Introduction

Significant air pollution challenges have emerged in China due to the country’s rapid economic development, industrialization, and urbanization (Geng et al. 2021; Song et al. 2017; Kan et al. 2012). People often live in polluted air, which can lead to poisoning or cancer and is a major risk factor for the disease burden in China (Li and Dai 2022). In particular, the different sources and components of pollutants have a more pronounced impact on health during the cold season (Cao et al. 2009; Chen et al. 2010). According to the 2021 Actions on Quality, if governments took action to improve air quality, it would prevent 7 million premature deaths each year and help improve the overall health and economic well-being of 92% of the world’s population. Air pollution, especially fine particulate pollution, has become a serious environmental problem in China (Hu et al. 2014) and poses a threat to human health (Geng et al. 2019; Zhao et al. 2018a; Li et al. 2018a; Cohen et al. 2017). PM2.5 concentrations for all of China’s population still exceed the WHO air quality guidelines, and 81.1% of them still live in areas that exceed WHO Interim Target 1 (the least stringent WHO interim air quality objective of 35 μg m−3), demonstrating that China still faces challenges with ambient particulate matter pollution (Yin et al. 2020). Since 2013, certain large cities’ national air quality monitoring stations have started to receive approval from the Ministry of Environmental Protection for the measurement of six criterion air pollutants, including sulfur dioxide (SO2), nitrogen dioxide (NO2), carbon monoxide (CO), ozone (O3), and particulate matter (PM2.5 and PM10, with an aerodynamic diameter equal to or < 2.5 and 10 μm, respectively). These data are a prerequisite for the investigation and resolution of China’s air pollution issue, and a variety of issues have been researched using data from these monitoring sites (Fan et al. 2020; Ma et al. 2019; Guo et al. 2019; Zhou et al. 2018). Since the implementation of emission control strategies mainly in industry, residential, and transportation, China has gradually accumulated experience and achieved certain results in air pollution control. Air pollutants (except O3) concentrations in China have decreased significantly over the past few years, but concentrations of SO2, CO, PM2.5, and PM10 still remain high in northwest China (Liu et al. 2020; Fan et al. 2020). By 2017, PM2.5 concentrations over the heavily polluted regions of eastern China showed a noticeable downward trend relative to 2013 (Zhang et al. 2019; Wang et al. 2020a). The concentration of air pollutants is significantly influenced by emissions and meteorological conditions (Jhun et al. 2015; Wang et al. 2018; He et al. 2017). Unfavorable meteorological conditions can increase air pollutants when emissions remain constant, and even play a dominant role in some cases (Zhang et al. 2015, 2018). Moreover, even if local emissions are reduced, regional transport may increase air pollutants at the location of interest (Leibensperger et al. 2011; Wang et al. 2009a). Research on the six air pollutants is urgently required. Intensive academic study has currently concentrated on air pollution issues in regions of China with rapid economic development and high human activities, such as Beijing–Tianjin–Hebei, the Yangtze River Delta, the Pearl River Delta, and the Sichuan Basin (Zhou et al. 2020; Qi et al. 2017; Yuan et al. 2021; Zhao et al. 2018b); however, less attention has been paid to the northern slope of the Tianshan Mountains (NSTM).

The NSTM is the political, economic, and cultural center of the Xinjiang Uygur Autonomous Region (hereinafter Xinjiang), as well as the northwestern agglomeration of the country. In the last decade, the region’s annual GDP has exceeded 50% of the GDP of Xinjiang, gathering many industrial enterprises, with a coal-based energy structure, an industrial structure with a large proportion of heavy industry, and a transportation structure with road freight emitting enormous pollutants (Fang 2019). In the study area, the non-heating period is characterized by a large environmental capacity, favorable diffusion conditions, and positive regional air quality. However, as the heating period approaches, the local environment’s capacity for handling pollutants decreases, diffusion conditions deteriorate, and the combination of unfavorable meteorological factors influences, resulting in more frequent heavily polluted weather during the heating period. The NSTM has one of China’s highest annual average PM2.5 concentrations (Shen et al. 2019). Wang et al. (2020b) explored the spatio-temporal variation characteristics of PM2.5 concentrations in this place and concluded that PM2.5 concentrations were high and increased significantly from 2001 to 2016. Wang et al. (2014a) observed that the annual mean concentrations of PM10 on the NSTM ranged from 55 ± 28 to 92 ± 75 μg m−3 (2007–2013), with heavy anthropogenic air pollution in winter. Ma et al. (2022) used near-surface observations obtained from the national air ambient monitoring network in the NSTM region from 2015 to 2019 to examine the variation and drivers of PM2.5 and found an exceedance rate of 23.3% (54.3 μg m−3) in this region. Although some studies, as described above, have been conducted to investigate pollutants in the NSTM region, they only analyzed one or two pollutants, and there has been a paucity of studies on all six pollutants over a long period of time. Thus, this study examined six pollutant concentrations over a 5-year period to draw more attention and further research to the serious environmental problems in this region.

Using data from air quality monitoring stations (January 2017–December 2021) in the NSTM, this study analyzed the spatial and temporal characteristics of air pollutants (SO2, NO2, CO, O3, PM2.5, and PM10). Rupakheti et al. (2021) studied a similar dataset for 16 cities throughout Xinjiang from 2013 to 2019. Additionally, this study investigated the relationship between concentrations of air pollutants and meteorological factors, as well as the source distribution. In order to comprehend the origins of particulate matter in the study area, backward trajectory, PSCF (potential source contribution function), and CWT (concentration weighted trajectory) analyses were carried out based on HYSPLIT (Hybrid Single-Particle Lagrangian Integrated Trajectory) model in heavily polluted weather. Meteorological conditions are one of the most critical factors affecting air quality (Kang et al. 2019; Tian et al. 2019). Understanding the transport of air pollutants and the impact of potential source areas on the local environment is also essential. Therefore, a more comprehensive air quality study of multiple pollutants on the NSTM is necessary to better understand the air pollution situation in the region. These findings will serve as a prerequisite and the foundation for the scientific development of prevention and control strategies for air pollution.

Data and methods

Study area

The NSTM is the most developed region in Xinjiang in terms of economy, education, science, and technology, as well as an essential hub of the Silk Road Economic Belt to Central Asia and Europe (Development and Reform Commission of Xinjiang Uygur Autonomous Region 2017). The region is distributed in the arid and semi-arid climate zone, a typical temperate continental climate with scarce precipitation, intense evaporation, dusty weather, and large annual and daily temperature differences (Liu et al. 2018). The Taklamakan Desert, China’s largest desert, and the Gurbantunggut Desert, China’s second-largest desert, respectively, border the northern and southern slopes of the Tianshan Mountains, together with rapid urbanization in recent years, has led to serious urban and regional air pollution problems (Han 2018). Ten cities in the study area, from Hami in the east to Ili Kazakh in the west, are shown in Fig. 1 along with their locations and elevations. Table S1 provides more details on these cities, including longitude, latitude, elevation, population, GDP, and car parc.

Fig. 1
figure 1

Location of the study area. The DEM data (Shuttle Radar Topography Mission, SRTM) were downloaded from the Geospatial Data Cloud (http://www.gscloud.cn/) and processed using ArcGIS 10.2 software. a Map of China and location map of the Xinjiang deserts provided by National Tibetan Plateau Data Center (http://data.tpdc.ac.cn): I. Taklimakan Desert, II. Gurbantunggut Desert. b Enlarged topographic map of NSTM

Full size image

Data source and quality control

The concentrations of six criteria pollutants (SO2, NO2, CO, O3, PM2.5, and PM10) and the AQI (Air Quality Index) from 36 monitoring stations in ten cities in the NSTM from 2017 to 2021 used in this study were provided by the China National Environmental Monitoring Center. Daily temperature, relative humidity, wind speed, and wind direction were among the meteorological data that were collected from the China Meteorological Data Service Centre. Downloaded from NOAA (National Oceanic and Atmospheric Administration), the GDAS (Global Data Assimilation System) dataset was used on a global scale in the HYSPLIT model, with a spatial resolution of 1° × 1°. Detailed data information is shown in Table S2.

The quality control of the data was validated in accordance with the requirements for the National Ambient Air Quality Standards of China (NAAQS, GB 3095–2012). The values that are missing or ≤ 0 are set to invalid. The calculation of the 24-h average concentrations of each pollutant was only done where there were more than 20 h of valid data available. Eight-hour O3 concentrations were only estimated when there were at least 6 h of valid data available for every 8 h. Based on less than four invalid days of data every month, the pollutants’ monthly average concentrations were computed. When a year had more than 324 days of valid data, annual averages were declared to be valid. Citywide average concentrations were calculated by averaging the concentrations from all stations in each city. The People’s Republic of China’s National Environmental Protection Standard (HJ 630–2011) technical standards for environmental monitoring quality management were used to guide the quality assurance and control (QA/QC) of data. This approach has also been employed in a number of previous studies to demonstrate the reliability of the data quality; more details can be seen in Wang et al. (2021) and Fan et al. (2020).

Backward trajectory analysis

Using the HYSPLIT (Hybrid Single-Particle Lagrangian Integrated Trajectory) backward trajectory model in conjunction with the GDAS dataset, 72-h backward trajectories were calculated to identify potential long-range transport paths of air masses approaching cities (Draxler and Rolph 2003). The backward trajectories calculated 500 m above the ground at 00:00, 06:00, 12:00, and 18:00 (UTC) every day were obtained by cluster analysis. A Lagrangian integrated model called HYSPLIT simulates the trajectories of air parcels to determine the paths that air masses travel during transport. In the research of air mass transportation, HYSPLIT is frequently utilized (Yin et al. 2021). Additionally, TrajStat is a Geographic Information System (GIS)-based tool for identifying pollution sources, enabling visualization and statistical analysis of air quality trajectories, and is used extensively in MeteoInfo software as a valuable tool for HYSPLIT simulations (Wang et al. 2009b; Wang 2014).

PSCF and CWT method

The potential source contribution function (PSCF) method is a method developed to qualitatively identify potential sources of pollution based on a conditional probability function. By combining the trajectory of an air mass with the value of an element (e.g., the hourly concentration of a pollutant) corresponding to that trajectory, PSCF is frequently used to identify potential source areas of high concentrations of pollutants affecting receptor sites. A higher value indicates that the grid cell has a higher probability of being a potential source area causing air pollution in the target city. The calculation formula is as follows (Nicolas et al. 2011; Wang et al. 2009a, b):

$${\mathrm{PSC}}{\mathrm{F}}_{\mathrm{ij}} \, \mathrm{=} \, \frac{{\mathrm{m}}_{\mathrm{ij}}}{{\mathrm{n}}_{\mathrm{ij}}}$$
(1)

where PSCFij is the ratio of the trajectory endpoints mij exceeding the threshold criterion to all trajectory endpoints nij in grid ij. To reduce the effect of small values of nij in the grid cells, multiply Wij (the weight function) by the PSCF value as follows:

$${\mathrm{WPSC}}{\mathrm{F}}_{\mathrm{ij}} \mathrm{=}{\mathrm{PSCF}}_{\mathrm{ij}}\mathrm{\times} {\mathrm{W}}_{\mathrm{ij}}$$
(2)
$$W_{ij}\left\{\begin{array}{cc}1.00,&{\mathrm n}_{\mathrm{ij}}\;>80\\0.70,&80\;\geq\;{\mathrm n}_{\mathrm{ij}}\;>\,20\\0.42,&\;20\;\geq{\,\mathrm n}_{\mathrm{ij}}\;>\;10\\0.05,&\begin{array}{c}10\;\geq\;{\mathrm n}_{\mathrm{ij}}\end{array}\end{array}\right.\begin{array}{lc}\;\;\;\;\;\;\;\;\;\;\;\;\;\;\;\;\;\;\;\;\;\;\;\;\;\;\;\;\;\;\;\\\;\;\;\;\;\;\;\;\;\;\;\;\;\;\\\;\;\;\;\;\;\;\;\;\;\;\;\;\\\;\;\;\;\;\;\;\;\;\;\;\;\;\;\;\;\;\;\;\;\;\;\;\;\;\;\;\;\;\;\end{array}$$
(3)

The PSCF method cannot distinguish between cases where the trajectory corresponds to a concentration slightly above the threshold or much above the threshold. Therefore, this deficiency is compensated by combining CWT, which can quantitatively reflect the pollution level of the different trajectories. Higher values of CWT indicate a greater likelihood that the potential source will contribute to high pollution levels at the target cell, as follows (Hsu et al. 2003; Dimitriou and Kassomenos 2015):

$${\mathrm{CWT}}_{{\mathrm{ij}} }\;\mathrm{=} \, \frac{1}{\sum\nolimits_{{\mathrm{l}}= \mathrm{1} }^{\mathrm{M}}{\mathrm\tau}_{\mathrm{ijl}}}\sum\nolimits_{{\mathrm{l}}= \mathrm{1} }^{\mathrm{M}}{{\mathrm{C}}}_{\mathrm{l}}{\tau }_{\mathrm{ijl}}$$
(4)
$${\mathrm{WCWT}}_{\mathrm{ij}} \, \mathrm{=} \, {\mathrm{CWT}}_{\mathrm{ij}}\;\mathrm{\times}{ \, {\mathrm{W}}}_{\mathrm{ij}}$$
(5)

where CWTij is the average weighted concentration of cell grid ij, M is the total number of trajectories, τijl is the time that trajectory l stays on grid ij, and Cl is the concentration of trajectory l when it passes through grid ij. As with PSCF, Wij is introduced to reflect the uncertainty of the values in each grid cell. The resolution of each grid cell in the PSCF and CWT methods is 0.5° × 0.5°, and the PM2.5 and PM10 criteria were set to 75 and 150 μg m−3 in the PSCF analysis according to the NAAQS Grade II.

Results and discussion

Characteristics of air pollutants

Spatio-temporal characteristics of the air pollutants

The 5-year averages (Table S3) and year-to-year pollutant concentration variations (Fig. 2) were shown in the NSTM and each city (January 2017–December 2021). The annual mean concentrations of SO2, NO2, CO, O3, PM2.5, and PM10 in the NSTM region ranged from 8.61 to 13.76 μg m−3, 26.53–36.06 μg m−3, 0.79–1.31 mg m−3, 84.29–87.62 μg m−3, 37.98–51.10 μg m−3, and 84.15–97.47 μg m−3, respectively. Except for O3, the concentrations of SO2, NO2, CO, PM2.5, and PM10 exhibited a decreasing trend from 2017 to 2021, at an average rate of − 1.265, − 1.779, − 0.128, − 3.042, and − 3.135 per year, respectively, while the concentration of O3 showed a slight increasing trend (at an average rate of 1.90% per year), and the trends of pollutants in the NSTM and the cities are shown in Table S4. The significant decreases in the annual average concentrations of pollutants (except O3) were attributed to specific environmental protection efforts. Among them, the concentrations of NO2, PM2.5, and PM10 in Urumqi decreased by 23.10%, 42.25%, and 36.11% in 2021 compared to 2017, respectively, with the largest decreasing trends (− 3.047, − 6.496, and − 11.828 μg m−3 year−1), indicating that the governmental regulation of air quality in the NSTM was initiated from Urumqi, especially for more effective control of particulate matter. The increase in O3 concentration was associated with VOCs and NOx (nitrogen oxides) formation from vehicle exhaust, as well as the reduction of PM2.5 and NO2, which was beneficial to O3 formation (Li et al. 2018b; Liu et al. 2020). In addition, the annual mean concentrations of NO2 and PM10 in 2020 were lower than in 2021, which is associated with a significant reduction in urban traffic due to the stagnation of production activities caused by COVID-19 (Wang et al. 2020c). Rupakheti et al. (2021) had done air pollutants characterization research throughout Xinjiang, and the results showed a slight decreasing trend in the concentrations of SO2, CO, and PM2.5 from 2015 to 2018, while the concentrations of NO2, O3, and PM10 showed an increasing trend. To more clearly compare the changes over a long period of time, we added concentrations of air pollutants in 2015 and 2016. From 2015 to 2021 (Fig. S1), SO2 concentrations have been decreasing; NO2, CO, PM2.5, and PM10 show an increasing trend followed by a decreasing trend; and O3 concentrations show an increasing trend. The comparison found that the trend in the NSTM from 2015 to 2018 was generally consistent with the whole of Xinjiang. The NO2/SO2 ratio is a crucial indicator for identifying whether the air pollutant originates from stationary (e.g., industrial emissions) or fuel-powered mobile sources (Aneja et al. 2001). The higher NO2/SO2 ratio indicates a greater contribution to air pollution from mobile sources, primarily from motor vehicle emissions (Song et al. 2017). The Yangtze River Delta, Huai River, Chengdu–Chongqing, and the Pearl River Delta NO2/SO2 ratios were higher in the pre-2020 period with 9.07, 7.40, 7.14, and 6.69, respectively (Bai et al. 2022). Due to the economic reliance on transportation in these high-tech industrialized urban agglomerations, mobile sources produce more air pollution than stationary sources (Zhao et al. 2018a; Song et al. 2017). While a relatively low ratio of 2.82 was observed in this study, this was due to the fact that the cities in the NSTM region were resource-based cities with an aggregation of heavy industries, and thus stationary source emissions such as coal and biomass burning, which were the primary factors restricting air quality (Turap et al. 2019).

Fig. 2
figure 2

Average values of concentrations of air pollutants during 2017–2021 in the NSTM region

Full size image

Among the ten cities in the NSTM, the lowest annual average SO2, NO2, PM2.5, and PM10 concentrations were observed in Tacheng, with values of 4.63 μg m−3, 11.96 μg m−3, 12.65 μg m−3, and 41.71 μg m−3, respectively. Tacheng’s northernmost distant and sparsely populated location (area: 4356 km2, population: 1.14 million, CAR PARC: 0.36 million, http://www.xjtc.gov.cn/info/2727/183468.htm) in the NSTM was the reason for the lowest concentrations of pollutants there. The highest annual average SO2 and CO concentrations were found in Ili Kazak, with values of 17.44 μg m−3 and 1.46 mg m−3, due to the typical coal-smoke air pollution in Ili Kazak (Zhong et al. 2022), where the main source of coal-smoke pollution is coal combustion and the main pollutants are SO2, CO, and smoke, in addition to NOx. The highest annual average concentrations of NO2 (41.91 μg m−3), PM2.5 (66.52 μg m−3), and PM10 (160.57 μg m−3) were observed in Urumqi, Wujiaqu, and Turpan, respectively. The highest concentrations of NO2 were found over Urumqi, an economic and political center with more anthropogenic activities than other places, and high NO2 values were mainly caused by NOX from traffic-related sources. Notably, according to the Chinese NAAQS annual average PM2.5 concentration threshold of 35 μg m−3, the 5-year average PM2.5 concentrations in Shihezi and Wujiaqu were 59.23 μg m−3 (69.23% exceedance) and 66.52 μg m−3 (90.01% exceedance), which even exceeded the Beijing-Tianjin-Hebei PM2.5 concentration of 57.98 μg m−3 in 2015–2019 (Luo et al. 2022). Moreover, the concentration of PM10 in Turpan significantly exceeded the NAAQS guideline. This indicated that the particulate matter pollution in the NSTM was extremely serious, and more attention should be paid to air pollution in this area.

The average AQI in the NSTM from 2017 to 2021 was 86.61, which reached the national secondary standard. This result included 67 and 290 yearly average days that met the primary and secondary standards, respectively, accounting for 18.41% and 61.32% of the total number of days. Over the past 5 years, the annual mean AQI has been 91.98, 90.44, 85.04, 82.70, and 82.91, respectively, indicating an improving trend in ambient air quality in the NSTM. Tacheng exhibited the lowest annual average AQI of 51.77 ± 3.99, with the lowest annual mean value of 48.04 in 2020. However, the annual average AQI of Turpan, Wujiaqu, and Shihezi were 115.38 ± 7.04, 110.27 ± 5.28, and 103.73 ± 3.81, respectively, with an annual average of 45.8, 24.2, and 58.2 days of heavily polluted weather, indicating that the three cities suffered from severe pollution. Turpan was severely affected by sand and dust, especially during the spring, when PM10 concentration even exceeded 2000 μg m−3 and AQI reached the upper limit (with a value of 500). The pollution in Wujiaqu and Shihezi was particularly severe in winter, when the rate of heavy pollution (AQI > 200) was as high as 57.21% and 46.12%, respectively, so the government should implement appropriate air quality improvement measures according to the pollution characteristics of each city.

PM2.5/PM10 ratios can identify different source approaches and processes of formation (Blanco-Becerra et al. 2015). Low PM2.5/PM10 indicated that natural processes associated with sand particles were primarily responsible for air pollution (Abuelgasim and Farahat 2020), whereas high ratios were associated with NO3−, SO42−, NH4+, and organic matter, which are secondary particles generated (Chan et al. 2005). For all of China, the ratio was 0.59 (Song et al. 2017), but the ratio was much higher in megacities such as Beijing, Shanghai, and Guangzhou, even exceeding 0.7 (Wang et al. 2014b), indicating a major contribution of finer anthropogenic particles (Xu et al. 2017). With a decreasing interannual variation, the mean PM2.5/PM10 ratio in the NSTM was 0.45 ± 0.02; higher ratios were found in Ili Kazakh, Urumqi, and Shihezi (0.56 ± 0.02, 0.52 ± 0.04, and 0.52 ± 0.01, respectively), while lower ratios were found in Hami, Tacheng, and Turpan (0.34 ± 0.05, 0.35 ± 0.02, and 0.37 ± 0.03, respectively). Apart from the above, the lowest PM2.5/PM10 ratio of 0.27 was observed in Hami in 2021, while the highest value of 0.63 was found in Changji in 2017. During the 5 years of the study, the NSTM’s PM2.5/PM10 ratio was lower than that of the majority of Chinese cities. The atmosphere was more impacted by coarse particulate matter, indicating that the NSTM is vulnerable to sandstorms in the Taklamakan and Gurbantunggut Deserts.

Seasonal variations

Figure 3 shows seasonal fluctuations in air pollution concentrations over various cities in the NSTM. Winter > autumn > spring > summer was the order of the seasonal variation of SO2, NO2, and CO concentration values in almost all cities. Summer had the lowest SO2, NO2, and CO concentrations, which ranged from 5.13 ± 2.25 μg m−3 (Tacheng) to 9.65 ± 2.09 μg m−3(Bozhou), 8.85 ± 3.37 μg m−3 (Tacheng) to 29.13 ± 7.88 μg m−3 (Changji), and 0.34 ± 0.18 mg m−3 (Bozhou) to 0.73 ± 0.28 mg m−3 (Tacheng). Accordingly, winter was the season with the highest SO2, NO2, and CO concentrations, ranging from 4.62 ± 1.57 μg m−3 (Tacheng) to 33.37 ± 13.06 μg m−3 (Ili Kazak), 13.89 ± 2.68 μg m−3 (Tacheng) to 62.60 ± 12.72 μg m−3 (Urumqi), and 1.08 ± 0.39 mg m−3 (Tacheng) to 2.87 ± 0.67 mg m−3 (Ili Kazak). The highest concentration of NO2 in Urumqi was due to relatively high human activity and coal burning in winter, while the lowest values were found in Tacheng with minor anthropogenic activities. The combined impacts of primary emissions from domestic heating, adverse diffusion circumstances, and photochemical processes were responsible for this outcome (Zhao et al. 2015). However, the trend for O3 was the opposite of the above, with the seasonal order of concentrations being summer > spring > autumn > winter. Similar findings in China were obtained by Li et al. (2019). The mean O3 concentration during summer ranged from 95.35 ± 8.05 μg m−3 (Tacheng) to 124.04 ± 9.47 μg m−3 (Shihezi), while in winter, the concentrations ranged from 45.25 ± 17.62 μg m−3 (Urumqi) to 77.60 ± 17.23 μg m−3 (Tacheng). This was primarily caused by the intense solar radiation, high temperatures, and potent photochemical processes that characterize summer, all of which favor the creation of O3.

Fig. 3
figure 3

Seasonal characteristics of concentrations of air pollutants from 2017 to 2021

Full size image

For almost all of the cities, the concentrations of PM2.5 and PM10 were highest in the winter and lowest in the summer. However, Hami and Turpan were found to have the highest PM10 concentrations in the spring season, which may be frequently influenced by dust storms and floating dust. The highest PM2.5 concentration occurred in Wujiaqu in winter (172.79 ± 42.48 μg m−3) and the lowest in Tacheng in summer (9.21 ± 1.96 μg m−3), with a ratio of up to 19 times between them. Similarly for PM10, the value in Wujiaqu in winter was 7 times higher than that in Tacheng in summer. That was mainly caused by the use of large quantities of coal for heating, poor weather conditions, and a significant rise in traffic over the Chinese New Year, all of which can lead to a rise in the accumulation of particulate matter and pollution events during winter.

Seasonal mean AQI values ranged from 67.42 ± 7.27 (summer) to 130.84 ± 52.78 (winter) in the study area, indicating that the air quality was markedly better in summer than in winter. Due to the high density of heavy industry in the NSTM region and the massive amount of coal that will be used for heating in the winter, the pollution levels are quite high. The total amount of pollutant emissions exceeds the local environmental capacity during the winter due to decreased environmental capacity, worsening diffusion conditions, and the addition of unfavorable meteorological conditions. In summer, there was no coal consumption for central heating, although coal consumption for power generation was also higher, due to stronger airflow convection and more precipitation, resulting in lower pollutant concentrations. As a result, air quality is worse in winter than in summer. The highest AQI value was found in the winter in Wujiaqu (216.98 ± 47.21), while the lowest AQI value was found in the autumn in Tacheng (47.90 ± 7.43). However, the values for Hami were higher in spring, and this difference can be attributed to the impact of coarse particles brought by dust storms. Seasonal mean PM2.5/PM10 ratios ranged from 0.36 ± 0.05 (summer) to 0.66 ± 0.14 (winter) in the study area. The PM2.5/PM10 ratios showed the highest ratios in winter, which was associated with anthropogenic pollution, with the highest value occurring in Urumqi (0.87 ± 0.13), followed by Ili Kazak and Shihezi with values of 0.79 ± 0.05 and 0.79 ± 0.07, while lower rates were observed in spring and summer in Hami (0.28 ± 0.05, 0.30 ± 0.05), Tacheng (0.33 ± 0.06, 0.28 ± 0.03), and Turpan (0.31 ± 0.04, 0.32 ± 0.03). That was caused by significant emissions of coarse particulate matter from sandy and wasteland areas as well as the frequent dust storms in semi-desert and desert regions of northwest China (Zou and Zhai 2004; Zhang and Cao 2015; Song et al. 2017).

Influence of meteorological conditions

Concentrations of air pollutants depend primarily on the influence of meteorological conditions when local emissions are relatively stable (Li et al. 2012). Qi et al. (2021) found that the influence of individual meteorological factors on concentrations of air pollutants was limited, while a combination of meteorological factors was more effective. Due to the availability of data, Urumqi, the capital of Xinjiang province, was chosen as a representative to analyze the relationship between pollutant concentrations and meteorological conditions. The meteorological factors include daily temperature (T), relative humidity (RH), wind speed (WS), wind direction (WD), and pressure (P).

Pearson correlation coefficients between the main meteorological factors and concentrations of air pollutants (Fig. 4) were calculated using SPSS (Guo et al. 2019). The correlations between pollutant concentrations and meteorological factors varied significantly with the seasons, with the smallest correlation in winter, which was due to the high concentrations of pollutants (except O3) in winter that were also associated with pollution emissions, topography, etc. In all seasons, temperature and O3 showed a significant positive correlation, with high temperatures and solar radiation favoring the formation of O3 in the atmosphere. Wind speed exhibited a positive correlation with concentrations of air pollutants, because an increase in wind speed under unique geographic conditions and wind direction led to the accumulation of concentrations of pollutants. In addition, relative humidity showed a negative correlation with concentrations of air pollutants, and the correlation between air pressure and concentrations of pollutants was minimal. We performed the analysis of temperature, relative humidity, wind speed, and wind direction to learn more about the meteorological factors that influence air pollutants (Figs. 5 and 6).

Fig. 4
figure 4

Correlation coefficient of meteorological factors and concentrations of air pollutants by season (diamond shape indicates p < 0.05)

Full size image
Fig. 5
figure 5

Relationship between concentrations of six air pollutants and meteorological factors

Full size image
Fig. 6
figure 6

Windrose of six air pollutants

Full size image

NO2, CO, and PM2.5 were found to be significantly negatively correlated with temperature, and atmospheric convection can be found to play an important role in their concentration regulation. Under high-temperature conditions, strong turbulence and other thermal activities of the atmosphere accelerate their diffusion (Yang et al. 2016a, b), and conversely, low temperatures weaken atmospheric convection and enhance their accumulation (Chen et al. 2020; Li et al. 2015, 2014). The highest concentrations of NO2, CO, and PM2.5 were reached at temperatures of − 13 to − 6 °C rather than at the lowest temperatures (Fig. 5). Due to the lower surface temperatures in winter, an increase in temperature might cause a temperature inversion layer to form that would prevent atmospheric motion and result in an accumulation of pollutants (Xu et al. 2019). SO2 and PM10 concentrations showed a slightly negative correlation with temperature. Similarly, at temperatures between –13 and –6 °C, the maximum concentration of PM10 (> 300 μg m−3) was recorded, but the SO2 concentration remained lower (< 50 μg m−3, below the NAAQS Grade Ι). O3 showed a strong positive correlation with temperature. When the temperature exceeds 15 °C, the concentration of O3 can reach more than 100 μg m−3, and some even exceed the NAAQS Grade II (> 160 μg m−3). The same results as Atkinson (2000) were obtained that high temperatures and low humidity favor the formation of O3. Pollutants were all slightly negatively correlated with RH. The highest mass concentrations were found at 30 ~ 50% RH, which could be attributed to the surge of pollutants emitted by centralized heating and the suppression of the atmospheric boundary layer during the winter. Humidity greater than 80% was concentrated in March, April, May, and October, when pollutants were more easily dispersed due to higher wind speeds. In addition, as the relative humidity rises, suspended particles assemble and become heavy enough to enable both dry deposition (particle drop) and wet deposition (precipitation) (Li et al. 2015).

Wind speed plays a vital role in the concentration of pollutants. Wind can enhance the flow of the atmosphere, which promotes the diffusion, dilution, and transmission of pollutants, effectively lowering their concentration. As seen in Fig. 4, the correlations between wind speed and concentrations of air pollutants were weak, indicating that concentrations of pollutants may be more closely related to wind direction than wind speed, which determines the source of pollution. In order to better understand the effect of wind direction and wind speed on concentrations of air pollutants, they were analyzed with the support of the wind rose diagram (Fig. 6). It demonstrated that the northeast and west winds are mostly responsible for the high concentrations of SO2, NO2, and CO when the wind speed is less than 2 m s−1. When the wind speed was more than 2 m s−1, the north winds were followed by the northwest, southeast, and west winds as the directions with the highest SO2, NO2, and CO concentrations. The highest O3 concentrations were primarily from the northwest and south winds when the wind speed was less than 3 m s−1; while the wind speed was greater than 3 m s−1, the highest O3 concentrations were primarily from the southwest wind, followed by the northeast wind. The PM2.5 concentrations were higher in the east wind at all wind speeds. In addition, high PM2.5 concentrations (> 75 μg m−3) also came from the northwest, north, and southwest winds at high wind speeds. The results suggest that the concentrations of air pollutants may be mainly influenced by local emissions and external transport. West winds were correlated with higher concentrations of NO2, CO, and PM2.5, which may be brought about by massive industrial emissions and frequent human activities in the western economic development zones and may also be transmitted from Changji and Shihezi upward in the west. High PM10 concentrations (> 150 μg m−3, exceeding NAAQS Grade II) were mainly associated with high wind speeds, mainly from the northwest, northeast, and southeast winds. Furthermore, at various wind speeds, PM10 concentrations mostly exceeded the NAAQS Grade Ι (50 μg m−3). It indicated that the local PM10 was mainly transported regionally and that the pollution was more serious.

Transport path and potential sources of particulate matter

Backward trajectory analysis during winter

The backward trajectory and cluster analysis reveal the transport paths of air masses reaching Wujiaqu, Shihezi, Changji, Urumqi, and Turpan, where particulate matter concentrations exceed NAAQS Grade II and air quality is poor in winter (Fig. 7). The distributions of the backward trajectory cluster in these five cities over the winter are shown in Fig. 8. Using MeteoInfo software, trajectories from the winter of 2017–2021 were classified according to the speed and direction of their transmission. Furthermore, a total of 1804 trajectories were obtained for each city, and the multiple trajectories were divided into four clusters. To account for the impact of each pathway on air quality in the five cities, the average pollutant concentration for each cluster can be calculated using the cluster statistics function of MeteoInfo’s trajStat plug-in.

Fig. 7
figure 7

Air quality conditions and primary pollutants in heavily polluted cities

Full size image
Fig. 8
figure 8

The trajectories of heavily polluted cities in winter

Full size image

In Wujiaqu, 60.09% of the air masses (clusters 1 and 4) originated in eastern Kazakhstan and then traveled via Ili Kazak. 22.45% (cluster 2) originated in Tacheng, passed through the southern Karamay and Changji, arrived in eastern Urumqi, and then turned around to head toward Wujiaqu after being obstructed by the Tianshan Mountains. Cluster 3 (17.46%) mainly came from northeastern Kazakhstan via northeastern Tacheng. Cluster 2 in Wujiaqu contained the highest PM2.5 (196.4 μg m−3) and PM10 (281.24 μg m−3) concentrations (Fig. 9), indicating that the cluster contributes significantly to the PM2.5 and PM10 concentrations in Wujiaqu. It showed that regional environmental cooperation should be strengthened, and joint prevention and control measures for air pollution should be implemented. The second highest concentrations were in cluster 1, with PM2.5 and PM10 concentrations as high as 192.58 μg m−3 and 270.58 μg m−3, respectively. With high pollution levels in Wujiaqu, it demonstrated that particulate matter was still a significant air pollutant to be concerned about.

Fig. 9
figure 9

Trajectory ratios and PM concentrations of each trajectory cluster

Full size image

In Shihezi, air clusters were mainly from the west, with a total of 75.88% (cluster 2 and cluster 1). Cluster 2 (39.41%) originated in eastern Kazakhstan and passed through Ili Kazak, Tacheng, and Changji. Cluster 1 from near the Ili River accounted for 36.74%, then crossed Tacheng and Karamay. Among them, clusters 1 and 2 were interspersed in the northern part of Ili Kazak. Cluster 3 originated in the eastern part of Tacheng, and Cluster 4 was from the north of Changji and local emissions, accounting for 14.69% and 9.42%, respectively. With 153.86 μg m−3 of PM2.5, Cluster 2 had the highest concentration. Clusters 1 and 3 corresponded to PM2.5 concentrations of 147.38 μg m−3 and 147.26 μg m−3, respectively, indicating that the areas through which the airflow passes play an essential role in the PM2.5 concentration in Shihezi. Cluster 4, of which the trajectories made up only 9.42% of all trajectories, contained the highest PM10 concentration (204.79 μg m−3). The above indicated that the surrounding areas contribute significantly to the concentration of PM10.

In Changji, clusters 1, 2, and 4 all originated in eastern Kazakhstan, with contributions of 46.18%, 12.20%, and 27.55% to the total trajectories (totally accounting for 85.93%). Clusters 1 and 2 were two trajectories of different speeds, passing through Ili Kazak and south of Tacheng. Cluster 4 crossed central Tacheng and Karamay. 14.08% of the total trajectories (cluster 3) were from the surrounding areas of Changji, which crossed Urumqi and Turpan. Cluster 2 contained the highest PM2.5 concentrations (165.87 μg m−3) and the highest PM10 concentrations (228.88 μg m−3). Cluster 1 contained the second highest concentrations of PM2.5 and PM10 (151.28 μg m−3 and 207.94 μg m−3, respectively). The results showed that eastern Kazakhstan, Ili Kazakh, and Tacheng were the most crucial transport pathways affecting the concentrations of particulate matter in Changji in winter.

In Urumqi, the trajectories of clusters 1, 2, and 4 were similar to those of clusters 1, 4, and 2 in Changji, accounting for 66.13%, 8.92%, and 17.52%, respectively, adding up to 92.57% of all trajectories from eastern Kazakhstan. Unlike the Changji trajectories, cluster 3 (7.43%) originated from the Turpan Basin. Compared to the other four cities, Urumqi showed the slightest difference in particulate matter concentrations between the four clusters, with PM2.5 (PM10) concentrations ranging from 116 to 135 μg m−3 (134 to 154 μg m−3). It showed that eastern Kazakhstan, Ili Kazak, and southern Changji were the most influential transmission pathways for particulate matter concentrations in Urumqi in winter.

In Turpan, the backward trajectories mainly originated from local emissions. Cluster 1 represented the main trajectory, which accounts for 39.86% of the total trajectories. Besides, the airflow was dominated by the northwest direction. Cluster 4 (31.71%) originated in eastern Kazakhstan and passed through Ili Kazak, Bozhou, Tacheng, Changji, and Urumqi. 14.30% (cluster 3) came from Karamay and across Tacheng, Changji, and Urumqi. 14.14% of the total trajectories (cluster 2) were from the north of the Taklamakan Desert. With a mean concentration of 93.19 μg m−3 across all four clusters, PM2.5 concentrations ranged from 92 to 95 μg m−3. The highest PM10 concentrations were found in clusters 4 and 2, with values of 217.79 μg m−3 and 195.07 μg m−3, respectively. The desert, which could deliver sand and dust to Turpan, was the source of these two clusters. Thus, the concentration values of PM10 were higher.

It is worth noting that all five cities have clusters originating in eastern Kazakhstan and have a significant share of the total trajectories. In order to supply heavy industry, there were coal mines in eastern Kazakhstan. The two primary industrial sectors that produce the majority of emissions are electricity generation and metallurgy. The eastern cities of Almaty and Ust-Kamenogorsk were the two cities with the highest monthly average concentrations of pollutants in 2019 (Assanov et al. 2021a), where the severe pollution in Ust-Kamenogorsk was not primarily caused by transport but rather by large industrial and power plants (Assanov et al. 2021b). And the two biggest coal-fired power plants were located in Ekibastuz in the northeast, indicating that the NSTM also received regional air pollutants from its borders. However, due to the long distance of the trajectories from Kazakhstan and greater air mass speed, the probability of pollution transfer was reduced, and the issue of pollution probability will be specified in the subsequent PSCF and CWT analyses. Furthermore, the shorter air clusters suggested that the air masses flowed slowly and that pollutants were ready to accumulate (Yu et al. 2019). The shorter clusters of these five cities’ clusters are as follows: cluster 2 in Wujiaqu, clusters 3 and 4 in Shihezi, cluster 3 in Changji, cluster 3 in Urumqi, and cluster 1 in Turpan. The observation was attributed to heavy local emissions along with natural sources of atmospheric pollutants.

Potential sources of particulate matter

The backward trajectory model can be used to identify the paths that air masses travel during transport, but it cannot be used to quantify the contribution of potential sources that could affect particulate matter. Therefore, to determine the potential regions, we employed the CWT and PSCF methods. The findings of the weighted potential source contribution function (WPSCF) and weighted concentration weighted trajectory (WCWT) maps for PM2.5 and PM10 in Wujiaqu (a), Shihezi (b), Changji (c), Urumqi (d), and Turpan (e) are shown in Figs. 9 and 10.

Fig. 10
figure 10

PSCF maps of particulate matter in winter (The left graphs depict PM2.5, while the right graphs depict PM10)

Full size image

High WPSCF scores (Fig. 10) suggest potential source areas for severe pollution in the corresponding regions. During the winter period in Wujiaqu, the distribution and morphology of potential contribution sources of PM2.5 and PM10 were similar, except that the potential contribution sources of PM2.5 corresponded to higher WPSCF values. The junction of Turpan, Urumqi, Changji, Tacheng, Ili Kazak, and Bayingol Mongolian Autonomous Prefecture (hereinafter Bayingol Mongolian) corresponds to the area of higher WPSCF values for PM2.5 (WPSCF > 0.9) and PM10 (WPSCF > 0.8). For Shihezi, the regions with high WPSCF values for PM2.5 (WPSCF > 0.9) and PM10 (WPSCF > 0.8) were similar, mainly in northeastern Aksu, northwest of Turpan, and the junction of Aksu and Bayingol Mongolian. Compared to PM2.5, the area of high-value WPSCF was reduced for PM10. For Changji, the Bayingol Mongolian’s southern region and the intersection of Hami and Turpan had the highest concentrations of PM2.5 WPSCF values, which were more than 0.9. Besides, areas with high WPSCF values were discovered along Ili Kazak and in the southeast of Kazakhstan, which were in the path of leading clusters 1 and 2 of higher PM2.5 concentrations (151.28–165.87 μg m−3). This was associated with Almaty, a large heavy industry city in southeast Kazakhstan, which had one of the highest annual average concentrations of total suspended particulate matter and also had the largest population and transport vehicles (Assanov et al. 2021a, b). Northwestern Bayingol Mongolian was the main potential contribution source for PM10, with a WPSCF value higher than 0.8. For Urumqi, the intersection of Aksu and Bayingol Mongolian had the highest PSCF values for PM2.5 (WPSCF > 0.9). The WPSF values for PM10 were generally not high, and higher values (WPSCF > 0.5) were found in northern Bayingol Mongolian. For Turpan, the WPSCF values for PM2.5 were higher than 0.7, mainly in the eastern of Aksu, the northern of Bayingol Mongolian, Turpan local emission, and the junction of Urumqi, Changji, and Bayingol Mongolian. As for PM10, the primary potential sources (WPSCF > 0.8) in Turpan were western Hami, the junction of Aksu, and Bayingol Mongolian. Due to its long-distance migration, the Taklamakan Desert region was one of the main potential contributors to both PM2.5 and PM10 in Turpan, which was associated with the contribution of sandstorms (Wang et al. 2015).

The PSCF method can only account for the contribution rate of the potential source area and cannot distinguish between trajectory concentrations that are slightly above the threshold and those that are substantially above the threshold. Therefore, CWT is utilized to determine the weight concentration of trajectories to remedy this deficiency and to reflect the pollution level of different trajectories quantitatively (Fig. 11).

Fig. 11
figure 11

CWT maps of particulate matter in winter (The left graphs depict PM2.5, while the right graphs depict PM10, Unit: μg m−3)

Full size image

In Wujiaqu, the higher WCWT values for PM2.5 and PM10 were distributed in the same regions, Urumqi, northwestern Turpan, northern Bayingol Mongolian, and Ili Kazak, where PM2.5 > 200 μg m−3 and PM10 > 280 μg m−3 were found. In Shihezi, the high WCWT values of PM2.5 (> 270 μg m−3) were located at the intersection of Aksu and Bayingol Mongolian, while high WCWT values of PM10 (> 200 μg m−3) were distributed in northeastern Aksu and southwestern Changji. In Changji, the region of high WCWT values for PM2.5 (> 100 μg m−3) was similar to the area corresponding to WPSCF > 0.5, covering almost the entire study area and southeastern Kazakhstan. The area of high PM10 WCWT was also wide, with NSTM, northeastern Aksu, and northwestern Bayingol Mongolian contributing to PM10 > 200 μg m−3. In Urumqi, the regions of high WCWT values of PM2.5 (WCWT > 140 μg m−3) were similar to the regions with high WPSCF values (WPSCF > 0.8). In addition, the highest WCWT values (> 200 μg m−3) were found in the intersection of northeast Urumqi and Changji. The highest WCWT values for PM10 were positioned in western Bozhou and central Tacheng, where cluster 2 (PM10 concentration of 134.66 μg m−3) passes through the highest WCWT value area. In Turpan, the potential source areas with high WCWT values of PM2.5 were in general agreement with the results of WPSCF. The high WCWT values (> 100 μg m−3) were distributed in the Aksu, Turpan, and Bayingol Mongolian. The areas with the highest PM10 WCWT (> 210 μg m−3) were the Urumqi–Changji–Shihezi region, southeast Changji, and the junction of northwest Bayingol Mongolian and southeast Ili Kazak. The results indicated that regional pollution transport could significantly impact the air quality in the study area.

Conclusion

The characteristics of six criteria pollutants, influence of major meteorological factors on concentrations of air pollutants, and potential source regions of particulate matter in heavily polluted cities were investigated for the NSTM. The concentrations of air pollutants (SO2, NO2, CO, PM10, and PM2.5) showed a downward trend from 2017 to 2021, while the opposite was observed for O3. The highest concentrations of air pollutants (apart from O3) were found in winter, especially in Wujiaqu, Shihezi, Changji, Urumqi, and Turpan, where the particulate matter concentrations exceed NAAQS Grade II. Temperature showed a significant negative correlation with concentrations of air pollutants (excluding O3), and relative humidity showed a weak negative correlation to air pollutants, which may be related to the deposition due to increased humidity. High concentrations of pollutants were more influenced by the west wind, which was related to the emission of pollutants from local economic development zones and the transmission of pollutants from the Changji–Shihezi region. At higher wind speeds, the high concentration of coarse particles was more associated with the long-distance transmission of sand and dust. The results of PSCF and CWT indicated that the WPSCF values of PM2.5 were higher in heavily polluted regions during winter, which was mainly influenced by centralized heating and unfavorable diffusion conditions, while the WPSCF values of PM10 were larger in Turpan, indicating that sandstorms have more influence on it. The potential source areas were mainly distributed in Urumqi–Changji–Shihezi, Turpan, the northern Bayingol Mongolian, and eastern Kazakhstan. The next step aims to combine local emission data for a more comprehensive source analysis of pollutants. In general, air quality improvement should be based on reducing local emissions, strengthening regional cooperation, and researching transboundary transport of air pollutants.