Influences of a Quasi-stationary Front on Particulate Matter in the Low-latitude Plateau Region in China

By utilizing observation and reanalysis data and statistical methods to comprehensively assess the main characteristics of particulate matter (PM) on the low-latitude plateau (LoLaP) of southwestern China, this study analyzed the influence of the Kunming quasi-stationary front (KMQSF) on the PM concentration during winter in this region. We found that the location and intensity of the KMQSF significantly affected the distribution pattern of the PM and induced high concentrations in some areas during a typical pollution event in 2016. Furthermore, we investigated all of the KMQSF synoptic weather that occurred from 2014 till 2019 and categorized it into nine types, which produced different pollution patterns. When the KMQSF moved westward from the center of the LoLaP, the easterly wind transported PM to the western part of the region, which led to an increase in the mean daily PM10 and PM2.5 concentrations in Kunming, a capital city in that area; simultaneously, however, the PM10 and PM2.5 levels remained high in Guiyang, a capital city in the eastern part of the LoLaP. When the KMQSF gradually retreated from the western to the eastern part of the LoLaP, the PM in Kunming dropped as the westerly wind increased in strength, whereas the PM in Guiyang continued to rise as the planetary boundary layer (PBL) height (PBLH) and boundary layer dissipation (BLD) rate decreased. When the KMQSF disappeared, temperature inversion, a shallow PBL and weak BLD across the entire region caused pollutants to accumulate, resulting in the highest mean daily PM10 and PM2.5 concentrations for both cities.


INTRODUCTION
Air pollution has been a global issue over the past several decades (Ding et al., 2013;Duncan et al., 2016;Bian et al., 2017;Davis, 2017) due to its harmful influences on human health (Lelieveld et al., 2015;Raaschou-Nielsen et al., 2017) and ecosystems as well as its severe impacts on climate (Knippertz et al., 2015;Acosta Navarro et al., 2016;Li et al., 2016;Lau et al., 2017) and synoptic systems (Gong et al., 2014;Fan et al., 2015). In addition to emissions, meteorological factors, such as wind, relative humidity (RH), precipitation, turbulent dispersion, the planetary boundary layer (PBL), vertical temperature structure and synoptic systems, have been found to play very important roles in the day-to-day variation and the long-term trend in air quality (Ding et al., 2009;Thishan Dharshana et al., 2010;Özbay, 2012;Appelhans et al., 2013;Quan et al., 2013;Wolf et al., 2014;Miao et al., 2015;Hou and Wu, 2016;Jiang et al., 2017;Liang et al., 2017;Tie et al., 2017). Therefore, understanding the influences of meteorological factors and atmospheric circulation on air quality is important both for improving air quality

Meteorological Data
The hourly air temperature, RH, wind speed, wind direction and precipitation data that were observed in Kunming and Guiyang were used in this study. The two additional stations are managed by the Kunming Meteorological Bureau; the first site is approximately 1 km southeast of the YNU Donglu station (102.71°E, 25.06°N), and the second site is the standard meteorological station (102.65°E, 25°N), which is located approximately 6 km southwest of the YNU Donglu station.
In addition, ERA-Interim reanalysis data four times day by day on November-next February from 2014 to 2019 between 97-110°E, 21-30°N at a resolution of 0.125° × 0.125° were used in this research and included the wind speed at a height of 10 m above the surface, temperature at a height of 2 m, low cloud cover, PBL height (PBLH) and boundary layer dissipation (BLD), wind speed, temperature and RH in the upper 37 levels above 1000 hPa (Dee et al., 2011;Tavolato and Isaksen, 2011).

Meteorological Diagnostic Method
The scalar frontogenesis function was introduced to display the development and change in the KMQSF (Zhu et al., 2000), and this function is presented in Eq. (1): where |∇θ| can be expanded using Eq. (2): Eq.
(1) includes the individual effects on the scalar frontogenesis function from the adiabatic heating term, vertical motion tilt term, horizontal divergence term and horizontal deformation term. In addition, F > 0 indicates frontogenesis, and F < 0 indicates frontolysis. The scalar frontogenesis function is used to confirm the location of the KMQSF. The usual criterion of the KMQSF can be represented by the isotherm intensive area and the west-east wind transition zone, which is associated with the ground region |∇θ| > 5 K 100-km −1 , the scalar frontogenesis function F > 0 (Duan et al., 2017).
The bulk Richardson number, Rb, is defined as the ratio of turbulence associated with buoyancy to that induced by mechanical shear (Stull, 1988), which is expressed as: where z denotes height above ground, g is the acceleration of gravity, θ is the potential temperature, u and v are the component of wind speed. In the limit of layer thickness becoming small, the bulk Richardson number approaches the gradient Richardson number, for which a critical Richardson number, Ric, is roughly 0.25. Gradient Richardson numbers less than this critical value are dynamically unstable and likely to become or remain turbulent.

General State of this Event
The observed average PM10, PM2.5 and PM1 mass concentration at the Donglu station located in southeast Kunming were 62.3, 34.4 and 28.0 µg m −3 , respectively, from November 2016 to May 2017. However, the daily mean concentration of PM2.5 at Donglu on December 9, 2016, was 89.6 µg m −3 , which was higher than 75 µg m −3 exceeding the Grade II standard in China and was the only pollution event in the research period. In addition, PM10 and PM2.5 concentrations synchronously increased during December 4-9, 2016, in the whole LoLaP (Fig. 2), while the highpollution event usually appeared in eastern LoLaP. A typical pollution event was selected for deeper analysis to reveal its characteristics and causes.
The continuously accumulated of PM in the typical pollution event was between two KMQSF processes on December 2-3 and after December 11, respectively (Fig. 3). The distribution of PM concentrations showed the similar structure to the KMQSF, and the isolines of PM concentrations was northwest-southeast. A discontinuity in wind direction and a significant horizontal potential temperature gradient could be used to diagnose KMQSF, and the pre-frontal and post-frontal wind speed decided the location of KMQSF, which usually caused the different weather in west and east parts of LoLaP. The KMQSF marked by a significant horizontal potential temperature gradient (|∇θ| > 5 K 100-km −1 ) and wind shear. When the KMQSF appeared and the westerly was stronger, the low values of PM in western LoLaP will be push to eastern LoLaP by high wind speed. When the KMQSF disappeared and this condition maintained several days, the westerly and the easterly weakened (Fig. 3), the PM accumulated.
In general, the PM concentrations were higher in eastern LoLaP and lower in western LoLaP (Fig. 2). On December 1-3, the low-values area of PM in western LoLaP gradually expanded eastward (Figs. 2(a)-2(c)), when the westerly gradually strengthened in western LoLaP and the KMQSF appeared in the middle of LoLaP (Figs. 3(a)-3(c)). From December 4-9, the PM in the whole LoLaP increased (Figs. 2(d)-2(j)), and the daily mean concentration of PM2.5 reached 80 µg m −3 near Kunming and 120 µg m −3 to the northeastern LoLaP on December 8. The wind speed was always lower in whole LoLaP in this period and the KMQSF disappeared in the middle of LoLaP (Figs. 3(d)-3(j)). After December 10, the low-values area of PM in western LoLaP expanded eastward again, when the westerly strengthened. A PM area with PM2.5 of 120 µg m −3 appeared on December 11 and 12, which was trapped in eastern LoLaP (Figs. 3(k)-3(l)).
In general, the highest PM concentration mainly occurred at night, associating with a stable nocturnal boundary layer. The lower PBLH, smaller BLD and Rb > 0.25 could represent a stable boundary layer. Fig. 4 shows distribution of PBLH, BLD and Rb at 18:00 UTC (02:00 BJT) from December 1 to 12, 2016. When the KMQSF appeared in the middle of LoLaP before December 3, the PBLH was higher, the BLD was stronger, dynamically unstable area was larger (Figs. 4(a)-4(c)), and the PM concentration was lower. During night from December 4 to 9, except for December 5, the stable PBLH were lower than 50 m and the BLD were lower (Figs. 4(d)-4(i)); when the PM reached the highest concentrations both in Kunming and Guiyang on December 8, the BLD were lower than 3 × 10 4 J m −2 (Fig. 4(h)). Obviously, local pollution accumulation under the stable PBL and weak dissipation was the main reason inducing this pollution episode. When the KMQSF reappeared in the middle of LoLaP after December 11, the PBLH and the BLD increased both (Figs. 4(k)-4(l)), the low PM area in western LoLaP was pushed to eastern LoLaP. The high-PM Fig. 3. The distribution of horizontal potential temperature gradient (K 100-km −1 ; shaded, horizontal bar), wind (m s −1 ) and potential temperature (K; vertical bar) at the surface in LoLaP during the period of (a-l) December 1-12, 2016. pollution was obviously blocked to the east of the frontal line of the KMQSF. The change of KMQSF was well matched with the change of PM concentrations, and the KMQSF played an important role in this pollution episode.

Temporal Evolution of PM Concentration in Western and Eastern LoLaP
In order to analyze the causes of this pollution event in detail, the temporal changes in PM10, PM2.5 and PM1 at Donglu in Kunming and the temporal changes in PM10 and PM2.5 in Guiyang, as well as the meteorological parameters (Fig. 5), are selected for further analysis, and a comparison between the western LoLaP site and the eastern LoLaP site can be made. The concentrations of PM10, PM2.5 and PM1 at Donglu and Guiyang during December 1-3 were lower (Figs. 5(d) and 5(h)). Under the control of the KMQSF, the water vapor in the west side of the stationary front is sufficient, which caused the low cloud (Fig. 6). The atmosphere was cleaner after the frontal precipitation process during December 1-2 at Donglu and during December 2-3 in Guiyang, and this state was the pre-pollution period of this episode.
The period of December 4-9 was the pollution accumulation phase of the pollution episode. The hourly maximum concentrations of PM10, PM2.5 and PM1 on December 4 were 148.7, 86.8 and 69.5 µg m −3 , respectively, and they increased to 228.0, 133.1 and 110.2 µg m −3 on December 9 at Donglu (Fig. 5(d)). The hourly maximum concentrations of PM10 and PM2.5 reached 273.9 and 188.9 µg m −3 on December 8 in Guiyang (Fig. 5(h)). The growth rates of hourly minimum PM10, PM2.5 and PM1 at Donglu changed from 7.9 µg m −3 24-h −1 on December 4 to 25.8 µg m −3 24-h −1 on December 7, from 4.0 µg m −3 24-h −1 to 19.0 µg m −3 24-h −1 , and from 3.2 µg m −3 24-h −1 to 15.6 µg m −3 24-h −1 , respectively. The growth rates of hourly minimum PM10 and PM2.5 in Guiyang During the pollution accumulation phase, wind speed remained at lower levels at Donglu, with an hourly maximum speed of 1.3 m s −1 before December 8, and it is southerly in daytime and northerly at night ( Fig. 5(a)). The nocturnal wind from the north of the Donglu station might contain a lot of pollution gas from the Jindinshan industrial area (Shi et al., 2016), which caused the hourly PM maximum at night at Donglu. The mean wind speed in Guiyang was larger than that at Donglu, several changes of wind direction happened from December 4 to 10 with relative lower wind speed (Fig. 6(e)), the air from surrounding region could be transported to Guiyang. RH was 81.3% at night on December 6 at Donglu, and then RH continually increased during both day and night until reaching 96.0% at night on December 9 ( Fig. 5(b)). The maximum RH at night on December 8 and 9 also reached 98% and 97% without rain in Guiyang (Fig. 5(f)). Under high RH (i.e., > 60-80%), the volume of aerosol particles can significantly increase when water vapor is absorbed onto the surface of these particles, and the enlarged aerosol surfaces/volumes lead to more rapid multiphase reactions and secondary aerosol formation, which results in elevated aerosol concentrations (Tie et al., 2017). In addition, the ground inversion layer (TIL) always existed (Figs. 5(c) and 5(h)). The low wind speed, high RH, and TIL induced stagnant conditions, which corresponded to the disappearance of the KMQSF and favored the accumulation of aerosol particles and secondary aerosol formation.
The RH was still high on December 10 at Donglu (Fig. 5(b)), and a stable TIL maintained (Fig. 5(c)). These conditions caused high concentrations of fine particles until midnight, and then the particle mass pollutant was dispersed by higher wind speeds and stronger turbulence on  December 11 and 12 (Figs. 5(a) and 5(d)). The RH abruptly decreased, which meant a weather change and a warm and dry airflow immigration in western LoLaP. The TIL disappeared in Guiyang on December 10 ( Fig. 5(g)), wind speed and RH gradually increased on December 11 and 12 (Figs. 5(e) and 5(f)), the southerly changed into easterly and wind speed increased on December 12 ( Fig. 5(e)), and the low cloud moved westward from eastern LoLaP (Fig. 6(c)), which meant a cold and moist airflow appeared in eastern LoLaP. The PM in Guiyang also decreased in higher wind speeds but was still higher than that at Donglu in influence of pollution transport from surrounding region. This final period of the pollution episode was called the collapsing period, which corresponded to the appearance of the KMQSF (Fig. 6). The wind speed remained at a relatively high level during December 13-14, with hourly maximum speeds of 1.8 and 2.7 m s −1 at Donglu and hourly maximum speeds of 3.9 and 5.5 m s −1 in Guiyang (Figs. 5(a) and 5(e)). This change was associated with a decrease in the maximum of nocturnal RH from 85.9% to 81.1% at Donglu (Fig. 5(b)), and with an increase in the maximum of nocturnal RH from 91% to 99% in Guiyang (Fig. 5(f)). These facts indicate that warm and dry westerly winds in western LoLaP and cold and moist easterly winds in eastern LoLaP had strengthened. The KMQSF moved westward with low clouds (Fig. 6), and rainfall occurred in Guiyang and removed PM (Figs. 5(f) and 5(h)).
Temperature inversion can restrain the convection and dispersion of air pollutants due to weaker turbulence, and temperature inversion can also lead to a weaker surface wind speed relative to neutral and unstable stratification according to the logarithmic velocity profile. These conditions can also further decrease the transport of air pollutants. Therefore, temperature inversion favors the accumulation of air pollutants. From December 4-12, there was a ground inversion layer in Kunming at 08:00 and 20:00, and this inversion layer was usually thinner and stronger at 20:00 than at 08:00 (Fig. 5(c)), which was produced at night and disappeared in  daytime. The surface cooling process at night might be caused by the strong upward infrared radiation and the cold air sliding down the hillside during nocturnal hours. These results support the formation of a ground temperature inversion (Duan et al., 2014;Wolf et al., 2014). The rate of the TIL appearing in Kunming is obviously higher than that in Guiyang, because Kunming is located in a small basin surrounded by mountains and along the northwestern edge of a lake with the valley wind, warm advection, and infrared radiation emission of the surface (Duan et al., 2014). The temporal characteristics of TIL indicated that there was absolute stability near the ground during this pollution episode, in particular from December 5 to 10, which supplied enough time for the accumulation of air pollutants, production of chemical reactions among pollutant species, and formation of secondary aerosols. The thickness of the inversion layer reached 1024 m at 08:00 on December 9, which matched the timing of the subsequent peak value of aerosols. The thicker inversion layer was not easy to disappear immediately, so the hourly maximum concentrations of PM appeared at 11:00 before it started to rain. The TIL appeared on December 4-9 and there was a ground inversion layer at 08:00 and 20:00 on December 8 in Guiyang (Fig. 5(g)), when clear sky also appeared in Guiyang (Fig. 6(c)). The similar calm weather supported the pollution event appeared in Kunming and Guiyang both. Otherwise, there were some different between western and eastern LoLaP, the stability in western LoLaP was obviously higher than that in eastern LoLaP, which shown lower wind speed and long-term maintain of TIL. Fig. 7 shows a backward trajectory analysis. Using the Hybrid Single-Particle Lagrangian Integrated Trajectory (HYSPLIT) model driven with the Global Data Assimilation System (GDAS) output, the 48-h backward trajectories of Kunming and Guiyang from December 4-10, 2016, were analyzed. The air over Kunming was mainly sourced from cleaner southwest direction during the study period, except for December 6 and 7. During the pollution accumulation period from December 4-9, the wind speed decreased gradually and slowly changed to ground-toground airflow, and the wind nearly stopped on December 9. Although the wind speed increased on December 10, it remained a ground-to-ground airflow. There were more complicated air sources in Guiyang. Except for December 6 and 7, the air over Guiyang in 48 h was mainly from the north and south direction, and the wind also nearly stopped on December 8 and 9. On December 6, the air over Guiyang in 48 h was from Inner Mongolia and turned to northeast wind near 30°N, and it was easterly in east of Guiyang ( Fig. 3(f)). The path on December 7 was similar to that on December 6, but the wind speed decreased and it turned to southerly before reaching Guiyang. Thus, it is reasonable to assume that the source of this pollution event in Kunming was local, and there were pollution sources from surrounding areas in Guiyang, except for local source.
This typical pollution episode was triggered by the combined effects of synoptic conditions, decreasing wind speed, increasing RH, and ground temperature inversion. During the sustained synoptic process, fine particles could continuously accumulate, and secondary aerosols could continuously form and wet-grow into larger aerosols; therefore, a light haze episode formed. When the synoptic conditions changed again, e.g., when the KMQSF migrated, wind disrupted the static and stable weather, and the haze dissipated.

KMQSF PATTERNS AND INFLUENCES ON AIR POLLUTION IN RECENT FIVE YEARS
The KMQSF often appears in the eastern region of LoLaP during winter and spring, and it seems that the location and intensity of KMQSF influenced the concentrations and distribution of PM from Section 3. The meteorological factors and PM data on November-next February 2014-2019 were used to analyze the relationship between them; there are 601 days in total. The selforganizing maps (SOMs) were used to distinguish the patterns of KMQSF (Mattingly et al., 2016;Gibson et al., 2017;Francis et al., 2018), and the horizontal potential temperature gradient and zonal wind were used as input data of SOM. Three different SOM configurations were tested: a 6-node SOM (3 columns × 2 rows), a 9-node SOM (3 columns × 3 rows) and a 12-node SOM (3 columns × 4 rows), applied to standardized horizontal potential temperature gradient data and zonal wind speed data. From these options, the output from the 9-node SOM is easier to interpret than the 6-node and 12-node SOM, and it also includes a sufficient number of nodes to represent the KMQSF; therefore, the 9-node SOM was chosen as the best classification for further analysis (Table 1).
The horizontal potential temperature gradient was matched with the transformation zone of westerly and easterly. When the easterly was strong, the KMQSF would be pushed to west, and when the westerly was strong, the KMQSF would be pushed back to east. The whole eastern LoLaP and middle LoLaP were controlled by the strong and cold northeast wind under Pattern 1 ( Fig. 8(a)). There were 246 days in Pattern 2-5, in which the KMQSF located between western and eastern LoLaP, and eastern LoLaP prevailed cold northeasterly and western LoLaP was covered by warm southwest wind (Figs. 8(b)-8(e)). The southwesterly was very strong under Pattern 6 and 7, southerly occupied the whole eastern LoLaP and the KMQSF was pushed to the northern LoLaP (Figs. 8(f) and 5(g)). Pattern 8 includes 82 days, the KMQSF located middle LoLaP, but the KMQSF was weaker, associating with southerly in eastern LoLaP (Fig. 8(h)). Fig. 9 shows distribution of PBLH, BLD and Rb at 18:00 UTC (02:00 BJT) and Fig. 10 shows distribution of daily mean PM10 and PM2.5. In general, the PBLH was higher and BLD was stronger near the KMQSF, and the high-values areas of the PBLH and the BLD shown northwest-southeast distribution. When the southwest wind strengthened, the high-values areas of the BLD in eastern LoLaP would move to north; atmospheric stability was bounded by the KMQSF, and the surface layer was dynamically stable (unstable) on the west (east) side of the KMQSF; the PM gradually increased from west to east, which was lower in westerly area and higher in easterly area. When  the KMQSF moved to the region west to Kunming or near Kunming represented by Pattern 1 and 2 (Figs. 9(a) and 9(b)), the high PM area was obviously pushed toward western LoLaP (Figs. 8(a) and 8(b)) with daily mean PM10 and PM2.5 of 55.8 ± 20.9 and 35.0 ± 13.7 µg m −3 in Kunming (Table 1). When both the southwest wind and northeast wind was the strongest, represented by Pattern 3 and 4, the BLD between western and eastern LoLaP was higher than 15 × 10 4 J m −2 , associated by a PBLH higher than 300 m in the region east to the frontal line of the KMQSF (Figs. 9(c) and 9(d)), the PM was dissipated in whole LoLaP, with low values of PM covering LoLaP from southwest to northeast (Figs. 10(c) and 10(d)), in which the daily mean PM10 and PM2.5 concentrations were 44.6 ± 19.5 and 28.6 ± 12.3 µg m −3 in Guiyang in Pattern 3, and 51.4 ± 15.1 and 26.0 ± 8.7 µg m −3 in Kunming in Pattern 4 (Table 1). When the southwesterly gradually strengthened, represented by Pattern 6 to 7, the BLD in LoLaP increased from 9 × 10 4 J m −2 to 21 × 10 4 J m −2 , associated by an increase of the PBLH from 100 m to 200 m (Figs. 9(f) and 9(h)), the daily mean PM2.5 concentrations decreased from 28.3 ± 10.9 to 23.6 ± 7.4 µg m −3 in Kunming and decreased from 46.7 ± 25.4 to 38.4 ± 16.0 µg m −3 in Guiyang. Pattern 8 was similarly with Pattern 9 with much stronger wind and lower PM. There were 72 days in Pattern 9, in which the KMQSF disappeared between western and eastern LoLaP, associating with lower southerly in LoLaP (Fig. 8(i)). The horizontal wind shear still existed in eastern LoLaP, but the horizontal potential temperature gradient was very low. The PM in LoLaP reached the highest value in Pattern 9 ( Fig. 10(i)), associating with the lowest PBLH below 100 m and the weakest BLD less than 6 × 10 4 J m −2 , and almost the whole region was dynamically stable (Fig. 9(i)), in which the difference of the PBLH and the BLD between western and eastern LoLaP was ignorable. The PM10 and PM2.5 significantly accumulated under such stable boundary layer to 72.3 ± 26.6 and 39.6 ± 15.4 µg m −3 in Kunming, and 96.1 ± 34.1 and 60.7 ± 22.1 µg m −3 in Guiyang, respectively (Table 1). Furthermore, Pattern 9 showed that it was westerly between Kunming and Guiyang on average; in fact, there were 9 days when it was weak easterly and 63 days when it was weak westerly. The daily mean PM10 and PM2.5 concentrations were 67.6 ± 31.1 and 43.6 ± 22.7 µg m −3 in Kunming with weak easterly. The daily mean PM10 and PM2.5 concentrations were 72.9 ± 26.2 and 39.1 ± 14.3 µg m −3 in Kunming with weak westerly.  In general, when the KMQSF reached Kunming, the pollution could be transported to western LoLaP from eastern LoLaP by easterly under unstable atmosphere. When the KMQSF was weaker or disappeared between western and eastern LoLaP, the wind was lower in LoLaP, associating with a low PBLH and a weak BLD; these facts resulted in stable atmosphere near the surface, which restrained pollutant dispersion and led to high PM values. When PM in eastern LoLaP was higher than in western LoLaP indicated by Pattern 8 and 9, the pollutant could not be transported by the southwest wind from eastern LoLaP to western LoLaP; therefore, the local accumulation should be the important reason inducing high PM concentration in Kunming. In general, the pollution in the Pattern 8 and 9 was induced mainly by local sources in the western LoLaP and there was pollution transport in the eastern LoLaP, there was long-range transport from east to west in Pattern 1 and 2, there was transport from west to east in Pattern 6 and 7, the air converged near the KMQSF in other patterns, and the KMQSF could block airflow between western and eastern LoLaP. The pollution event on December 2016 exactly occurred in meteorological condition of Pattern 9 and was ended in Pattern 6. Pattern 9 appeared 20 days in November 2016 to February 2017, and the frequency was higher than other years. Furthermore, the PBLH was always lower in the west side of the KMQSF, and this fact indicated more stable air near the surface in western LoLaP than eastern LoLaP, but the PM in eastern LoLaP was always higher than in western LoLaP, which indicated the local pollution emission and pollution transport in eastern LoLaP should be stronger than in western LoLaP. In Pattern 8, the PM10 and PM2.5 concentrations in Guiyang was 23.7 and 21.1 µg m −3 higher than those in Kunming, and in Pattern 9, the PM10 and PM2.5 concentrations in Guiyang was 23.8 and 21.1 µg m −3 higher than those in Kunming.
In addition, a pollution event lasting for several days might include different KMQSF patterns. We find an example from December 24 to 27, 2017, with the daily mean PM2.5 concentrations in Kunming of 89.7, 103.4, 80.3 and 81.7 µg m −3 , respectively experienced temporal changes of Pattern 9, Pattern 8, and Pattern 9, respectively. Though this event also went through the weaker and even disappeared KMQSF with lower PBLH and weaker BLD, the atmospheric circulation was different from the event on December 4-10, 2016. A high pressure appeared in the middle north of the LoLaP on December 26, 2017; the sinking restrained the pollution dispersion and promoted this pollution event. The pollution source and meteorological condition might be different in different pollution event, which should be particularly analyzed through the whole continuous process.

CONCLUSIONS
Analyzing the PM10, PM2.5 and PM1 concentrations for November 2016-May 2017 in downtown Kunming (Donglu), we identified the causes of a pollution episode during 2016 and evaluated the relationship between the KMQSF and the fine particle concentrations on the LoLaP between November and February for the years 2014-2019. Our main conclusions can be summarized as follows: 1) The change in weather triggered the regional pollution episode that affected Kunming and Guiyang on December 8 and 9, 2016, during which the PM10 and PM2.5 concentrations synchronously increased across the entire LoLaP. Prior to this episode, both Kunming and Guiyang had experienced a brief precipitation event and low aerosol concentrations. Additionally, from December 4 till 9, the KMQSF weakened to the east of Kunming, and both Kunming and Guiyang exhibited low wind speed, low PBLH, weak BLD, high RH and ground temperature inversions. The meteorological conditions favored the accumulation of primary PM and secondary organic aerosol precursors as well as the formation and growth of fine secondary particles, which resulted in a light haze. Once the KMQSF intensified, the static and stable weather was interrupted by a strong westerly wind, and the haze dissipated. 2) From November till February between 2014 and 2019, the weakest westerly and easterly winds, lowest PBLH and weakest BLD occurred when the KMQSF was very weak and disappeared in the center of the LoLaP. This weather pattern was also associated with a dynamically stable atmosphere near the surface at night and maximum PM concentrations due to unfavorable boundary layer dispersion conditions. When the KMQSF approached Kunming from the eastern LoLaP, easterly winds transported pollutants to the western part of the region beneath an unstable atmosphere. When the KMQSF was more intense, both the easterly and the westerly winds were stronger, leading to low PM concentrations across the LoLaP. Identifying nine KMQSF synoptic weather patterns, we discovered that the pollution observed during Patterns 8 and 9 was mainly contributed by local emission on the western LoLaP and there was pollution transport from surrounding areas on the eastern LoLaP. We also observed long-range transport from the east to the west during Patterns 1 and 2 and in the opposite direction during Patterns 6 and 7. During the other weather patterns, the air masses converged near the KMQSF, which blocked airflow between the western and eastern parts of the LoLaP. Furthermore, the consistently low PBLH to the west of the KMQSF indicated that the air near the surface was more stable in the western than the eastern part of the LoLaP. Hence, we can infer that the PM pollution tends to be far more severe on the western LoLaP. The near-surface wind speed determines the intensity of a front. The average wind speed near the surface has been decreasing by 0.087 m s −1 decade −1  across the world (excluding Australia; Dunn et al., 2016), by 0.07 m s −1 decade −1  in East Asia (Dunn et al., 2016) and by 0.08 m s −1 decade −1 in South Asia (Vautard et al., 2010). Additionally, the average wind speed dropped at a rate of 0.16 m s −1 decade −1 from 1966 till 2011 in China (70-135°E, 15-54°N; Liu et al., 2014), with a decrease of 0.24 m s −1 decade −1 from 1969 till 2009 in southwestern China (75-110°E, 15-35°N;Yang et al., 2012). The frequency of stagnant air days and episodes in the Sichuan-Chongqing region, which is east of the KMQSF, also increased between 2007 and 2016, with linear slopes of 0.61 and 0.26 y −1 , respectively (Liao et al., 2018b). Therefore, the intensity of the KMQSF has diminished, allowing more opportunities for air pollutants to accumulate.
This study analyzed the influence of the KMQSF on the formation and development of a light haze during early December 2016, which arose partly from favorable meteorological and terrain conditions and increased PM concentrations across the entire LoLaP (Fig. 2). Although only one event occurred during the observation period, our results suggest that the LoLaP will experience frequent pollution episodes in the future if anthropogenic emissions continue increasing in combination with a continual weakening of the KMQSF. Thus, future research should emphasize obtaining long-term regional aerosol data to further understand the role of the KMQSF in the generation and dissipation of air pollution.

Funding
This work was supported by the National Natural Science Foundation of China (41865001, 41875178, 41675149 and 41905120), Natural Science Foundation of Jiangsu Province (BK20181099) and the Program for Key Laboratory at the University of Yunnan Province.

Declaration of Competing Interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.