2023; Qi et al. 2023; Xu et al. 2022). When all rank values are integers, the rank correlation coefficient can be calculated by the formula Eq. 1.

$${r}_{s}=1- \frac{6\sum {d}_{i}^{2}}{n\left({n}^{2}-1\right)}$$
(1)

\({d}_{i}\) is the rank difference between the two variables in each set of observations, \(n\) is the total number of paired samples. This study uses Spearman’s correlation analysis to clarify the relationship between meteorological indicators and air pollutant indicators in order to understand the strength of the correlation between variables and focus on controlling pollutants that have a greater impact on air quality.

Interpolation techniques

Spatial interpolation is a method of converting discrete point measurement data into a continuous data surface and comparing it with other spatial distribution patterns. The most commonly used methods of spatial interpolation include Inverse Distance Weighting (IDW) interpolation, Kriging interpolation, Spline function method, etc. (Kumar et al. 2020; Lotfata 2022; Pinto et al. 2020). Among them, Kriging is a prominent interpolation technique and the most commonly used spatial interpolation algorithm. Kriging interpolation is more accurate than other interpolation methods. By considering spatial autocorrelation and the weight of data points, this method can generate a more accurate spatial prediction model. In addition, this method not only considers the numerical value of the data point, but also the spatial correlation of the variable. Finally, Kriging interpolation can select different variogram models according to different data distribution and autocorrelation patterns, which is highly flexible (Shukla et al. 2020; Yang & Hu 2018). Kriging is a geostatistical technique that is an interpolation method for unbiased optimal estimation of variables within a limited area (hukla et al., 2020; Yang & Hu 2018). In the kriging method, it is assumed that the joint probability distribution of the entire research space is stationary, and uniform in all directions. In this method, \(\left(x, y\right)\) set as the coordinates of a certain space data point and \(z\left(x, y\right)\) represents its value. For \({z}_{0}=z\left({x}_{0}, {y}_{0}\right)\), the Kriging interpolation formula is shown in (Eq. 2) (Beauchamp et al. 2018; Kumar et al. 2022; Lotfata 2022).

$${\widehat{z}}_{0}= \sum_{i=1}^{n}{w}_{i}{z}_{i}$$
(2)

Since the Kriging method is unbiased and has the smallest variance, the undetermined weight coefficient \({w}_{i} \left(i=\text{1,2},\dots ,n\right)\) of the unbiased condition satisfies the relational expression (Eq. 3):

$$\sum_{i=1}^{n}{W}_{i}=1$$
(3)

When there is no bias and the kriging variance is the smallest, the variance group of undetermined weight coefficients \({w}_{i}\) can be obtained (Eq. 4). \({\widehat{z}}_{0}\) represents the estimation value at the point \(\left({x}_{0}, {y}_{0}\right)\), \({z}_{i} \left(1\le i\le n\right)\) represents the observation value of \(n\) points near of the \(\left({x}_{0}, {y}_{0}\right)\), \({w}_{i}\) represents the weight coefficient of the observation value \({z}_{i} \left(1\le i\le n\right)\) and the relative point \(\left({x}_{0}, {y}_{0}\right)\). \(C\left({x}_{i}, {y}_{i}\right)\) is the covariance function of \({z}_{i}\) and \({z}_{j}\).

$$\left\{\begin{array}{c}\sum_{i=1}^{n}{w}_{i}C\left({x}_{i}, {x}_{j}\right)+ \mu =C\left({x}_{0}, {x}_{j}\right), \left(j=\text{1,2},\dots ,n\right)\\ \sum_{i}^{n}{w}_{i}=1\end{array}\right.$$
(4)

Backward trajectory clustering

Due to the spatial similarity of air mass trajectories, the backward trajectory clustering method can group all trajectories and calculate the spatial dissimilarity (SPVAR) and total spatial difference (TSV) of each trajectory from the clustered average trajectory (Foy et al. 2021; Li et al. 2017a, b). Then, according to the relationship between TSV and n, the number of clusters is judged, and air mass trajectories reaching at the receiving point of the model are grouped and clustered (Sasmita et al. 2022). The SPVAR of each cluster is expressed as Eq 0.5:

$$SPVAR= \sum_{i=1}^{x}{D}_{i}$$
(5)

Trajectory clustering is mainly calculated by two clustering methods, angular distance and Euclidean distance (Cao et al. 2020). This study adopts the Euclidean distance method, and the principle of grouping and clustering is to make the difference in moving speed and direction between the trajectories in the group extremely small, and at the same time, the difference between groups is maximized. The Euclidean distance between two trajectories within a particular cluster is defined as Eq 0.6 (Li et al. 2017a, b; Sen et al. 2017). \({X}_{1}\left({Y}_{1}\right)\) and \({X}_{2}\left({Y}_{2}\right)\) is the trajectory 1 and 2.

$${d}_{12}= \sqrt{\left(\left({X}_{1} \left(i\right)- {X}_{2} {\left(i\right)}^{2}+ \left({Y}_{1} \left(i\right)- {Y}_{2}{\left(i\right)}^{2}\right)\right.\right.}$$
(6)

PSCF model

The proportion of pollution trajectories and the distribution of potential contributing source areas of grid cells to high pollutant loads at receptor sites can be found through PSCF values. The high value area of PSCF is the main area of high pollutant concentration (Stojić & Stanišić, 2017). In PSCF model, Chengdu (30.66°N, 104.10°E) was set as the starting point and the simulation start altitude is set to 500 m (Deng et al. 2020; Kaskaoutis et al. 2019). The thresholds of PM2.5 and O3 in the model are set at 75 μg/m3 and 160 μg/m3, respectively (Chen et al. 2022a, b, c; Zhang et al. 2021a, b). Finally, the grid cover area is set within the range of 25.83–43.13°N and 75.40–111.59°E, and grid cells of 1° × 1° were contained in the study. PSCF model divides the research area into \(i \times j\) grids, and \({n}_{ij}\) is mean the number of all trajectories passing through the grid \(\left(i, j\right)\), \({m}_{ij}\) is the number of pollution trajectories, \({PSCF}_{ij}\) (the occurrence probability of pollution track in grid \(\left.\left(i, j\right)\right)\) is expressed an Eq 0.7 (Bai et al. 2021a, b; Berriban et al. 2022).

$${PSCF}_{ij}= \frac{{m}_{ij}}{{n}_{ij}}$$
(7)

Besides, \({W}_{ij}\) (weight factor) is used to calculate PSCF to reduce uncertainty (Zhang et al. 2022a, b, c; Zhou et al. 2023). By counting the number of all trajectory nodes and the total number of grids, the average number of nodes per grid is obtained, which is defined as \({n}_{ave}\), then the calculation formula of:

$${W}_{ij}=\begin{array}{c}1.00, \left({n}_{ij} > {3n}_{ave}\right)\\ 0.70, \left({1.5n}_{ave} < {n}_{ij} \le {3n}_{ave}\right)\\ \begin{array}{c}0.42, \left({n}_{ave} < {n}_{ij} \le {1.5n}_{ave}\right)\\ 0.2, \left({n}_{ij} \le {n}_{ave}\right)\end{array}\end{array}$$
(8)

The PSCF value after adding the weight can be expressed as Eq 0.9:

$${WPSCF}_{ij}= {W}_{ij} \times {PSCF}_{ij}= {W}_{ij} \times \frac{{m}_{ij}}{{n}_{ij}}$$
(9)

CWT model

The study area set in the CWT model calculation almost covers the entire airflow transport path, with a resolution of 1.0° × 1.0°. In addition, the threshold settings for PM2.5 and O3 in the CWT model are consistent with those in PSCF. The weighted concentration of each grid cell is \({C}_{ij}\) (Eq 0.10) (Bai et al. 2021a, b; Cao et al. 2023):

$${C}_{ij}= \frac{\sum_{l=1}^{m}{C}_{l}{\tau }_{ijl}}{\sum_{l=1}^{m}{\tau }_{ijl}}$$
(10)

In the formula, \(i\) and m represent the trajectory index and the total number of trajectories, respectively. \({C}_{l}\) is the corresponding pollutant concentration when the trajectory \(l\) passes through the grid unit \(\left(i, j\right)\), and \({\tau }_{ijl}\) is the trajectory in the grid unit \(\left(i, j\right)\) Duration of stay. The same weight factor \({W}_{ij}\) as PSCF were used in the study to reduce the uncertainty of CWT. A cell with a higher Weighted Concentration Weighted Trajectory (WCWT) value indicates a greater contribution to the pollutant concentration (Cao et al. 2023; Tian et al. 2022) (Eq 0.11).

$${C}_{ij}= \frac{{W}_{ij}}{\sum_{l=1}^{m}{\tau }_{ijl}} \sum_{l=1}^{m}{C}_{l}{\tau }_{ijl}$$
(11)

Results and discussion

Spatial and temporal heterogeneity of pollutants

Spatial heterogeneity of pollutants

Concentrations of air pollutants are not always at a steady level, but is characterized by multiple temporal and spatial scales. In 2014, the areas with high average concentration of PM2.5 were mainly concentrated in the central urban area, while the low-value region were distributed in Dujiangyan (Fig. S3). In addition, the central region (Wuhouqu, Jinjiang and Shuangliu etc.) is a high CO value area. This finding is similar to those of Kuang et al., (2018), they pointed out that lots of automobiles and industrial enterprises in the central area of Chengdu are the reasons for the higher average CO concentration. On the contrary, the high-value areas of O3 are mainly concentrated in Dujiangyan and Pengzhou, while the central urban areas are relatively low. In 2015, the pollutants PM2.5, PM10, NO2 and CO all gradually decreased from the central urban area to the east and west, and the low-value areas all appeared in Dujiangyan (Fig. S4). While, the distribution characteristics of SO2 and O3 concentration values are characterized by lower values in the central urban area and higher values in the northwest direction. In addition, the SO2 concentration is higher in the northwest, mainly because there are a large number of industries and power plants in these areas (Kuang et al., 2018). The high-concentration area of O3 presents a zonal distribution in the northwest direction and is mainly caused by motor vehicle emissions in the central urban region, and in the northern region the impact of industry on O3 is more important.

Compared with 2014, the PM2.5 concentration in 2016 has improved to a certain extent, and the value in most areas is higher than 53.6 μg/m3 (Fig. S5). In addition, compared with 2015, the area with high NO2 and O3 concentration showed an expanding trend. It may be due to the substantial increase in VOCs emissions in industrialized areas, which facilitated the reaction with NOx (Kuang et al., 2018). Zhou et al., (2021) identified that OVOCs emitted from wood-based panels, furniture manufacturing, pharmaceutical industries and automobile manufacturing accounted for more than 50% of Chengdu. Therefore, controlling the emission of VOC in the manufacturing industry is the key to improving the O3 concentration. In 2017, the high-value areas of PM2.5 and PM10 has shown a trend of shrinking compared with 2014–2016 (Fig. S6). However, the overall pollution is still relatively serious, and the central urban area is still a key area for emission reduction. Air pollutants have serious impacts on human health, and the lethal effects are mainly manifested in total mortality, respiratory diseases, cardiovascular diseases and lung cancer (Alexeeff et al. 2022; Dimitriou et al. 2022). Therefore, it is particularly necessary to reduce pollutant emissions.

Except for O3, the concentration values of other pollutants show a decreasing trend in 2018 (Fig. S7). In total, about 98.0% of the region’s PM2.5 concentration values failed to reach the limit of CAAQS Grade II. The areas with the highest concentration are Wuhou, Jinjiang, Shuangliu and Qingyang, which is 10 times higher than the WHO, 2021AQGs standard. Xu et al. (2024) pointed out that PM2.5 pollution is a major environmental risk factor that can bring about huge disease burden and economic losses (Xu et al. 2024). Song et al., (2022a, b) emphasized that the formation of secondary aerosols (43.4%) in the winter of 2018 was the main cause of the high concentration of PM2.5. The central urban area is also a heavily polluted area of PM10, which is consistent with the observations of Qi et al., (2022) and Luo et al., (2020). Numerous studies have shown that air particulate matter is one of the main causes of mortality, lung and cardiovascular complications, and its long-term effects on the human body are more serious than short-term effects (Alexeeff et al. 2022; Liu et al. 2020a, b; Xu et al. 2024). Therefore, controlling particulate matter has great health benefits. Overall, in 2019 each pollutant showed a certain downward trend compared with 2018, but the distribution characteristics were basically the same (Fig. S8). NO2 is still the key object of pollutant emission reduction. The research of Qi et al., (2022) also emphasized that the rapid development of urbanization in Chengdu-Chongqing region led to the gradual increase of motor vehicles, which is one of the main contributor of NO2. About 98.0% of the regions have O3 concentrations exceeding the limit of WHO, 2021AQGs standards. This result was similar to the findings of Chen et al., (2022a, b, c), Which also underscore that vehicle exhaust and solvent utilization contributed more to O3 production. O3 is produced by photochemical reactions of precursors such as NOx and VOCs, so the O3 concentration depends largely on the VOC/NOx value.

In 2020, except for SO2 and O3, the distribution characteristics of other pollutants are generally larger in the central urban area (Fig. 5). The areas with high PM2.5 values are distributed in the northwest of the central urban region, and gradually decrease eastward and westward. The concentration of NO2 is higher in the south of the central urban area, and gradually decreases eastward and westward from this center. This may be due to the high urbanization and high motor vehicle emissions in the central region. The average value of O3 has increased slightly, and the average value in 2020 is 169 μg/m3, which is lower in the central region and relatively higher in the new cities in the eastern and western suburbs. It is intriguing that, compared with 2019, during the lockdown of COVID-19 in 2020, except for O3 showing a slight increase trend, the concentration values of other pollutants decreased significantly, which may be due to the reduction in emissions, rather than due to the meteorological conditions with a large range (Gao et al. 2023; Konstantinoudis et al. 2021; Xia et al. 2022). Overall, all pollutants in each region showed a decrease trend from 2014 to 2020, except for the O3 concentration which showed a certain fluctuation and growth. The spatial patterns of pollutants (PM2.5, PM10, CO, NO2) all show higher concentrations in the central region. This may be due to frequent human activities and high emissions from industrial and vehicles in the central urban area (Tan et al. 2023; Zhu et al. 2019). While, O3 and SO2 showed a relatively low distribution pattern in the central urban area. Overall, the air pollution in Chengdu from 2014 to 2020 showed the characteristics of both high PM2.5 and O3 concentrations. This is generally similar to the findings of the Qi et al., (2022) and Tan et al., (2023), which emphasized that high concentrations of PM and O3 are closely related to dense traffic flow and basin topography. In addition, existing studies have shown that PM2.5 and O3 are the main pollutants in most cities. High concentrations of PM2.5 and O3 may pose a huge threat to public health and are closely related to premature death (Gómez González et al. 2023; Liu et al. 2022). Some studies have also shown that strategies to reduce PM2.5 concentrations may change the NOx-VOC ratio, thereby, adversely affecting the control of O3 (Liang et al. 2021; Xu et al. 2023). Therefore, it is particularly important to further explore the complex relationship, sources and health effects of PM2.5 and O3.

Fig. 5
figure 5

Spatial variations of six pollutants in 2020

Full size image

Temporal variation characteristics of pollutants

Based on the spatial distribution characteristics, it can be concluded that the pollutant concentration in Chengdu showed a downward trend from 2014 to 2020, and the air quality has been significantly improved. The daily concentration variation characteristics of six pollutants in 2014 and 2020 are shown in Fig. 6 and Fig. 7. It can be seen that the pollutant concentration values in 2014 were significantly higher than those in 2020, which is obviously related to the epidemic control measures in 2020 (Bashir et al. 2020; Qi et al. 2022). Among them, SO2, CO, NO2 and PM2.5 decreased more significantly, but the concentrations of O3 and PM2.5 were still high. It can be seen from the figure that the particle concentration values are higher in autumn and winter, while the O3 concentration values are higher in summer. This is consistent with the research results of Lu et al., (2022a, b), that is, PM2.5 shows a "U"-shaped change trend, while O3 shows an inverted "U"-shaped variation characteristic. High concentrations of O3 can lead to the formation of photochemical smog pollution, and long-term exposure to high concentrations of O3 will not only harm human health, but also have an adverse effect on plant growth and crop yields (Chen et al. 2022a, b, c; Wang et al. 2022a, b). Therefore, in winter, the control of precursors such as SO2, NH3 and primary PM2.5 should be strengthened in a coordinated manner to achieve coordinated prevention and control of PM2.5 and O3. In addition, the atmospheric environment is already in the transition stage from primary pollution to secondary pollution. While controlling primary pollution, it is necessary to prevent the occurrence of secondary pollution (Liang et al. 2023a, b; Mostafa et al. 2021; Zhang et al. 2023).

Fig. 6
figure 6

Temporal variations of six pollutants in 2014

Full size image
Fig. 7
figure 7

Temporal variations of six pollutants in 2020

Full size image

Influence of meteorological parameters on the air pollutants

Variation characteristics of meteorological factors

Wind speed and sunshine duration

Wind affects the diffusion and transportation of air pollutants, and plays a prominent role in the spatial distribution of air pollution. Chengdu is located in the SCB, with low wind speed all year round. The annual average wind speed between 2014 and 2020 is 1.0 m/s, 1.1 m/s, 1.3 m/s, 1.27 m/s, 1.35 m/s, 1.29 m/s, 1.27 m/s, respectively, all lower than 1.4 m /s. Overall, there are seasonal differences in wind speed during the study period, with larger values appearing in spring and summer (Fig. 8), and the highest values appearing in April-August (Fig. S9). Besides, a notable characteristic of the climate in Chengdu is cloudy with short sunshine hours. During the study period, the average annual sunshine duration was 1,038 h, and the average annual sunshine duration in 2019 was the lowest (857 h). Obvious seasonal differences can be found in the sunshine hours, with the largest values in spring and summer every year. The longest sunshine time in each year is mainly concentrated in April-August (Fig. 9).

Fig. 8
figure 8

Seasonal Wind speed

Full size image
Fig. 9
figure 9

Monthly SSD

Full size image

Relative humidity and precipitation

Fluctuations in relative humidity can affect atmospheric stability and lead to changes in pollutant concentrations. There are some differences in the average relative humidity and minimum relative humidity of various meteorological stations in Chengdu. The average relative humidity is between 68% and 88.6%, and most sites are higher than 70% (Fig. 10). The annual average relative humidity is 79.6%, and the minimum value is about 54.4%. The relative humidity shows seasonal variations, which are relatively high in autumn and winter and generally low in spring (Fig. S10). The seasonal heterogeneity of relative humidity is also one of the main bases for explaining the seasonal variation of pollutants. In addition, the maximum relative humidity is concentrated in September to December, all higher than 80%, and the minimum value occurs in February to April in most years. Precipitation is also one of the main parameters affecting the concentration of atmospheric pollutants, and wet deposition can clean the atmospheric pollutants (Jia et al. 2019; Wang et al. 2016). During 2014–2020, the precipitation at each station was inconsistent, and the precipitation was relatively high at Dujiangyan station and Pujiang station (Fig. 11). There are obvious differences in precipitation in different seasons, and the overall trend is “inverted U shape” (Fig. S11). Summer precipitation is obviously higher than other seasons, and winter is the lowest. The precipitation is mainly concentrated in July and August, and the precipitation in November, December, January and February is relatively small (Fig. S12).

Fig. 10
figure 10

RHU in the station

Full size image
Fig. 11
figure 11

Annual PRE

Full size image

Temperature, ground surface temperature and pressure

The transport and distribution of regional air pollutants are obviously affected by temperature. Higher temperature can enhance the thermodynamic conditions of the atmosphere, intensify the turbulent exchange of the atmosphere, and promote the vertical diffusion and horizontal transport of the atmosphere (Li et al. 2021; Zhan et al. 2019). From 2014 to 2020, the average temperature of each station is basically the same, about 17 °C (Fig. S13). The daily maximum temperature values at each site were lower than 25 °C, and the minimum temperature fluctuated between 12.5 °C and 15 °C. The annual average temperature is around 16.4 °C to 17.6 °C, with the highest value in summer and the lowest in winter (Fig. 12). The average monthly temperature in each year presents an inverted “U-shaped” change pattern. The highest average temperature appears in July and August, and the extreme minimum temperature appears in January, which is -6.7 °C.

Fig. 12
figure 12

Monthly TEM

Full size image

From 2014 to 2020, the annual changes of GST are relatively stable, and the annual average is about 17 °C. In addition, GST showed an inverted “U-shaped” seasonal variation in each year (Fig. S14). The monthly variation trend of GST in each year is similar, showing an inverted “U shape”. The average maximum GST occurs in July and August and the maximum GST occurs in May–August (Fig. 13). In general, the PRS annual value fluctuation of range is 946 hPa-956 hPa. High atmospheric pressure values are mainly concentrated in winter and lowest in summer. The monthly trends of each year from 20014 to 2020 are consistent, showing a “U-shaped” trend, with the maximum values concentrated in January, February and December, and the minimum values appearing in June and July (Fig. S15).

Fig. 13
figure 13

Monthly GST between 2014 and 2020

Full size image

Relationship between PM pollutants and meteorological factors

The relationship between meteorological factors and PM concentration is not a simple linear relationship, and each meteorological factor has different correlations with air quality conditions. It can be seen from Table 3 that among the meteorological parameters, only PRS is positively correlated with the concentrations of PM2.5 and PM10. Wind is a meteorological parameter closely related to the level of air pollution, and it plays a positive role in the dilution and transport of air pollutants (Flores et al. 2020; Peng et al. 2022; Zhan et al. 2019). It is clear that that wind is negatively correlated with PM2.5 and PM10, and the correlation coefficients are -0.445 and -0.427, respectively (Table 3). Furthermore, a significant negative correlation between wind speed and PM was also observed in summer, indicating that higher wind speed can accelerate the dilution of PM, thereby reducing PM concentration (Fig. 14). On the contrary, the static and steady wind will increase the mass concentration of PM. Besides, PM2.5 and PM10 had a significant negative correlation with air temperature, with correlation coefficients of -0.730 and -0.697, respectively, indicating that high temperature is also conducive to reducing PM concentration. Relevant studies have shown that one of the main reasons for the serious air pollution in the Chengdu Plain in winter is the high frequency of temperature inversions in winter and the weak intensity of cold air, which makes it difficult for pollutants to diffuse in the basin (Kong et al. 2020). High relative humidity is conducive to the secondary transformation of precursors and the retention of atmospheric particulate matter in the air, thereby promoting the chemical transformation and hygroscopic growth of aerosols (Kong et al. 2020). The findings of Fan et al.(2021) also found that the ratio of PM2.5/PM10 tends to be higher in southern Chinese cities due to higher relative humidity in winter. Kong et al.(2020) found that secondary aerosol is an important source of PM2.5 in Chengdu, which is sensitive to meteorological factors.

Table 3 Correlation between PM and meteorological factors in Chengdu
Full size table
Fig. 14
figure 14

The relationship between PM (PM10, PM2.5) and meteorological factors

Full size image

In the absence of precipitation, relative humidity is generally positively correlated with PM2.5 concentrations (Fan et al. 2021; Li et al. 2019a, b). There was a negative correlation between PM and RHU (PM2.5:R = -0.257, PM10:R = -0.334). The correlation is extremely strong in summer, indicating that relative humidity has the most significant effect on PM in summer. Between 2014 and 2020, PM2.5 and PM10 showed a negative correlation with PRE (PM2.5:R = -0.505, PM10:R = -0.521), indicating that rainfall can reduce the concentration of PM. The increase in precipitation accelerates the wet removal of pollutants, which in turn reduces the concentration of pollutants (Yang et al. 2019a, b). Most of the rainfall in Chengdu is concentrated in summer, when the concentration of PM is the lowest (Huang et al. 2015; Liu et al. 2018). The correlation coefficients between PRE and PM2.5 and PM10 in summer reached -0.598 and -0.669, respectively. Furthermore, PM concentration was also affected by SSD, showing a negative correlation between them (PM2.5:R = -0.375, PM10:R = -0.332). The higher the SSD, the higher the temperature will be, which is more conducive to the secondary conversion of PM. The fluctuation of PM concentration is opposite to that of GST (Fig. 14), and there is a significant negative correlation between them (PM2.5:R = -0.724, PM10:R = -0.668). Overall, particulate matter was only positively correlated with PRS (PM2.5:R = 0.654, PM10:R = 0.638), indicating that an increase in air pressure would promote an increase in PM concentration. The downward airflow generated by high pressure can inhibit the upward movement of PM, leading to particle accumulation, which is more obvious in winter. In general, weather conditions such as strong wind, high temperature, and low humidity can promote the dilution and diffusion of particulate matter, which is promote the reduction of particulate matter concentration.

Relationship between gaseous pollutants and meteorological factors

Wind facilitates the dispersion and transport of pollutants. It can be observed that SO2, NO2, and CO all illustrated negative correlations with wind speed, and the correlation coefficients are -0.412, -0.401 and -0.475, respectively (Table 4). However, there was a significant positive correlation between O3 and wind speed (R = 0.624). The possible reason is that higher wind speed can dilute PM and enhance solar radiation, thus accelerating the formation of O3 (Romshoo et al. 2021; Yang et al. 2020; Zhang et al. 2015). SO2, CO and NO2 were only positively correlated with PRS, and showed different negative correlations with other meteorological factors (Table 4). Furthermore, it can be observed that O3 is only negatively correlated with RHU and PRS. In addition, the concentration of SO2, NO2 and CO is relatively low when the TEM is high, while the fluctuation of O3 is basically consistent with the TEM and presents an “inverted U-shape” (Fig. 15). This is consistent with the correlation shown in Table 4, SO2, NO2 and CO have different negative correlations with TEM, and the correlation is: CO > NO2 > SO2. However, a strong positive correlation was observed between O3 and TEM (R = 0.827), studies also shown that high temperatures are one of the factors leading to severe ozone pollution (Qiao et al. 2021; Yang et al. 2020). The photochemical reaction rate in the troposphere is significantly accelerated by high temperature, which promotes the conversion rate between O3 and precursors, thereby generating abundant O3 (Kong et al. 2020; Kuang et al., 2018).

Table 4 Correlation between gaseous pollutants and meteorological factors in Chengdu
Full size table
Fig. 15
figure 15

The relationship between gaseous pollutants and meteorological factors

Full size image

Existing research has emphasized that pollutants tend to accumulate under conditions of high relative humidity and promote various secondary processes, further exacerbating regional air pollution (Li et al. 2022a, b; Ma et al. 2022). The gaseous pollutants showed different negative correlations with RHU. Besides, SO2, NO2 and CO displayed different negative correlations with PRE, and the correlation was NO2 > CO > SO2. However, a positive correlation can be observed between O3 and PRE (R = 0.457), indicating that precipitation was beneficial to the accumulation of O3. O3 is also greatly affected by SSD and shows a strong positive correlation (R = 0.819). Because the sun’s ultraviolet radiation is strong when there is sufficient sunshine, it is very conducive to the photochemical reaction to generate O3. In winter, solar radiation is significantly weakened, and the scattering and absorption of high-concentration particulate matter slows down the production of O3 (Kuang et al., 2018; Lu et al. 2022a, b). In addition, the significant negative correlations can be found between sunshine duration and pollutants SO2, NO2, and CO. In winter, except for O3, other pollutants’ concentration in this region is relatively high. SO2, NO2 and CO all expressed a significant negative correlation with GST, and there were seasonal differences in the correlation. On the contrary, it can be found that O3 has a strong positive correlation with GST (R = 0.863), which is consistent with the change trend shown in Fig. 15. This indicates that the increase in temperature is beneficial to the formation of O3. It can be observed that the concentrations of SO2, NO2 and CO are only positively correlated with PRS (Table 4). While, O3 showed a strong negative correlation with PRS (R = -0.894). Under low pressure conditions, it is easy to form a static and stable weather pattern, which in turn leads to a surg in regional O3 pollutants (Yang et al. 2020). Overall, O3 concentrations were only significantly negatively correlated with RHU and PRS. In addition, the reduction rate of VOC and NOx was proved to be the main way to reduce O3 and PM2.5 emissions in Chengdu (Kuang et al., 2018).

Pollution trajectory and potential source analysis

Back trajectory cluster analysis

Cluster analysis can not be done by only derive different classes of trajectories, but also provide valuable information for identifying the history of air masses and air pollution in specific regions (Berriban et al. 2022; Yu et al. 2019). According to the track direction and transport area distribution, the trajectories of air masses in 2020 were divided into different clusters in the four seasons (Fig. 16). In spring, air mass trajectories are clustered into seven categories, among which cluster C2 has the longest trajectory, accounting for 9.22%. Clusters C1, C5, and C6 are all from the northeast of Chengdu, accounting for 14.85%, 13.67% and 10.45%, respectively. Cluster C3 has a shorter track length but the largest proportion (30.61%). In summer, the trajectories are clustered into four categories, among which cluster C2 has the fastest transport speed, accounting for 18.98%, mainly from the northeast direction of the acceptor point. The one with the shortest track length is cluster C3, accounting for the largest proportion (33.06%). Cluster C1 is mainly transported in Chongqing and Guiyang, accounting for 26.20%. In addition, shorter air masses indicate that the air mass moves slowly and is prone to accumulation of pollutants (Berriban et al. 2022; Zhan et al. 2019). Cluster C2 is the largest among the six types of trajectories in autumn, accounting for 25.57%. The longest trajectory is cluster C6, which mainly comes from the southwest direction of Xi’an. Cluster C3 has the shortest trajectory length, indicating that the air mass moves slowly and the pollutants are mainly emitted locally. Most of the trajectories in autumn originated from the north and northeast (Xi’an, Lanzhou, Chongqing). In winter, the trajectory cluster C2 accounted for 28.05%, with the longest trajectory length. It shows that the long-distance transportation in winter mainly comes from Xi’an, Lanzhou. The track length of Cluster C3 is the shortest, indicating that the air mass transportation speed is the slowest, and it is mainly local emissions.

Fig. 16
figure 16

Seasonal trajectory cluster in 2020. Note: The gray lines in the figure represent the backward trajectory distribution of external air masses in different seasons

Full size image

Seasonal PSCF analysis of PM2.5 and O3

Areas with higher WPSCF values have a higher probability of passing pollution trajectories, that is, the trajectories passing through this area are the main transport paths that affect the concentrations of PM2.5 and O3 (Gogikar & Tyagi 2016; Zhou et al. 2023). Through the analysis, it can be observed that the concentration of PM2.5 in each hour in the summer of 2020 is lower than 75 μg/m3 and the hourly average concentration of O3 in autumn and winter is lower than 160 μg/m3. To deeply and quantitatively explore the potential source areas of PM2.5 and O3 in theses seasons, the thresholds of PM2.5 and O3 in these seasons were set at 35 μg/m3 and 100 μg/m3, respectively (GB3095-2012). From the PSCF simulation results, it can be observed that the potential source distribution regions of PM2.5 and O3 are different in different seasons. This is mainly due to significant differences in air flow in different seasons. However, the overall distribution of potential source regions is basically consistent with the trajectory transport distribution of air masses.

The potential sources of PM2.5 in 2020 have obvious seasonal differences, in which the largest WPSCF value appear in winter, and the smallest value occurs in spring and summer (Fig. 17). In spring, the potential sources of PM2.5 are generally concentrated in the east of Chengdu and the southwest of Xi’an, and the WPSCF values in these areas are all lower than 0.4. The concentration of PM2.5 in summer is lower than 75 μg/m3. Figure 17 (b) shows the potential source areas where PM2.5 is higher than 35 μg/m3 but lower than 75 μg/m3, and the main contributing sources are located in the east of Chengdu and the southwest of Chongqing. Local sources dominate in autumn, and the potential source areas of PM2.5 are mostly concentrated in the east and northeast of Chengdu. In winter, PM2.5 is affected by regional transport in addition to local contributing sources. The northeastern part of Lanzhou is the main source of contribution, and the WPSCF value is greater than 0.6. In addition, the eastern of Chengdu, the southwestern of Chongqing, and the northern part of Guiyang are the main potential sources of O3 in the spring and summer (Fig. S16). The WPSCF value in spring is higher in the southwestern part of Chongqing (0.5 < WPSCF < 0.7), indicating that this area contributes more to O3 in spring. The areas with higher WPSCF values in summer are distributed in the northwest of Guiyang, but the overall WPSCF values in summer are lower than 0.4. There is no area with O3 concentration higher than 160 μg/m3 in autumn and winter, but there are some potential source areas higher than 100 μg/m3 in the central and eastern part of Chengdu (WPSCF < 0.2), indicating that O3 concentration is lower in autumn and winter.

Fig. 17
figure 17

Source region of PM2.5 identified by PSCF in four seasons of 2020

Full size image

Seasonal CWT analysis of PM2.5 and O3

CWT model has the advantage of being able to distinguish between regions of high and moderate PM2.5 and O3 concentration. The potential source areas reflected by CWT are more abundant than the PSCF simulation results (Li et al. 2017a, b). Overall, the CWT simulation in 2020 is basically consistent with the PSCF model results, and the potential sources area of PM2.5 were found in the north of Chengdu and southwest of Chongqing (Fig. 18). However, the potential influence area of the CWT model is larger than that of the WPSCF distribution. The WCWT value was the largest in winter, and the values in all regions were lower in spring and summer. In autumn, the eastern part of Chengdu is the main potential source area, but the WCWT value of the whole potential source area is between 10 μg/m3 and 50 μg/m3. In winter, the WCWT value is larger (higher than 100 μg/m3) in the southeast of Lanzhou, indicating that regional transport has a significant impact on the distribution of PM2.5. For O3, there are also seasonal differences in the distribution of CWT simulation results in 2020, with wider distribution areas in spring and summer (Fig. S17). The WCWT value in spring is higher than 100 μg/m3 in most parts of southern Chongqing, indicating that this region is a source of high concentration of O3. In summer, the high O3 concentration source area appears in the border area between western Chongqing and Chengdu. The distribution area of WCWT in autumn and winter is relatively concentrated. The WCWT values in autumn are mainly distributed in the northeast of Chengdu, southwest of Xi’an, west of Chongqing and southeast of Lanzhou. The distribution area of winter simulation results is also small, mainly in the north of Chengdu and the south of Lanzhou, and the WCWT value is less than 50 μg/m3. High concentrations of O3 will also accelerate the formation of pollutants such as particulate matter, thereby, increasing the frequency and intensity of severe pollution weather. Therefore, PM2.5 and O3 pollution prevention and control measures must be spatially and temporally differentiated.

Fig.18
figure 18

Spatial distributions of WCWT values for PM2.5 in four seasons of 2020

Full size image

Conclusions

Scientific emission reduction in complex terrain areas under the goal of coordinated environmental and climate governance is challenging. In this research, the spatial fluctuation characteristics of the six criteria pollutants were investigated, also the relationship between the air pollutants and meteorological factors are explored in detail. And through the backward trajectory model the transmission paths and contribution of potential sources of PM2.5 and O3 in Chengdu in 2020 were analyzed. The conclusions and recommendations are as follows:

  1. (1)

    The spatial patterns of pollutants (PM2.5, PM10, NO2, CO) all show higher concentrations in the central region. While, O3 and SO2 showed a relatively low spatial pattern in the central urban area. Overall, air pollution in Chengdu (2014–2020) was characterized by both high PM2.5 and O3 concentrations.

  2. (2)

    Concentrations of pollutants (PM2.5, PM10, NO2, SO2, CO and NO2) only showed positive correlation with PRS, and displayed a negatively correlated with other meteorological elements. While, O3 is only significantly negatively correlated with RHU (R = -0.389) and PRS (R = -0.894).

  3. (3)

    Significant seasonal differences can be observed in the regional transport paths of pollutants (PM2.5 and O3) in 2020. In addition, these pollutants concentration are not only affected by local source emissions, but regional transport (Chongqing, Lhasa, Lanzhou, Xi’an, Guiyang and other cities) is also an important source of contribution. Besides, the high-concentration contribution source areas of PM2.5 and O3 mainly appear in winter and summer respectively, and the contribution source areas are centralized in Chengdu and its surrounding areas (SCB) and cities.

In general, methods such as kriging interpolation and backward trajectory models, can reflect more specific information on the spatial distribution characteristics of pollutants and potential source areas, and better represent the spatial heterogeneity of pollutants. Therefore, regional pollution prevention and control cannot be a "one-size-fits-all" approach. It requires a detailed analysis of regional distribution characteristics, and dynamic and coordinated emission reduction across the entire region based on time, area, grade, and category differences to improve the effectiveness of pollution control. Strengthen the management of key industries in the region, screen enterprises with good emission reduction effects and implement refined management and control, focusing on controlling emissions from transportation, construction, industry, etc. In addition, the results of the study show that in autumn and winter, the focus should be on reducing PM2.5 emissions, especially in the central urban area; in summer, more attention should be paid to reducing O3 pollution in surrounding counties and districts. Accurately quantifying the nonlinear relationship between precursors (such as NOx, NH3 and VOCs) and PM2.5 and O3 is the key to achieving coordinated prevention and control of pollutants, which will bring greater environmental and health benefits. Therefore, the main focus should be on the coordinated reduction of PM2.5 and O3 emissions, detailed analysis of heavy pollution incidents, and promotion of coordinated reduction of multiple pollutants. It also is necessary to continue to study the interaction mechanism of extreme weather, climate change, urbanization and pollutants. At the same time, the effectiveness evaluation of measures and policies is also crucial to improving air quality.

Finally, the impact of pollutant exposure on health should be further quantitatively evaluated to promote the positive health benefits of pollution prevention and control. Local Environment Agency Plans should also be developed as an integrated management plan to assess, prioritize and provide solutions to local environmental issues related to the air pollution and encourage stakeholders to work together to preserve and protect their local environment.