-
The Air Quality Index (AQI), used to assess ambient air quality in China (CEA, 2016), is classified into six grades: Grade I (0–50), Grade II (51–100), Grade III (101–150), Grade IV (151–200), Grade V (201–300), and Grade VI (> 300). For many years, the primary pollutant in Urumqi has been PM2.5, especially in winter (Li, 2013). An AQI > 200 corresponds to a daily average PM2.5 concentration > 150 μg m–3. In this paper, a daily average AQI > 200 defines a heavily polluted day. From 2013 through 2017, Urumqi experienced a total of 248 heavily polluted days (Table 1), with an annual average of 49.6 days, of which 40.8 days were in winter (82.3% of the whole year), 5.0 days in spring (10.0%), 0.6 days in summer (1.2%), and 3.2 days in autumn (6.5%).
Number of heavy pollution days Jan. Feb. Mar. Apr. May Jun. Jul. Aug. Sep. Oct. Nov. Dec. Total 2013 25 16 4 0 0 1 0 0 0 0 10 3 59 2014 11 1 2 5 2 0 0 1 0 0 0 12 34 2015 6 9 2 1 1 0 1 0 1 0 1 17 39 2016 22 19 4 0 0 0 0 0 0 0 4 14 63 2017 30 16 4 0 0 0 0 0 0 0 0 3 53 Avg. 18.8 12.2 3.2 1.2 0.6 0.2 0.2 0.2 0.2 0 3 9.8 49.6 Table 1. The number of heavy pollution days in Urumqi during 2013–17.
Pollution in Urumqi is a result of both anthropogenic emissions related to human activities and natural emissions from sources in the surrounding desert and bare land surfaces. The timing of heavy pollution events caused by these two types of emission sources rarely overlaps. Northern Xinjiang, where Urumqi is located, is covered by snow in winter, and as a result, dusty weather is rare. However, temperature inversions with light winds are prevalent in winter, which often leads to heavy pollution events. The incidence of winter inversions over Urumqi is as high as 89% of the days, and the average inversion layer thickness is 860 m above ground level (AGL) (Liu et al., 2007; Li, 2013). Some studies have linked air quality in Urumqi to the intensity of temperature inversions in this region (Li et al., 2007, 2015). Shallow foehn, which frequently occurs in winter, further strengthens the intensity and increases the thickness of the inversion layer over the city, promoting the convergence of airflows (Li, 2013; Li et al., 2015).
When cold air from the north enters Xinjiang during the warm half of the year (late spring through early fall), strong winds can lead to meso- to large-scale dust weather (Xu and Wang, 2002; Wu, 2003; Li et al., 2005; Shayiti et al., 2008), which is often referred to as northwesterly-induced dusty weather. In addition, the GBSEG originating from the MTMV can entrain sand and dust along the way to the urban area of Urumqi, leading to southeasterly-induced dust weather, also known as GBSEG-induced dusty weather (Wang and Zhang, 2014). In contrast to these dust pollution events associated with strong winds, heavy pollution events also occur under light winds and temperature inversions with or without the occurrence of ESEG.
-
A heavy pollution weather process is defined in this paper as the period when the average daily AQI changes from a minimum value to a peak value greater than 200, before dropping back to another minimum of less than 200. An examination of the AQI values between 2013 and 2017 yielded 70 heavy pollution weather processes in Urumqi that were distributed unevenly from one year to another (Table 2). Although only seven heavy pollution processes were recorded in 2017, one of them, with daily average AQI greater than 200 for 34 consecutive days, was the all-time record. Among the 70 total processes, 50, or 71.4%, belong to the type of light wind and inversion accompanied by ESEG. The next is the northwesterly-induced dusty weather type with 10 occurrences (14.3%). As for the other two types, the type of light wind and inversion without ESEG was experienced eight times (11.4%), while the GBSEG-induced dusty weather occurred only twice. As the predominant type of heavy pollution weather processes, the type of light wind and inversion along with ESEG warrants further investigation.
Causes Year Sum/Proportion 2013 2014 2015 2016 2017 Weak Wind with Inversion (WWI) 2 1 1 3 1 8/11.4% WWI with ESEG 17 8 8 11 6 50/71.4% Dusty weather by GBESG 0 1 1 0 0 2/2.9% Dusty weather by northwesterly wind 1 5 4 0 0 10/14.3% Sum 20 15 14 14 7 70 Table 2. The occurrence number of heavy pollution weather processes in Urumqi and process classifications during 2013–17.
-
Figure 4 shows temperature, humidity, and wind evolution between 1900 LST on 15 February and 1800 LST on 23 February observed from the wind profiler and the microwave radiometer in Urumqi. At the beginning of the period, on 15 February, relatively strong northwesterly winds with a peak speed of 14 m s–1 prevailed below 1800 m AGL following the passage of a trough across Xinjiang (Figs. 4c and 4d). The strong wind led to the removal of the inversion layer (Fig. 4a), lowering PM concentrations to a minimum of less than 100 μg m–3 during the midday on 16 February. The maximum vertical wind speed rose to 0.9 m s–1, which only represented the subsidence caused by the snow (Fig. 4d). Over the next two days (16–17 February, during the first phrase), winds gradually decreased to 3–4 m s–1 in the bulk of the boundary layer and less than 2 m s–1 in the surface layer (Fig. 4c), along with weak vertical motion of –0.2 to 0.0 m s–1. The PM concentrations gradually increased to just over 200 μg m–3.
Figure 4. Evolution of the boundary layer structure over Urumqi for the period 1900 LST 15 –1800 LST 23 February 2013. (a) temperature (°C), (b) humidity (%), (c) horizontal wind speed (m s−1), (d) vertical wind speed (m s−1), and (e) wind vector. The white and yellow lines denote different stages of the severe pollution event, similar to Fig. 2b.
From the evening of 17 February until the morning of 19 February (the second phase), strong easterly-southeasterly winds up to 18.2 m s–1, representing shallow foehn, occurred between 500–1700 m above Urumqi. This elevated foehn layer was sandwiched from below by weak (< 2 m s–1) northerly winds and above by moderate (4–5 m s–1) northwesterly winds. The warm foehn resulted in the development of an elevated temperature inversion layer of 0.4°C temperature difference across a 400-m-deep layer located just above the surface-based radiation inversion layer of 400–600 m deep during the nights of 17 and 18 February (Fig. 4a). This inversion was accompanied by a significant drop of 20% in relative humidity at 1000 m AGL (Fig. 4b). The strength of the southeasterly foehn wind decreased gradually and even disappeared from 18 to 19 February, while the top of the inversion layer descended from 1200 m to 900 m AGL and the temperature differences across the inversion layer increased to 11°C [an inversion intensity of 2.2°C (100 m)–1] as a result of the combination of reduced mechanical mixing at lower wind speeds and continuous warm advection aloft. It is worth noting that subsidence occurred between 300–2200 m AGL with a peak speed of 0.5 m s–1 during the active foehn stage, which could have driven pollutants from the upper boundary layer downward to the lower layers. During other stages without foehn, the ascending motion was very weak, with speeds of –0.2–0.0 m s–1 below 1500 m AGL (Fig. 4d).
Over the next two days (corresponding to the third phase), winds remained weak (less than 2 m s–1) near the surface despite the occurrence of a weak shallow foehn event with a maximum speed of 6 m s–1 that lasted from 1300 LST 19 February through 1200 LST 20 February. The inversion continued to strengthen and, together with the development of a strong ground-based nocturnal radiation inversion under weak wind conditions during the night of 19 February, significantly limited mixing and allowed the PM concentration to increase to nearly 600 μg m–3. There was a drop in PM values to below 400 μg m–3 following the mixing event that temporarily broke the inversion in the afternoon of 21 February. The next period, from 1800 LST 20 February to 1500 LST 22 February, was marked by another strong foehn event with wind speeds up to 17 m s–1 at around 800 m AGL and a maximum descending speed of 0.7 m s–1 at the same height (Fig. 4d). The strong warm advection further increased the strength of the inversion, raising PM concentration again to its peak of more than 750 μg m–3. The height of the maximum foehn wind speed rose from below 1000 m to 1200–1300 m on 22 February due to the low-level cold air advection associated with a shortwave disturbance over Xinjiang. And, the inversion layer also was lift by the cold-air. The rise of the inversion allowed the development of the mixing layer in the afternoon of 22 February, reducing PM concentration.
The final breakup of the inversion occurred on 23 February, when strong winds close to 16 m s–1 from the northwest blew cold air into the region, lowering PM concentrations to below 200 μg m–3 and marking the end of this heavy pollution episode. Table 3 displays some boundary layer parameter values and maximum pollutant concentration values during the four phases of this heavy pollution event.
Stage I II III IV 166−1710 1710−197 197 −2215 2215−237 Maximum PM2.5 concentration (μg m–3) 127 189 544 155 Maximum PM10 concentration (μg m–3) 163 267 794 196 Duration (hours) 28 45 80 16 Maximum γ [°C (100 m)–1] − 2.2 2.7 0.9 Maximum top temperature of IL (°C) − –7.5 2.9 –2.1 Maximum temperature difference of IL (°C) − 11.2 19 3.9 Note: 166, the number 16 represents the date, and the superscript 6 represents the hour. IL is inversion layer. Table 3. Boundary layer characteristics during four phases of the heavy pollution process in Urumqi on 15–23 February 2013.
-
The topography around Urumqi slopes from the southeast down towards the northwest. Once foehn exits in the MTMV, it continues to blow along the valley axis towards lower elevations in the northwest. To capture the changes of near-surface meteorological conditions during the foehn episode, hourly observations from five surface stations (see Fig. 1 and Table 4) close to the path of the foehn are analyzed (Fig. 5). During the initial phase (0600 LST 16–1000 LST 17 February), as cold air moved into the area, there was a sharp drop of ~4°C in surface temperature at all stations. The temperature differences between stations were dominated by their elevation differences. The differences in wind speed were also dominated by elevation differences, with there being weak winds of 2–3 m s–1 at Urumqi and other low-lying stations and stronger winds of 4–6 m s–1 at the WLB station located near the northern end of the MTMV. The divergence/convergence field over Urumqi, estimated using wind from ANQ and WLB (Fig. 1), showed weak divergence [(5.0–10.5) × 10–5 s–1], which was consistent with the low PM concentration during this period in spite of the generally low wind speeds.
Station Location Longitude (°E) Latitude (°N) Elevation (m) ANQ Northern suburb 87.51 43.98 564 GXQ City center 87.57 43.86 780 Urumqi Southeastern urban area 87.65 43.78 936 WLB Southern suburb 87.58 43.67 1079 SXG Hillside 87.49 43.47 1603 Table 4. Geographic locations and elevations of five representative weather stations around Urumqi.
Figure 5. Evolutions of surface meteorological parameters around Urumqi for the period 1900 LST 15–1800 LST 23 February 2013. (a) temperature, (b) wind speed, (c) wind vector, (d) divergence of wind speed components from WLB to ANQ (the blue lines denote positive value and red lines are negative value). The red dashed lines are similar to Fig. 2b
During the second period (1000 LST 17–0700 LST 19 February), Xinjiang was still under the control of a high pressure system. Surface temperatures at Urumqi and other stations were lower than during the first period, but temperatures at the high-elevation station, SXQ (> 1500 m ASL), rose significantly in the latter part of the period. This increase in temperature at high elevation was a result of midlevel warming by free-atmosphere foehn and the formation of the inversion (Fig. 4a). Near the surface, winds remained weak throughout this period and weak divergence [(5.0–10.5) × 10–5 s–1] remained over Urumqi.
During the next phase (0700 LST 19–1500 LST 22 February), there was a steady rise in temperature at all stations, and the warming rate was at 2°C each day at Urumqi station, increasing with elevation. Near-surface winds remained low (~2 m s–1) at all stations except at WLB, where strong southerly foehn winds developed, and surface wind speed increased rapidly to 14.8 m s–1 around 2100 LST on 21 February. Surface temperature at WLB also increased. It went from a low of –23.8°C at 0700 LST 19 February to a maximum of 2.9°C around 1300 LST 21 February, an increase of 27°C in just two days. Meanwhile, the rise in temperature at Urumqi station was only 10°C. The differences in the wind speeds between the upstream station (WLB) and the downstream station (ANQ) relative to Urumqi suggested a strong convergence of the flow field in the area, with a maximum value of –21.0 × 10–5 s–1. The converging flow, in combination with a strong inversion as suggested by the increase in surface temperature with elevation, explained the rapid increase in the PM2.5 and PM10 concentrations to 522 μg m–3 and 794 μg m–3, respectively, by the end of this period.
The final phase (1500 LST 22 – 0700 LST 23 February) was marked by a weakening of the pressure gradient across the Tianshan Mountain Range. The breakup of the temperature inversion was indicated by the merging of surface temperature time series from stations of different elevations. The flow field in Urumqi changed from convergence to weak divergence. Despite the collapse of the inversion, surface wind speeds were still low, causing pollutants to meander around the region.
-
Located near the northern exit of the MTMV, Urumqi is influenced by mountain–valley winds during 67% of the days in a year. In winter, northerly valley winds usually occur between 1100 LST and 1700 LST, while southerly mountain winds typically appear between 2100 LST to 0800 LST (Lv et al., 2016). This diurnal wind oscillation interacts with the shallow foehn, making wind patterns in the region highly complex during foehn episodes (Figs. 6 and 7). During the first phase (0600 LST 16 – 1000 LST 17 February), as cold air moved in from the northwest, surface winds were generally from the north and northwest. At higher-elevation (> 1400 m ASL) stations, northerly valley winds during the day (Fig. 7a) changed to southerly mountain winds at night (Fig. 6a). Weak northerly wind (< 2 m s–1) was blocked by surrounding terrain, allowing pollutants to accumulate.
Figure 6. The surface wind and temperature (colored dots) fields around Urumqi at 0600 LST on 16 February (a), 18 February (b), 21 February (c), and 22 February (d) 2013.
Figure 7. The surface wind and temperature (colored dots) fields around Urumqi at 1400 LST on 16 February (a), 18 February (b), 21 February (c), and 22 February (d) 2013.
During the second phase of the pollution episode (1000 LST 17 February – 0700 LST 19 February) (Figs. 6b and 7b), mountain–valley winds prevailed and the surface layer was decoupled from layers aloft by the formation of the inversion. The northerly valley winds started to develop around 0900 LST 17 February, reaching peak speed (3 m s–1) around 1300 LST at Urumqi station. The mountain (or downslope) winds started to develop around 1900 LST at higher elevations, and these southerly mountain winds were opposing the northerly valley winds at lower elevations, leading to convergence, which was particularly evident in areas north of the SXQ station. Another convergence area near GXQ station was enhanced by the urban heat island effect. A shallow foehn, which was first identified above 500 m (AGL) by the wind profiler around 1900 LST 17 February, penetrated down to the surface seven hours later around 0200 LST 18 February. The foehn wind speed was about 6 m s–1 at the CWP station, and it continued until about 1400 LST 18 February. Northerly valley wind to the north of WLB met with southerly foehn winds to the south of the station, producing a convergence zone between WLB station and CWP station that lasted until the mountain winds overwhelmed the foehn winds. Convergence reappeared after 2000 LST over the urban areas and persisted for 12 hours. The convergence resulting from the weak mountain winds (< 1 m s–1) and relatively strong foehn winds (6 m s–1), the weak inversion, and the urban heat island effect were all in favor of further accumulation of pollutants.
During the third phase of this episode (0900 LST 19–1500 LST 22 February) (Figs. 6c and 7c), as the increase of foehn winds further strengthened the inversion, surface winds in Urumqi decreased substantially to near-calm conditions at the five urban stations around 0200 LST 21 February. The weakening of winds was accompanied by an increase in convergence in the urban area. The shallow foehn re-appeared later in the day with the foehn front reaching the WLB station. Both the temperature (–4°C to –11°C) and wind (2–4 m s–1) in areas south of WLB were higher than the temperature (–16°C to –25°C) and wind (< 1.2 m s–1) in the urban area. Strong convergence appeared around 1400 LST resulting from the opposing valley wind and foehn wind. As the northerly valley wind sped up, the foehn front retreated to areas south of WLB. The south of the urban area continued to warm as the foehn strengthened during the night of 21 February, enhancing the convergence in the south of the urban area and around the WLB station until around 1200 LST 23 February. The strong convergence and weaker winds drove pollutants to meander and accumulate.
The surface wind field during the final phase of the episode (1500 LST 22–0700 LST 23 February) (Figs. 6d and 7d) was marked by northwesterly surface winds that began at the northern stations around 1400 LST 22 February and gradually extended southward. The foehn winds completely disappeared around 1800 LST 22 February. The removal of the inversion caused surface temperatures in the urban area to rise during the day, with the afternoon temperatures being higher than those during the previous days. The northwesterly surface winds transported pollutants downwind, leading to greatly improved air quality in Urumqi.
The following summarizes the surface wind patterns during this heavy air pollution episode. The initial phase featured cold air advection. Northerly winds occurred across most of the low-elevation areas, and diurnal mountain–valley winds appeared only at higher elevations. The second phase was marked by the initial appearance of a shallow foehn and development of an inversion. The foehn winds did not penetrate down to the surface in the southern suburbs, and the wind field in the urban area reflected the influence of diurnal mountain–valley winds and weak convergence enhanced by the urban heat island effect. During the third phase, the shallow foehn front propagated further northward and penetrated down to the surface just south of the Urumqi city. The foehn winds and diurnal mountain–valley winds led to persistent convergence, which, combined with the intensification and lowering of the inversion layer, accelerated the accumulation of pollutants and worsened air pollution. The final phase of the heavy pollution episode was characterized by the removal of the inversion as a result of relatively strong northwesterly winds and cold air invasion into the region.
-
The results above show that temperature inversion and wind field patterns play key roles in the development of heavy pollution episodes in Urumqi during the winter season. Three parameters are used to characterize temperature inversions, namely, temperature at the top of the inversion layer, temperature difference between the top and bottom of the inversion layer, and the temperature gradient across the inversion layer (the intensity of inversion layer). In addition to an inversion, convergence over the urban area caused by opposing wind directions between the diurnal mountain–valley winds and the foehn winds also helps to accumulate pollutants. The convergence can be estimated by the north–south wind vectors between the WLB and ANQ stations. Both the inversion strength and the surface convergence are closely related to the evolution of the north–south pressure difference across the Tianshan Mountains (Zhang et al., 1986; Li et al., 2012, 2020), which can be estimated using sea level pressure observations from the WLB station and the DK station in the Turpan Basin on the south side of the Tianshan Mountains (Fig. 1).
The PM2.5 concentration at Urumqi during this episode had a strong relationship with the three inversion parameters, with correlation coefficients of 0.61, 0.55, and 0.60 for inversion top temperature, temperature difference across the inversion layer, and temperature gradient, respectively. The PM2.5 concentration is negatively correlated to the convergence/divergence indicator (–0.50), the regional pressure difference (–0.56), and the mixed layer height (–0.42). Table 5 presents the correlation coefficients between the above parameters.
Inversion layer parameter Mixed layer Height Divergence Air pressure difference Temperature at the top Temperature difference Intensity PM2.5 0.612 0.547 0.591 −0.419 −0.497 −0.557 Table 5. Correlation coefficients between temperature inversion, north–south divergence, and air pressure difference and PM2.5 concentration in Urumqi on 16–23 February 2013.
Similarly, the AQI values also exhibit strong relationships with these indicators. Among them, the temperature at the top of the inversion layer along with the surface pressure difference (Fig. 8) appear to be the best indicators. A careful examination of the scatter plots reveals three zones that match with the three grades of air pollution. For instance, when AQIs are lower than Grade IV, the corresponding north–south surface pressure difference varies between 10 hPa and 15 hPa, while the temperature at the top of the inversion layer is between –20°C and –16°C. For AQIs in Grade IV, the pressure difference is 4–10 hPa and the temperature at the top of the inversion layer is –14°C to –10°C. High AQIs above Grade V correspond to pressure differences between –7.5 hPa and 10 hPa and inversion top temperature between –10°C and 0°C. More wintertime data are necessary to confirm these relationships and develop better indicators for predicting heavy pollution events.
Number of heavy pollution days | |||||||||||||
Jan. | Feb. | Mar. | Apr. | May | Jun. | Jul. | Aug. | Sep. | Oct. | Nov. | Dec. | Total | |
2013 | 25 | 16 | 4 | 0 | 0 | 1 | 0 | 0 | 0 | 0 | 10 | 3 | 59 |
2014 | 11 | 1 | 2 | 5 | 2 | 0 | 0 | 1 | 0 | 0 | 0 | 12 | 34 |
2015 | 6 | 9 | 2 | 1 | 1 | 0 | 1 | 0 | 1 | 0 | 1 | 17 | 39 |
2016 | 22 | 19 | 4 | 0 | 0 | 0 | 0 | 0 | 0 | 0 | 4 | 14 | 63 |
2017 | 30 | 16 | 4 | 0 | 0 | 0 | 0 | 0 | 0 | 0 | 0 | 3 | 53 |
Avg. | 18.8 | 12.2 | 3.2 | 1.2 | 0.6 | 0.2 | 0.2 | 0.2 | 0.2 | 0 | 3 | 9.8 | 49.6 |