STUDY OF THE VERTICAL WIND FIELD STRUCTURE AND ITS RELATIONSHIP WITH PM2.5 AIR POLLUTION OVER THE PEARL RIVER DELTA IN AUTUMN

The effect of meteorological structure is one of the key impact factors affecting air quality. In recent years, many pure and application-oriented research studies have been performed to scientifically and technologically support the prediction and management of air pollution. Many meteorological variables significantly affect air quality by controlling the transportation and dispersion of pollutants (Cogliani^[1]; Khedairia and Khadir^[2]; Tan et al.^[3]; Menut et al.^[4]). For example, studies have shown that wind speed and atmosphere boundary layer (ABL) height were key variables in analyzing the effect of meteorological structure on air pollution (Krautstrunk et al.^[5]; Davies et al.^[6]). In several cities, influenced by circulation patterns on regional transport pathways, regional emission control has become more important than local emission control (Schleicher et al.^[7]; Zhang et al.^[8]). Research has shown that pollutants from surrounding regions can be transported to urban areas (Lin et al.^[9]; Wang et al.^[10]). Based on measurements with advanced equipment, such as sodar and radar acoustic sounding systems, ABL scaling parameters and energy fluxes associated with poor air quality have been analyzed (Neff et al.^[11]; Hanna et al.^[12]; Emeis et al. ^[13]; Alappattu and Kunhikrishnan^[14]). These analyses have shown that the weak wind and low inversion conditions at night play important roles in poor air quality (Kolev et al.^[15]; Sang et al.^[16]) and that the weak and intermittent turbulence in the nocturnal ABL facilitates the transport of pollutants from higher altitudes to the surface (Pournazeri et al.^[17]; Salmond and McKendry^[18]).

As a subtropical coastal metropolitan area, the Pearl River Delta (PRD) suffers from major atmospheric environmental pollution issues. The meteorological conditions associated with air pollution in the PRD are quite different from those in other metropolitan areas in China (Chan et al.^[19]; Wu et al.^[20]). On the one hand, local circulation patterns, such as sea-land breezes, are crucial factors in pollution episodes (Ding et al.^[21]; Fan et al.^[22]). On the other hand, tropical cyclones are also important weather systems that influence the PRD in summer and autumn because tropical cyclone subsidence reduces the ABL height and produces surface flow stagnation, which in turn limits the advection or diffusion of locally emitted pollutants (Chen et al.^[23]; Feng et al.^[24]; Wu et al.^[25]). The synoptic situations and the ABL structure over the PRD have been discussed by Fan et al.^[26-27], who developed a conceptual model that shows the relationship between regional air quality and ABL meteorological conditions for the PRD (Fan et al.^[28]). Including Guangzhou, Shenzhen, and Hong Kong, more than eleven cities are clustered in a relatively small area, and long-range transport from regional emission sources also significantly affects the air quality in the PRD (Ding et al.^[29]; Xiao et al.^[30]).

This study discusses the influence of the vertical wind field structure (the key meteorological factor affecting air quality) on surface PM_2.5 concentrations in the PRD and investigates the boundary layer wind, daily change in ABL height, ventilation index (VI), recirculation and sea-land breezes. These variables and processes have a direct impact on PM_2.5 concentrations and have commonly been used to design emission reduction strategies in previous studies.

Air quality and upper-air and surface meteorological data were collected from October 1, 2012 to October 31, 2012. This study focuses on the surface PM_2.5 data collected at Panyu and on the surface and upper-air meteorological data collected at Qingyuan, Zengcheng and Nansha (Fig. 1). The surface and upper-air meteorological data at Zengchen and Nansha were collected by a radar wind profiler (RWP), whereas those at Qingyuan were collected by rawinsonde.

Figure 1.  Location of the observation stations.

The RWP at Zengcheng measured winds with a vertical resolution of 60 m from the height of 100 m to 5500 m, and the temporal resolution was 30 min. The RWP at Nansha had similar settings but from the height of 88 m to 3100 m. Temperature was measured by rawinsonde at Qingyuan at 08:00 and 20:00 LST every day, and the vertical resolution was 200 m to 300 m depending on meteorological conditions. The RWP data were used to determine the boundary layer wind, VI and recirculation factor (RF). The rawinsonde data were used to calculate the daily ABL height.

The ABL height was estimated from the rawinsonde releases with the use of a technique suggested by Heffteret^[31]. In this method, potential temperature profiles were computed for each sounding. The profiles were analyzed for the existence of a 'critical inversion', which is assumed to mark the top of the ABL. In this scheme, a critical inversion is defined as the lowest inversion that meets the following two criteria:

(ⅰ) Δθ/ΔZ ≥ 0.005°Cm^-1

(ⅱ) θ_vtop - θ_vbase ≥ 2°C

where Δθ /ΔZ= the potential temperature lapse rate in the inversion layer, θ_vtop= the potential temperatures at the top and θ_vbase= the potential temperature at the bottom.

As an example, Fig. 2 shows the ABL height at Qingyuan at 20:00 LST on October 13, 2012 estimated based on criteria (ⅰ) and (ⅱ) above. The ABL height was approximately 200 m, indicating that this method is applicable to the PRD.

Figure 2.  (a) Potential temperature, (b) specific humidity and (c) potential temperature lapse rate at 20: 00 LST on October 13, 2012.

As a useful tool for air pollution management, the VI is the product of wind speed and ABL height (Pasch et al.^[32]):

where i is the level at which wind speed values were recorded, h_i is the height of level i, v_i is the wind speed at level I and ABLH is the daily mean ABL height.

In this study, VI was computed for each day using the wind data at multiple levels within the ABL, and the daily mean ABL height was estimated based on the temperature profile obtained from the Qingyuan radio sounding measurements.

Recirculation is an event in which polluted air is initially carried away from the source region but subsequently returns to produce a high pollution event (Fig. 3a). The RF is the ratio of the resultant transport distance to the scalar transport distance (wind run) (Fig. 3b). The RF can be used to infer the air parcel movement and the dispersive characteristics (e. g., ventilation, stagnation or potential recirculation) of a given air flow (Allwine and Whitman ^[33]). When the RF is equal to 1, which is a straight line, steady transport occurs (Fig. 3c). However, when the RF is equal to zero, no net transport occurs, as described in Eq. (2).

Figure 3.  Diagrammatic sketches of the recirculation factor calculation.

where l is the real trace of wind, h is the linear distance of wind, i is the time of each data point, t_s is the time of the starting data point, t_e is the time of the ending data point, ΔT is the averaging time interval of the data, u_t is the x-velocity component (north is defined as the positive x-axis) and v_t is the y-velocity component (east is defined as the positive y-axis).

Figure 4 provides an overview of the hourly concentrations of PM₁₀ and PM_2.5, the hourly temperatures and the wind speeds at Panyu station. The concentrations of PM₁₀ and PM_2.5 were very high in October 2012 and exceeded the class 2 limit values of the National Ambient Air Quality Standard (NAAQS) for at least 18 days. Furthermore, the concentrations almost consistently exceeded the class 1 limit values of the NAAQS (class 1, 35 µg m^-3; class 2, 75 µg m^-3). At times, the hourly PM_2.5 concentrations were even higher than 160 µg m^-3. The trend of the PM_2.5 concentrations was highly similar to that of the PM₁₀ concentrations; the PM_2.5 concentrations reached more than 80% of the PM₁₀ concentrations most of the time. This result indicated that the pollution of fine particles was relatively serious and was the major factor influencing visibility.

Figure 4.  Time series of the hourly data of (a) PM₁₀ and PM_2.5, (b) temperature and wind speed, and (c) surface wind vectors at Panyu in October 2012.

The time series of hourly temperature and wind speed at the Panyu station showed that three cold air masses affected the PRD, the wind speed significantly changed and temperature progressively decreased during this period (Fig. 4b). The wind speed was weak during the periods with a high PM_2.5 concentration. When the wind speed was strong on October 10, 18 and 31, 2012, the concentrations of PM_2.5 and PM₁₀ decreased remarkably. Based on Fig. 4c, influenced by the meeting of cold and warm air masses, the wind in the PRD was quite weak, and the direction changed frequently, thereby weakening the transportation capacity of the wind field.

According to the changes in PM₁₀, PM_2.5, temperature and wind speed, three pollution processes were identified in the PRD in October 2012. Table 1 shows the features of each pollution process. The synoptic situation is a representation of micro scale characteristics in relation to the regional atmosphere conditions. Table 1 shows that the synoptic situation associated with PM_2.5 air pollution can be divided into two types: a southward weak cold air mass and a typhoon. When a weak cold air mass approached the PRD, the PRD was dominated by calm wind as a result of the meeting of cold and warm air masses, and the calm wind in the region reduced the pollutant diffusion. The subsidence was very strong when the typhoon was close to the PRD. As a result, pollutants constantly accumulated, and the air quality rapidly deteriorated. In particular, high PM_2.5 air pollution occurred in the PRD from October 12, 2012 to October 17, 2012 because of the combined influence of Typhoon Prapiroon and a cold air mass.

Figure 5 shows the weather maps on October 12, 2012 and October 23, 2012, during which two high PM_2.5 pollution processes occurred. On October 12, 2012, most of northern China was controlled by a cold high, and southern China was just in front of this cold high. At the same time, the center of Typhoon Prapiroon was over the ocean east of Taiwan, and the PRD was to the left of the path of the typhoon and was controlled by the subsidence associated with Typhoon Prapiroon.

Figure 5.  Weather maps during PM_2.5 air pollution days in the PRD on October 12 and October 23 (Red box represents focus area).

Under the influence of the cold high and subsidence, a uniform pressure field appeared in southern China, which resulted in poor diffusion conditions with weak airflow activities. On October 23, 2012, a new cold air mass moved south, and the PRD was in front of the center of this cold high. The straight and sparse isobars associated with a weak east wind limited the diffusion of pollutants in the PRD.

Figure 6 shows the time series of wind speed in each height in October 2012 at Zengcheng and Nansha. The macro scale trends of these two stations were quite similar but still showed some subtle differences between them. During PM_2.5 air pollution days, the wind speed at Zengcheng station was less than 3 m s^-1 at a height of 1400 m, and the wind speed at the surface layer was less than 1.5 m s^-1. These results indicated that the ventilation ability inside the ABL was very weak under the influence of the meeting of cold and warm air masses. During non-pollution days, the wind speed significantly increased. Similarly, at the Nansha station, the wind speed was less than 3 m s^-1 at a height of 1000 m during PM_2.5 air pollution days. The heaviest pollution was recorded from October 12, 2012 to October 17, 2012. Both stations featured a weak wind layer (WWL), which could reach thicknesses of 1500 m during pollution days. The thick WWL was not conducive to the horizontal transmission of pollutants and was the primary cause of air pollution during autumn. However, unlike Zengcheng, Nansha had a remarkable diurnal variation in wind speed at a height of 600 m during pollution days; this phenomenon was influenced by active sea-land breezes.

Figure 6.  Time-height cross sections of wind speed in October 2012 at (a) Zengcheng and (b) Nansha.

Figure 7 shows the difference in the average wind speed and direction profiles over the full month, on PM_2.5 pollution days and on non-pollution days in October at Zengcheng. As a general view, the structures of the average wind speed profiles for the whole day (00: 00-23:00), the daytime (06:00-17:00) and the nighttime (00: 00-05: 00 and 18: 00-23: 00) were similar, and all could be divided into three layers (Fig. 6a, b and c). In contrast, the average wind direction profiles could only be divided into two layers (Fig. 6d, e and f). During the pollution days (Fig. 6b and e), the average wind speed profiles were divided into three layers at heights of 600 m and 1700 m. In the lower layer, the average wind speed values of the whole day, the daytime and the nighttime were extremely similar and very small. The wind direction in all situations changed from westerly (270°) to southerly (180°), but the change during the nighttime was more substantial. In the middle layer, the average wind speed was lower during the nighttime than during the daytime, but in the upper layer, the average wind speed was higher during the nighttime than during the daytime; The average wind directions of each situation in the middle and upper layers were almost the same. The structure of the wind profiles during the non-pollution days (Fig. 6c and f) was quite similar to that during the pollution days, but the upper boundary of the lower layer rose to 1000 m, and the average wind speed was 2 m s^-1 higher.

Figure 7.  The average wind speed and direction profiles (a, d) during the full month of October 2012, (b, e) on PM_2.5 pollution days and (c, f) on non-pollution days at Zengcheng.

Figure 8 shows the average wind speed and direction profiles for the full month, on PM_2.5 pollution days and on non-pollution days in October at Nansha. During the pollution days (Fig. 7b, e), the average wind speed profiles can be divided into three layers at heights of 600 m and 1700 m. In the lower layer, the average wind speeds for the whole day, the daytime and the nighttime were quite different, and the average wind speed of the nighttime was 1~2 m s^-1 higher than that of the daytime. The wind direction was approximately E (180°) during the nighttime approximately ESE (112.5°) during the daytime. In the middle and upper layers, the average wind speeds and wind directions of the nighttime and daytime were almost the same. The structure of the wind profiles during the non-pollution days (Fig. 7c and f) was remarkably different from that during the pollution days. Furthermore, the wind speed in each layer during the nighttime was approximately 0.5~2 m s^-1 higher than that during the daytime. In contrast, the wind direction was similar between the daytime and the nighttime.

Figure 8.  The average wind speed and direction profiles (a, d) during the full month of October 2012, (b, e) on PM_2.5 pollution days and (c, f) on non-pollution days at Nansha.

In conclusion, the average wind profiles of Nansha were remarkably different from those of Zengcheng, especially during the pollution days, and the wind speed and direction profiles of Nansha exhibited remarkable diurnal variation. These observations suggest that sea-land breezes had an important effect on the air quality in the PRD coastal area.

Figure 9 shows the time series of the daily ABL height in October 2012 at the Qingyuan station. A significant difference in daily ABL height was detected between PM_2.5 air pollution days and non-pollution days. The mean ABL height was lower than 500 m during pollution days but higher than 1400 m during non-pollution days. From October 2, 2012 to October 15, 2012, the ABL height remained low under the influence of a cold air mass and the typhoon, with a minimum of only 150 m. In general, the ABL height was less than 600 m during most pollution days, and the low ABL height prevented the spread of pollutants. As a result, air pollutants accumulated inside the ABL, resulting in an increase in PM_2.5 concentration at the surface.

Figure 9.  Time series of the daily mean ABL height in October 2012 at the Qingyuan station.

Figure 10 shows the time series of daily PM_2.5 concentration and VI in October 2012. The VI was calculated with the vertical wind data at the Zengcheng station. A strong negative correlation existed between PM_2.5 concentration and the VI. When the VI was high, the PM_2.5 concentration was low. When the VI was low, the PM_2.5 concentration was high. This pattern indicated that the VI is a useful tool in describing the ventilation ability of the ABL. During PM_2.5 air pollution days, the VI was usually less than 2000 m² s^-1, with a minimum of 165 m² s^-1 on October 6. The VI was far higher during non-pollution days than during pollution days, with a maximum of 8970 m² s^-1.

Figure 10.  Time series of the daily PM_2.5 concentration and ventilation index in October 2012 (black dotted line: class 2 limit values of the NAAQS).

Table 2 shows the correlation coefficients between the daily PM_2.5 concentration and the ABL height, surface wind speed and VI. The ABL height, wind speed and VI all appear to be negatively correlated with the PM_2.5 concentration; in particular, VI exhibited a much stronger correlation than the others, with a correlation coefficient of approximately - 0.65. Thus, VI can better represent the transportation capacity of the wind field.

Figure 11 shows the vertical RFs at the Nansha station in October 2012. During the PM_2.5 air pollution days, the RF was quite small at a height of 800 m. Particularly, the low value centers of RF (i. e., < 0.6) always occurred at heights from 300 m to 600 m. The wind direction changed substantially throughout the pollution days in the lower level of the ABL, and recirculation was also significant. Therefore, the pollutants were initially transported away from the PRD but subsequently returned. However, the RF was usually larger than 0.8 throughout the ABL during non-pollution days. These high RF values indicated that the wind direction was very steady and little recirculation occurred from October 10 to 11 and from October 27 to 31 in 2012.

Figure 11.  Time-height cross sections of RF in October 2012 at Nansha station.

Figure 12 shows the mean RF profiles of PM_2.5 pollution days and non-pollution days at Nansha. The RFs during PM_2.5 pollution days were approximately 0.4 to 0.6 below a height of 800 m, and during non-pollution days, they were mostly greater than 0.8. This indicates that recirculation was highly reactive in the ABL over the PRD, causing the transportation capacity of the wind field to be only 40 to 60%.

Figure 12.  The mean RF profiles of (a) PM_2.5 pollution days and (b) non-pollution days at Nansha station.

Figure 13 shows the correlation coefficient between daily PM_2.5 concentration and RF at each height and the RFs at 600 m and 1500 m at Nansha. The correlation coefficient profile between daily PM_2.5 concentration and RF had a clear minimum value area (Fig. 13a), and the correlation coefficient values decreased from -0.4 to -0.7 with height and then gradually increased to zero above 600 m. This pattern indicated that the PM_2.5 concentration was strongly associated with the RF under 600 m. Furthermore, the PM_2.5 concentration was negatively correlated with the RF at 600 m (Fig. 13b). When RFs fell below 0.6, the transportation capacity of the wind field significantly weakened, and the PM_2.5 concentration increased quickly. The RFs at 1500 m were approximately 0.8~1.0, and the correlation coefficient with PM_2.5 concentration was approximately zero, showing no correlation with surface air quality.

Figure 13.  (a) The correlation coefficient between daily PM_2.5 concentration and RF at each height and (b) the series of daily PM_2.5 concentrations and the RFs of 600 m and 1500 m at Nansha.

Sea-land breezes, a typical type of recirculation, had a very important influence on air quality. Fig. 14 shows the hourly average temperature at Panyu and Nansha and the variation in wind vectors at 150 m during days with sea-land breezes. The thermal contrast between land and sea was the basic requirement for sea-land breezes (Fig. 14a). In the daytime, Panyu was warmer than Nansha, with a maximum temperature difference of approximately 2ºC, and such conditions are conducive to generating sea breezes. Because of the influence of sea-land breezes, the surface wind direction changed clockwise over time, and the wind speed gradually increased as the sea breezes developed in the afternoon (Fig. 14b).

Figure 14.  The series of (a) hourly average temperature at Panyu and Nansha stations and (b) the variation in wind vectors at 150 m at Nansha during days with sea-land breezes.

Figure 15 shows the vertical wind and RF profiles obtained on October 4. The results indicated the presence of sea-land breezes. The sea breeze started at approximately 17: 00 LST, and the wind direction was southeast. The sea breeze was maintained until 23: 00 LST, and its peak influence occurred at approximately 20: 00 LST at a height of approximately 600 m. During the daytime, the land breeze started at approximately 04: 00 LST, which caused the wind speed to decrease and the wind direction to turn easterly. The vertical distribution of RF values also reflected the change in the vertical wind field, and the RF values below a height of 600 m were low, with a minimum value at approximately 400 m. The characteristics of the RF distribution and wind profiles indicated that recirculation related to sea-land breezes was highly important in the polluted situations in the PRD.

Figure 15.  The series of (a) hourly average temperature at Panyu and Nansha stations and (b) the variation in wind vectors at 150 m at Nansha during days with sea-land breezes.

With the influence of sea-land breezes, the RF on PM_2.5 pollution days was approximately 0.2~0.6 within the ABL, and the transportation capacity of the wind field was very low, only 20~60%. Moreover, when sea breezes developed in the night, they may transport pollutants back inland and produce high PM_2.5 concentrations.

In conclusion, when the PRD was influenced by the meeting of cold and warm air masses, stagnated conditions emerged (Fig. 16). Under these stagnated conditions, sea-land breezes actively developed. The horizontal transmission capacity of the atmosphere was quite weak, promoting pollutant accumulation and causing an air pollution episode.

Figure 16.  Schematic diagram of the vertical wind field structure with PM_2.5 air pollution in autumn over the PRD.

Table 3 shows features of surface wind, RF, ABL height, VI with PM_2.5 air pollution in autumn over the PRD. As the key indicator of a stagnated air mass, the wind strength of the system was very low, and most wind conditions were categorized as calm (v < 3 m s^{- 1}). The calculation shows that the average ABL height in the PRD during pollution days was always lower than 400 m. Under the influence of calm wind and a low ABL height, the average VI was always less than 2000 m² s^-1. Furthermore, the air quality of the PRD had a significant relationship with recirculation, and the RF values at 100-400 m were approximately 0.5-0.6. Thus, the horizontal transmission capacity of the vertical wind field was weak under the influence of the sea-land breezes.

The vertical wind field characteristics associated with PM_2.5 air pollution in the PRD were studied on the basis of observational data on PM_2.5 concentrations, radar wind profiles, and other surface meteorological data in October 2013.

(1) Under the influence of the meeting of cold and warm air masses, the wind field of the ABL over the PRD showed different characteristics between PM_2.5 air pollution days and non-pollution days in October 2012. During pollution days, the wind field was very weak and mostly calm, and the vertical wind structure could be divided into three levels. In the near-estuary area, the wind speed and direction profiles at Nansha exhibited a remarkable diurnal variation. The vertical wind field featured a thick WWL with wind speeds less than 3 m s^-1 and a thickness of approximately 1000 m. The thick WWL was not conducive to the horizontal transmission of pollutants and was the primary cause of air pollution in autumn over the PRD.

(2) A strong negative correlation was detected between the PM_2.5 concentration and the VI, and the VI can effectively reflect the atmospheric diffusion ability. During the pollution days, the VI and atmospheric diffusion ability decreased under the influence of a thicker WWL and a lower ABL height. The calculation shows that the daily ABL height in the PRD during pollution days was approximately 200~500 m, and the daily VI was approximately 500~2000 m² s^-1, which means that the atmospheric diffusion ability was very weak, and the pollutants accumulated easily.

(3) Because of the weak wind system, sea-land breezes occurred frequently, and the RF could be used to quantitatively analyze the influence of sea-land breezes on the atmospheric horizontal transmission capacity. The RF values were low at a height of 800 m during pollution days, and the centers of low RF values always occurred between the heights of 300 and 600 m. The RF characteristics and wind profiles indicated that sea-land breezes were highly important in the development of PM_2.5 air pollution in the PRD. The RF values during PM_2.5 pollution days were approximately 0.4 to 0.6 below a height of 800 m, indicating that recirculation was highly active in the ABL over the PRD and that the transportation capacity of the wind field was only 40 to 60%. The sea breeze started at 16: 00 and reached a maximum at 20: 00, with the strongest influence at heights of approximately 600-800 m. The sea breeze may transport the pollutants back inland and produce a peak in PM_2.5 concentrations at night.

The present results show that the vertical wind field characteristics are associated with PM_2.5 air pollution in the PRD. The results are useful not only in developing atmospheric models and establishing related policies but also in providing information for similar studies conducted in other places. The regional air quality in the PRD is strongly dependent on the meteorological characteristics, which are very complex and volatile. Further observational studies and numerical simulations are necessary.