Temporal and Spatial Variation of PM2.5 in Xining, Northeast of the Qinghai–Xizang (Tibet) Plateau

PM2.5 was sampled from January 2017 to May 2018 at an urban, suburban, industrial, and rural sites in Xining. The annual mean of PM2.5 was highest at the urban site and lowest at the rural site, with an average of 51.5 ± 48.9 and 26.4 ± 17.8 μg·m−3, respectively. The average PM2.5 concentration of the industrial and suburban sites was 42.8 ± 27.4 and 37.2 ± 23.7 μg·m−3, respectively. All sites except for the rural had concentrations above the ambient air quality standards of China (GB3095-2012). The highest concentration of PM2.5 at all sites was observed in winter, followed by spring, autumn, and summer. The concentration of major constituents showed statistically significant seasonal and spatial variation. The highest concentrations of organic carbon (OC), elemental carbon (EC), water-soluble organic carbon (WSOC), and water-soluble inorganic ions (WSIIs) were found at the urban site in winter. The average concentration of F− was higher than that in many studies, especially at the industrial site where the annual average concentration of F− was 1.5 ± 1.7 μg·m−3. The range of sulfur oxidation ratio (SOR) was 0.1–0.18 and nitrogen oxidation ratio (NOR) was 0.02–0.1 in Xining. The higher SO42−/NO3− indicates that coal combustion has greater impact than vehicle emissions. The results of the potential source contribution function (PSCF) suggest that air mass from middleand large-scale transport from the western areas of Xining have contributed to the higher level of PM2.5. On the basis of the positive matrix factorization (PMF) model, it was found that aerosols from salt lakes and dust were the main sources of PM2.5 in Xining, accounting for 26.3% of aerosol total mass. During the sandstorms, the concentration of PM2.5 increased sharply, and the concentrations of Na+, Ca2+ and Mg2+ were 1.13–2.70, 1.68–4.41, and 1.15–5.12 times higher, respectively, than annual average concentration, implying that aerosols were mainly from dust and the largest saltwater lake, Qinghai Lake, and many other salt lakes in the province of Qinghai. Time-of-flight secondary ion mass spectrometry (ToF-SIMS) was utilized to study the surface components of PM2.5 and F− was found to be increasingly distributed from the surface to inside the particles. We determined that the extremely high PM2.5 concentration appears to be due to an episode of heavy pollution resulting from the combination of sandstorms and the burning of fireworks.


Introduction
Due to rapid urbanization, industrialization, and economic growth in China, many cities (especially in some developed or industrial areas like northern and eastern China) experience frequent haze The industrial site (GID; N36°32′, E101°31'' 2590 m above sea level) was at the Ganhe industrial district, about 36 km southwest away from FPH. The air sampler was about 30 m away from the nearest street, Ganqinger, and set on the top of a building over 10 m above ground level. A variety of industrial activities (such as metal smelting and fertilizer production) are performed in this area.
The suburban site (QHU; N36°43′, E101° 44′; 2330 m above sea level) was at Qinghai University in the northern suburb of Xining, approximately 5 km away from the Science and Technology industrial district and 15 km north of FPH. The air sampler was on the roof of Department of Chemical Engineering building of Qinghai University, approximately 18 m above ground level. The nearest road, Haihu, was almost 50 m away. There was no large industrial source nearby.
The rural site (YLV; N36°42′, E101°31′; 2426 m above sea level) was at the village of Yula, which is northwest of Xining and about 30 km from FHP. This site is representative of rural areas, without any industrial emissions present. The air sampler was set on the roof of a farmer's house, 3 m above ground level.
PM2.5 was collected by medium-volume PM2.5 samplers (TH-150C China) at a flow rate of 100 L min −1 . All filters were 90 mm quartz (Whatman Inc., Maidstone, UK) and prebaked at 600 °C for 4.5 h before sampling. After being stabilized at 20 ± 1°C temperature and 30% ± 2% humidity, filters were weighed before and after sampling with an analytical scale (Mettler Toledo XP205DR, Zurich, Switzerland; precision: 0.01 mg). All filters were individually packed with aluminum foil, sealed in clean plastic bags, and stored at −18 °C until analysis. In total, 311 samples were collected from 7 January 2017 to 28 May 2018. Details are outlined in Table 1.

Chemical Analysis
Half of each filter was cut into small pieces and then ultrasonically extracted with 20 mL ultrapure Milli-Q water (18.2 MΩcm −1 ) for 40 min. Following filtration using a microporous membrane filter (pore size: 0.45 μm), all filtrates were stored at 4 °C in precleaned glass bottles until analysis. Five cations (Na + , NH4 + , K + , Mg 2+ , and Ca 2+ ) were analyzed using Dionex ICS5000 (Thermo Fisher Scientific, Waltham, USA). Five anions (F − , Cl − , NO3 − , C2O4 2− , and SO4 2− ) were analyzed using Dionex ICS1100 (Thermo Fisher Scientific, Waltham, USA). To efficiently separate the ions, a gradient The suburban site (QHU; N36 • 43 , E101 • 44 ; 2330 m above sea level) was at Qinghai University in the northern suburb of Xining, approximately 5 km away from the Science and Technology industrial district and 15 km north of FPH. The air sampler was on the roof of Department of Chemical Engineering building of Qinghai University, approximately 18 m above ground level. The nearest road, Haihu, was almost 50 m away. There was no large industrial source nearby.
The rural site (YLV; N36 • 42 , E101 • 31 ; 2426 m above sea level) was at the village of Yula, which is northwest of Xining and about 30 km from FHP. This site is representative of rural areas, without any industrial emissions present. The air sampler was set on the roof of a farmer's house, 3 m above ground level. PM 2.5 was collected by medium-volume PM 2.5 samplers (TH-150C China) at a flow rate of 100 L min −1 . All filters were 90 mm quartz (Whatman Inc., Maidstone, UK) and prebaked at 600 • C for 4.5 h before sampling. After being stabilized at 20 ± 1 • C temperature and 30% ± 2% humidity, filters were weighed before and after sampling with an analytical scale (Mettler Toledo XP205DR, Zurich, Switzerland; precision: 0.01 mg). All filters were individually packed with aluminum foil, sealed in clean plastic bags, and stored at −18 • C until analysis. In total, 311 samples were collected from 7 January 2017 to 28 May 2018. Details are outlined in Table 1.

Chemical Analysis
Half of each filter was cut into small pieces and then ultrasonically extracted with 20 mL ultrapure Milli-Q water (18.2 MΩcm −1 ) for 40 min. Following filtration using a microporous membrane filter (pore size: 0.45 µm), all filtrates were stored at 4 • C in precleaned glass bottles until analysis. Five cations (Na + , NH 4 + , K + , Mg 2+ , and Ca 2+ ) were analyzed using Dionex ICS5000 (Thermo Fisher Scientific, can be found in our previous study [28]. Surface analysis of PM 2.5 was performed with a ToF-SIMS V instrument (ION-ToF GmbH, Germany). A small punch (10 mm × 8 mm) of each filter was taken to match the sample holder as detailed in our previous study [29]. In this study, we obtained clear images of Na + , NH 4  were unclear because of a matrix effect, and these images are not shown.

Potential Source Contribution Function (PSCF) and Positive Matrix Factorization (PMF)
In this study, the potential source contribution function (PSCF) was applied to identify the potential source regions that contributed to the elevated PM 2.5 episodes. Positive matrix factorization (PMF) was used for source apportionment in this study. Uncertainties for individual species were c ij ≥ MDL j , calculated as (1); c ij ≤ MDL j , calculated as Equations (2) and (3).
where c ij , u ij , and s ij are the concentration, uncertainty, and analytical uncertainty of species j in the i-th sample, and MDL j is the method detection limit for species j [30,31]. This method has been previously described in detail in a previous report [32]. The input observable parameters included OC, EC, WSOC, and 7 ions (Na + , NH 4 + , K + , Ca 2+ , Cl − , NO 3 − , and SO 4 2− ). In this study, F − , Mg 2+ , and C 2 O 4 2− were excluded due to the low signal-to-noise (S/N) ratios that were acquired.

PM 2.5 and Chemical Compositions
The annual average concentration of PM 2.5 in Xining was 40.6 ± 34.6 µg·m −3 ; the highest concentration was 51.5 ± 48.9 µg·m −3 , observed at FPH, and the lowest in YLV, with an average of 26.4 ± 17.8 µg·m −3 . The average PM 2.5 concentration of GID and QHU was 42.8 ± 27.4 and 37.2 ± 23.7 µg·m −3 , respectively. All sites except YLV had concentrations above the ambient air quality standards of China (GB3095-2012). In this study, the determined PM 2.5 concentrations was lower than the recorded levels for provincial capitals in western China, including Xi'an [33], Lanzhou [34], and Chengdu [35], and much higher than those of studies in the Qinghai-Tibet Plateau [24,36,37].
In addition, the PM 2.5 concentration also shared similar seasonal variation across sampling sites as those of most studies, being highest in winter and lowest during summer. On the one hand, during winter, coal combustion heating systems can release large amounts of air pollutants. Lower temperature and wind speed in winter can also result in a lower mixing layer, which contributes to particle accumulation. The average PM 2.5 concentrations in spring were higher than those in autumn at the suburban and rural sites, and quite different from those at the other two sites (FPH and GID). These findings were the opposite of results obtained for Ningbo [38]. This can be ascribed to the following reasons: (1) in the past, straw was burned randomly in autumn, but burning straw in fields is now prohibited by government legislation; and (2) dust is prone to occurring in spring, and YLV and QHU are surrounded by bare fields, which makes them more likely to form dust.

OC/EC
At FPH, OC concentrations ranged from 1.6 to 24.6 µg·m −3 with an annual average of 9.3 ± 5.5 µg·m −3 , and EC concentrations ranged from 0.3 to 10.8 µg·m −3 with an annual average of 2.2 ± 2.0 µg·m −3 . At GID, OC ranged from 1.3 to 22.5 µg·m −3 (average: 6.2 ± 4.4 µg·m −3 ), and EC ranged from 0.2 to 5.9 µg·m −3 (average: 1.6 ± 1.2 µg·m −3 ). At QHU, annual average OC concentration was 6.3 ± 5.1 µg·m −3 , and average EC was 1.8 ± 1.0 µg·m −3 . At YLV, the annual average concentration of OC was 6.7 ± 5.0 µg·m −3 , whereas average EC was 2.1 ± 1.0 µg·m −3 . Generally, concentrations of OC and EC in Xining were lower than those in Beijing [39] (25.9 OC and 6.1 µg·m −3 EC) and Shanghai [40] (14.1 OC and 8.5 µg·m −3 EC in winter). OC was about four times higher than EC, especially at FPH (6.1 ± 6.0) and GID (7.4 ± 3.9). It is reported that, if OC/EC > 2, secondary organic carbon (SOC) may be formed [41][42][43]. SOC could be calculated using the following equations: where POC is primary organic carbon, and (OC/EC) min is the minimal OC/EC ratio excluding special data for the days when snow, rainstorms, and sandstorms caused drastic changes to the OC/EC ratio during the period of observation [42,44]. SOC concentration at the four sampling sites was highest in winter, which might be attributed to higher levels of gaseous precursor pollutants and the favorable oxidation conditions for secondary conversion in winter. The char-EC/soot-EC ratio is thought to be an indicator for identifying sources from biomass and coal burning or vehicle exhausts [45]. Char-EC is the major constituent of EC and is mainly derived from biomass and coal burning, while soot-EC usually results from vehicle emissions with higher-temperature combustion. According to previous studies, the char-EC/soot-EC of vehicle exhausts is about 0.60, while the char-EC/soot-EC of biomass and coal combustion is about 22.6 [46]. In this study, the average of char-EC/soot-EC shared similar seasonal variation across the sampling sites, with the highest in winter and lowest during summer, the highest average char-EC/soot-EC was 9.6 ± 5.5 at YLV and the lowest average char-EC/soot-EC was 1.3 ± 0.6 at GID in summer, annual average of char-EC/soot-EC at YLV and at GID were 5.1 ± 4.8, 2.8 ± 1.7 indicating that EC was mainly influenced by biomass and coal combustion at YLV, and by vehicle emissions at GID.

WSOC
The annual WSOC concentration in Xining was 4.0 ± 2.0 µg·m −3 , less than that in Beijing [47] and Xi'an, probably due to Xining's smaller population and lower plant coverage than compared to other cities. There was no significant difference in the annual WSOC concentrations at the four sampling sites: while the seasonal variation of WSOC at the four sampling points were significant statistically (ANOVA, p < 0.01). WSOC was highest in the winter and lowest in the summer across all four sites, which might be due to the higher intensity of biomass burning for heating and cooking in winter [48][49][50]. The annual mean WSOC/PM 2.5 ratio was 10.7% ± 5.0% across all four sites. The WSOC/PM 2.5 ratio was highest at YLV (14.2% ± 7.6%) and lowest at FPH (8.8% ± 5.3%). WSOC was more strongly correlated with PM 2.5 at all four sampling sites in spring, autumn, and winter than in summer ( Figure 2). The correlation coefficients of PM 2.5 and WSOC were higher in winter and lower in summer across the four sampling sites, which can be ascribed to the higher WSOC in winter being formed by the conversion of atmospheric pollutants. WSOC was an important part of PM 2.5 , especially at YLV, because biomass burning is the main source of heating at YLV. the higher concentrations of Ca 2+ and Mg 2+ in spring are mainly due to the frequent occurrence of sand and dust. The concentration of F − at Xining was higher than that at other Chinese cities [52,53], especially at GID, where the average concentration of F − was 1.5 ± 1.7 μg·m −3 . The high levels of F − measured in Xining might be associated with material manufacturing, such as electrolytic aluminum and phosphate fertilizer [53,54].
As shown in Figure 3, WSIIs were highest in winter and lowest in summer across all four sites, and seasonal variations were similar to those in our previous research in Shanghai [51]. In this study, the higher concentrations of Ca 2+ and Mg 2+ in spring are mainly due to the frequent occurrence of sand and dust. The concentration of F − at Xining was higher than that at other Chinese cities [52,53], especially at GID, where the average concentration of F − was 1.5 ± 1.7 µg·m −3 . The high levels of F − measured in Xining might be associated with material manufacturing, such as electrolytic aluminum and phosphate fertilizer [53,54].
Atmosphere 2020, 11, x FOR PEER REVIEW 7 of 16 PM2.5 mainly originates from external inputs, local emissions, and secondary conversions. The formation and transformation of secondary aerosols are usually characterized by the sulfur oxidation ratio (SOR) and nitrogen oxidation ratio (NOR) [55][56][57]. SOR and NOR can be calculated using the following equations: where n is the molar concentration. The range of SOR was 0.1-0.18 and NOR was 0.02-0.1 in Xining. In this study, the highest values for SOR and NOR were observed at FPH in summer and winter, while the lowest SOR and NOR was observed at YLV in autumn and summer, respectively. Results indicated that secondary conversion was more likely to occur at FPH, and the conversion of SO2 to SO4 2− always occurs in summer, while conversion of NO2 to NO3 − was more intensive in winter. High relative humidity and intensive solar radiation in summer could be attributed to the increasing concentrations of SO4 2− [58]. The mass ratio of NO3 − /SO4 2− was 0.67, suggesting that coal burning had greater impact on PM2.5 in Xining than vehicle exhausts, similar to the results in Hangzhou [59].
On the basis of ToF-SIMS, Ca + , K + , and Mg + were found to be mostly distributed as larger particles at lower concentrations of PM2.5 at FPH in Xining, while Ca + , K + , and Mg + were mostly distributed as smaller particles at higher concentrations of PM2.5 relative to the larger particles ( Figure  4); the profile was different from our previous research in Shanghai [60], indicating that the formation mechanism and composition of PM2.5 in the two cities were different. To explore the sources of heavy pollution, two samples that were obtained during a pollution episode were selected for depth profiles, The formation and transformation of secondary aerosols are usually characterized by the sulfur oxidation ratio (SOR) and nitrogen oxidation ratio (NOR) [55][56][57]. SOR and NOR can be calculated using the following equations: where n is the molar concentration. The range of SOR was 0.1-0.18 and NOR was 0.02-0.1 in Xining. In this study, the highest values for SOR and NOR were observed at FPH in summer and winter, while the lowest SOR and NOR was observed at YLV in autumn and summer, respectively. Results indicated that secondary conversion was more likely to occur at FPH, and the conversion of SO 2 to SO 4 2− always occurs in summer, while conversion of NO 2 to NO 3 − was more intensive in winter.
High relative humidity and intensive solar radiation in summer could be attributed to the increasing concentrations of SO 4 2− [58]. The mass ratio of NO 3 − /SO 4 2− was 0.67, suggesting that coal burning had greater impact on PM 2.5 in Xining than vehicle exhausts, similar to the results in Hangzhou [59].
On the basis of ToF-SIMS, Ca + , K + , and Mg + were found to be mostly distributed as larger particles at lower concentrations of PM 2.5 at FPH in Xining, while Ca + , K + , and Mg + were mostly distributed as smaller particles at higher concentrations of PM 2.5 relative to the larger particles ( Figure 4); the profile was different from our previous research in Shanghai [60], indicating that the formation mechanism and composition of PM 2.5 in the two cities were different. To explore the sources of heavy pollution, two samples that were obtained during a pollution episode were selected for depth profiles, which showed that K + and the Na + signals were stronger than those of other ions ( Figure 5). The presence of the two ions gradually decreased going from the surface to the interior particles, with Na + decreasing more steeply than K + . NH 4 + was more evenly distributed on the surface and interior of particles. Other cations were distributed on the particle surface. SO 4 − was evenly distributed going from the surface to the interior of particles. F − increased going from the surface to the interior of particles, further research is needed to investigate the mechanism.

PM2.5 Source Apportionment in Xining
To exactly identify the spatial distribution of potential sources, the PSCF method was utilized on the basis of the results of backward-trajectory analysis of 72 h air masses. For each day, four trajectories (local time: 2:00, 8:00, 14:00, and 20:00) were employed with an interval of six hours. As shown in Figure 6, most potential source areas with higher PSCF values for PM2.5 were located west of Xining, including the Qaidam basin in Qinghai province, the Tarim basin in Xinjiang province, and the Pamirs plateau, where the air mass passed over the desert. Through mid-and large-scale transportation, the air mass that passed over this area made a large contribution to PM2.5 concentrations in Xining, the prevailing wind was from west during the sampling period in Xining.
In this study, PMF analysis was conducted to identity emission sources of PM2.5 in Xining. PMF input was the dataset (the concentrations of chemical compositions mentioned above) of the PM2.5 samples. Samples collected during heavy pollution events, such as sandstorms and the Lunar New year, were excluded. Five to eight factors were tested, and the source profile of the seven-factor solution was the most reasonable. The seven-factor solution was verified to be stable by performing 100 bootstrap runs, as 85% of the runs produced the same factors.

PM 2.5 Source Apportionment in Xining
To exactly identify the spatial distribution of potential sources, the PSCF method was utilized on the basis of the results of backward-trajectory analysis of 72 h air masses. For each day, four trajectories (local time: 2:00, 8:00, 14:00, and 20:00) were employed with an interval of six hours. As shown in Figure 6, most potential source areas with higher PSCF values for PM 2.5 were located west of Xining, including the Qaidam basin in Qinghai province, the Tarim basin in Xinjiang province, and the Pamirs plateau, where the air mass passed over the desert. Through mid-and large-scale transportation, the air mass that passed over this area made a large contribution to PM 2.5 concentrations in Xining, the prevailing wind was from west during the sampling period in Xining.
In this study, PMF analysis was conducted to identity emission sources of PM 2.5 in Xining. PMF input was the dataset (the concentrations of chemical compositions mentioned above) of the PM 2.5 samples. Samples collected during heavy pollution events, such as sandstorms and the Lunar New year, were excluded. Five to eight factors were tested, and the source profile of the seven-factor solution was the most reasonable. The seven-factor solution was verified to be stable by performing 100 bootstrap runs, as 85% of the runs produced the same factors.

Influence of Sandstorms on PM 2.5
During spring and winter, the frequent occurrence of sand and dust contributed to a higher concentration of PM 2.5 in Xining. As mentioned above, the dust mainly was from west, FPH located at the most east of Xining. Thus, FPH is the ideal site for investigating the effect of dust in Xining. During the sampling period, there were 10 severely polluted episode affected by sandstorms, particle samples during severely polluted episodes at FPH were chosen to analyze the characteristics of particles. During these sandstorms, the average concentrations of PM 2.5 and PM 10 sharply increased, and the concentration of PM 2.5 reached 1.42-10.21 times higher than the annual average values (Figure 8). The concentrations of Na + , Ca 2+ , and Mg 2+ were 1.13-2.70, 1.68-4.41, and 1.15-5.12 times higher than the average concentration, respectively. Na + mainly comes from marine and industrial emissions, but Xining is an inland city on a plateau; thus, Na + was likely to come from salt lakes and saltwater lakes such as Qinghai Lake, which is the largest saltwater lake. This is also the reason for the higher concentration of Mg 2+ . The surface profiles of PM 2.5 collected at FPH during a sandstorm and a clear day were determined by ToF-SIMS. Particulate matter in the sandstorm greatly increased and almost covered the entire filter area. The silicon fiber on the clean day was very clear, and there were no significant changes in the NH 4 + , Ca + , and Mg + images. K + and Na + in the images were more obvious on the clear day. The difference of the negative-and positive-ion spectrum between sandstorm and clear days in Xining was not as obvious as the comparison between haze and clear days in Shanghai [29]. TOF-SIMS was not suitable for analyzing filters polluted by sandstorms, which might be related to the stronger matrix effect caused by the nonconductivity of a large amount of dust.

Heavy-Pollution Episode
A severe pollution episode with an AQI of 500 occurred on 26-27 January 2017, where the concentration of PM2.5 exceeded 200 μg·m −3 . PM2.5 concentration peaked at 406.77 μg·m −3 on 26 January, which was much higher than the annual average concentration of PM2.5. Concentrations of Ca 2+ and Mg 2+ increased more sharply than that of Na + on 26 January. Ca 2+ and Mg 2+ are usually thought to be the indicator of dust and sand; thus, the higher levels of PM2.5 on 26 January were related to the sandstorm (verified by the Qinghai Meteorological Bureau). Additionally, K + on 27 January was almost 20 times higher than the annual average value, which is mainly because 27 January was the traditional Chinese New Year's Eve, where a large number of fireworks and firecrackers were burned. On 28 January, high wind speed was helpful for the dilution of air pollutants, so the concentration of PM2.5 decreased. Backward trajectory analysis also demonstrated that the air pollution process was affected by dust from the sandstorm that occurred in the Hexi corridor on 25 January, and the stable weather conditions in Xining contributed to the accumulation of atmospheric pollutants.

Heavy-Pollution Episode
A severe pollution episode with an AQI of 500 occurred on 26-27 January 2017, where the concentration of PM 2.5 exceeded 200 µg·m −3 . PM 2.5 concentration peaked at 406.77 µg·m −3 on 26 January, which was much higher than the annual average concentration of PM 2.5 . Concentrations of Ca 2+ and Mg 2+ increased more sharply than that of Na + on 26 January. Ca 2+ and Mg 2+ are usually thought to be the indicator of dust and sand; thus, the higher levels of PM 2.5 on 26 January were related to the sandstorm (verified by the Qinghai Meteorological Bureau). Additionally, K + on 27 January was almost 20 times higher than the annual average value, which is mainly because 27 January was the traditional Chinese New Year's Eve, where a large number of fireworks and firecrackers were burned. On 28 January, high wind speed was helpful for the dilution of air pollutants, so the concentration of PM 2.5 decreased. Backward trajectory analysis also demonstrated that the air pollution process was affected by dust from the sandstorm that occurred in the Hexi corridor on 25 January, and the stable weather conditions in Xining contributed to the accumulation of atmospheric pollutants.

Conclusions
To comprehensively study the characteristics of WSIIs and organic compositions in addition to the potential sources of PM 2.5 , an over-one-year field measurement study was conducted from January 2017 to May 2018 at four sites in Xining, China: an urban, industrial, suburban, and rural site. The annual mean of PM 2.5 was the highest at the urban site and lowest at the rural site, with an average of 51.5 ± 48.9 and 26.4 ± 17.8 µg·m −3 , respectively. The average PM 2.5 concentration of the industrial and suburban sites was 42.8 ± 27.4 and 37.2 ± 23.7 µg·m −3 , respectively. All sites except the rural had concentrations above the ambient air quality standards of China (GB3095-2012). The highest concentration of PM 2.5 was observed in winter, followed by spring, autumn, and summer at all sites. The concentration of major constituents of PM 2.5 showed statistically significant seasonal and spatial variation; the highest concentrations of OC, EC, WSOC, and WSIIs were found in winter at the urban site. The average concentration of F − was higher than that of many studies in other cities of China, especially at the industrial site, where the annual average concentration of F − was 1.5 ± 1.7 µg·m −3 . PSCF results showed that air mass from mid-and large-scale transportation from the western areas of Xining contributed to the higher level of PM 2.5 . On the basis of PMF, aerosols from salt lakes and dust were found to be important sources of PM 2.5 in Xining, accounting for 26.3% of aerosol total mass. During the sandstorms, the concentration of PM 2.5 increased sharply, and concentrations of Na + , Ca 2+ , and Mg 2+ were 1.13-2.70, 1.68-4.41, and 1.15-5.12 times higher than the annual average concentration, respectively, and were mainly derived from dust and the largest saltwater lake, Qinghai Lake, and many other salt lakes in Qinghai. The surface profiles of PM 2.5 showed F − was increasingly distributed from surface to inside of the particles. During the heavy pollution episode, the combination of sandstorms and burning of fireworks contributed to the occurrence of severe pollution and extremely high concentration of PM 2.5 .