Downdraft influences on the differences of PM2.5 concentration: insights from a mega haze evolution in the winter of northern China

A significant haze event occurred in northern China from 16 to 21 November 2022. This study analyzed the haze spatial evolution, and meteorological influences by integrating ground and satellite measurements. Most data were obtained using aerosol lidar and wind lidar observations in suburban (Nanjiao Observation Station, NJOS) and urban Beijing (Haidian Observation Station, HDOS). The observations at NJOS and HDOS indicate the presence of a distinct layer of haze restricted to a height of up to 1500 m above the surface. However, the aerosol intensity at HDOS was comparatively lower (aerosol extinction coefficient: 1.39 ± 0.27 km−1) than at NJOS (1.77 ± 0.38 km−1), with approximately one day of time lag in response to the southerly winds. Though NJOS and HDOS presented a similar wind stratification structure, the downdraft under 1000 m influenced the surface air quality were significantly different. The intense downdraft at the lower height at HDOS prevented the vertical upward diffusion of accumulated ground pollutants, whose effect was similar to that of the inversion layer. That led to a more stable increasing trend of PM2.5 at HDOS, with the shallowest planet boundary layer height of 242 m on 20 November. By contrast, NJOS in the transportation path was more regularly influenced by the southerly flow and presented cyclical PM2.5 concentration. This study shows downdraft in urban environments acting as an accelerator for urban episodic PM2.5 pollution, suggesting the complicated contribution from meteorological factors.


Introduction
Tropospheric aerosols can degrade air quality and disrupt regional and global climate by directly scattering or absorbing light and changing cloud abundance and longevity (Ramanathan et al 2001, Kaufman et al 2002).In particular, particles with an aerodynamic diameter of less than 2.5 µm (PM 2.5 ) are harmful to human health (Zheng et al 2015).Haze pollution due to extremely high concentrations of PM 2.5 in China has attracted worldwide attention from the public and scientific community (Zhang et al 2012b, Huang et al 2014).PM 2.5 concentration exceeding 75 µg m −3 is used as the criteria for haze pollution, according to the standard 'Observation and Forecasting Levels of Haze' issued by the China Meteorological Administration (QX/T113-2010).Though the Air Pollution Prevention and Control Action Plan and other measures to improve air quality in China have been performed progressively since 2013 (Huang et al 2018), haze pollution still occur in winter due to the elevated air pollutants generated in Northern China, especially in Beijing-Tianjin-Hebei (BTH) region under certain weather conditions (Jiang et al 2015, Ye et al 2016, Gao et al 2018, Jin et al 2020).
With Beijing at its center, the BTH area in northern China is one of China's most significant metropolitan clusters.The landscape of BTH is surrounded by Yanshan Mountain to the north, Taihang Mountain to the west and Bohai Sea to the east (figure 1 However, vertical observations of haze pollution from multiple instruments at different locations in Beijing are still scarce.In addition, the study of meteorological influences, especially wind vertical velocity on the aerosol evolution and ground air quality is insufficient.Therefore, this article aims to provide a thorough review of a haze event in BTH in winter 2022.The haze event from 16 to 21 November was chosen because the wind characteristic and their different influences on the near-surface PM 2.5 at two specific sites is a matter of significant concern.Furthermore, it lasted six days and investigation is required to study the haze movement, and eventual dissipation Specifically, we will focus on (1) the spatial-temporal distribution of aerosols and the local wind patterns; (2) comparison of the vertical wind profiles (especially the downdraft) and their influences on the near-surface PM 2.5 concentration at the outskirts and urban Beijing; (3) spatial analysis combined with satellite observation and synoptic analysis.Considering the above, the ground monitoring data (aerosol lidar, wind lidar and microwave radiometer (MWR)), together with the moderate resolution imaging spectroradiometer (MODIS) observations is utilized to illustrate the vertical distribution of haze, which helps in understanding the influences of aerosol loading on the surface air quality.This paper is structured as follows.Descriptions of the equipment and the methodologies are presented in section 2. Section 3 illustrates the results of the analyses using various data.Finally, conclusions are summarized in section 4.

Surface data
The CMA Meteorological Observation Centre (CMAMOC) started a surface site network in 2013 to continuously monitor meteorological data and air quality.In this study, the PM 2.5 concentration and the ratio of PM 2.5 and PM 10 are used to study the impact of haze pollution on the near-surface air quality.The Nanjiao Observation Station (NJOS, 39.80

Meteorological data
To study the meteorologic variables and synoptic circulation related to the haze plume, we utilized the ERA5-Interim reanalysis data from the European Centre for Medium-Range Weather Forecasts (Hersbach et al 2020).The data include the daily geopotential height and wind data gridded at 0.25 • × 0.25 • with a temporal resolution of 1 h at 925 hPa.More information about ERA5 can be obtained at the Copernicus Climate Data Store (https://cds.climate.copernicus.eu,accessed on 20 February 2023).

Aerosol lidar and wind lidar
The aerosol lidar system from CMAMOC generates pulsed light at the wavelength of 532 nm with 0.5 mJ, and a Schmidt-Cassegrain telescope is used to capture the backscattered signal.This instrument exhibits profiles of aerosol backscattering intensity with a vertical resolution of 30 m and a temporal resolution of 5 min, respectively.The wind lidar is also operated at the observing network of CMAMOC.Since June 2018, this system has worked in a continuous and automatic mode.It functions at 1290 MHz with a range resolution of 120 m under 3000 m and 240 m above 3000 m.The initial range gate is positioned at a height of 150 m.Here, a vertical range of 5000 m for the wind profiles and a temporal resolution of 6 min have been used.

Microwave radiometer (MWR)
The MWR utilizes direct detection receivers with K and V bands to determine the sky brightness temperature within 800 K.The inversion algorithms presented by Rose et al (2005) are applied to extract relative humidity (RH) and temperature profiles from these bands with the vertical resolution of the lowest atmospheric layers up to 500 m is 25 m, 50 m at the altitudes between 500 and 2000 m, and 250 m above 2000 m.The temperature profiles from MVR are used to detect the PBLH based on the gradient Richardson number (Ri) method.
Besides the ground observation, we also use the MODIS to capture the Aerosol Optical Depth (AOD) evolution during the haze event.More details about the MODIS, aerosol lidar, wind lidar and MWR can be found in the supporting information.

Surface measurements
Ground-based observations of the haze were made at two monitoring sites in Beijing's southern (NJOS) and urban (HDOS) areas.Figure 2 shows the hourly surface PM 2.5 concentration and PM 2.5 /PM 10 ratio through the event at NJOS and HDOS.At NJOS (figure 2(a)), the PM 2.5 concentrations increased to 74.6 µg m −3 at 18:00 LST on 16 November.Then it fluctuated and reached a maximum value of 234 µg m −3 at 11:00 LST on 20 November.Though the PM 2.5 values sharply decreased from 183.2 µg m −3 at 22:00 LST on 20 November to 57.3 µg m −3 at 02:00 LST on 21 November, it still demonstrated an increasing trend at noon of 21 November.It indicates the existence of the haze in southern Beijing.Compared to NJOS, the rise of PM 2.5 concentration at HDOS started later and ended earlier (figure 2(b)).For example, the PM 2.5 concentration increased discernibly from 04:00 LST on 17 November and reached the value above 75 µg m −3 at 13:00 LST on 18 November.It maintained the increasing trend and experienced two peaks of 109.4 µg m −3 and 102.6 µg m −3 at 19:00 LST on 19 November and 20 November, respectively.Afterward, it dropped gradually and reduced below 75 µg m −3 at 23:00 LST on 20 November.The proportion of days in which the surface PM 2.5 /PM 10 ratio exceeded 0.5 at NJOS and HDOS was 63% (90 out of 144 d) and 14% of days (20 out of 144), respectively.It suggests that the transported anthropogenic fine mode was the predominant contributor to atmospheric aerosols at NJOS.Significant differences can be observed comparing the PM 2.5 concentration variations at the two sites.Firstly, the PM 2.5 concentration and PM 2.5 /PM 10 ratio at NJOS are more fluctuated and periodic than at HDOS.Secondly, the first time HDOS exceeded the 75 µg m −3 threshold value of PM 2.5 concentration was two days later than NJOS.It indicates the existence of a time lag regarding the arrival time of haze pollution at two sites.In addition, it is interesting to see the sharp decrease of PM 2.5 concentration at NJOS (from 108.4 µg m −3 at 22:00 LST on 18 November to 57.3 µg m −3 at 05:00 LST on 19 November, the black circle in figure 2(a)).However small fluctuations of PM 2.5 concentration at HDOS, from 73.3 µg m −3 to 86.1 µg m −3 at the same time were observed (the black circle in figure 2(b)).These differences were mainly influenced by the different wind characteristic (especially the downdraft) at the two sites, which would be further discussed in section 3.2.

Wind lidar observations
Figures 3 and 4 present the horizontal wind feature (including wind speed and wind direction) at NJOS and HDOS.The wind speeds below 1000 m at NJOS and HDOS were generally low (<5 ms −1 , also shown in figure S1), which was favorable to the accumulation of pollutants.For the horizontal wind direction, NJOS and HDOS shared a similar wind stratification structure-westerly wind in the upper air above 1000 m and the southerly winds prevalent below 1000 m.But prominent differences at two sites are observed under 1000 m.Specifically, the winds at NJOS experienced a clockwise shift from the northeast to the southwest direction from 07:00 LST to 23:00 LST on 16 November (figure 3(a)).There was not much noticeable difference except the longer duration of northeast winds on 17 November (figure 3(b)).Then the winds continued the clockwise shift with the early intrusion of easterly winds in the morning and more frequent southerly winds on 18 November (figure 3(c)).It is worth noting that the prevailing southwesterly wind began to penetrate down from the height of 4000 m to 1000 m from the noon of 18 November to the morning of 19 November.It transmitted from the prevailing northwesterly to the southwesterly at about 1000-4000 m that day (red arrows in figures 3(c) and (d)).Therefore, wind consistencies increase since winds are southerly from the near surface to the higher height (6000 m) in the afternoon.On 19 and 20 November (figures 3(d) and (e)), NJOS was entirely controlled by calm southern winds.The small southerly winds contribute to the accumulation of pollutants, leading to the surface PM 2.5 peak on 20 November.Though the lower height of northerly winds alleviated the pollution in the morning of 21 November, it shifted towards southerly winds.That resulted in the reincrease of PM 2.5 concentration at NJOS in the afternoon at NJOS in figure 2(a).
Compared to the NJOS, winds at HDOS are more variable and inconsistent below 1000 m (figure 4).That is probably due to the multiple circulation patterns, significant interference of human activities in urban areas, and the nonlinear effect of urban heat islands on wind speed in urban Beijing.Usually, weaker local winds and lower wind consistency correspond to local circulations (Hu et al 2022).Only from the noon of 18 November to the early morning of 19 November NJOS and HDOS presented the most similar wind transition feature-the southwesterly wind at 1000-4000 m gradually passed down to the lower height (red arrows in figures 3(c) and (d) and figures 4(c) and (d)).However, it presented different influences on the near-surface air quality at two sites (black circles in figure 2), which will be investigated in the next paragraph.
The vertical wind motion at NJOS (figure 5) and HDOS (figure 6) under 1000 m is generally weak with wind speeds <2 ms −1 .HDOS (figure 6) experienced more frequent downdraft and more substantial vertical air motion than NJOS (figure 5).For example, from the noon of 18 November to the early morning of 19 November, the westerly winds at 5000 m at NJOS (figures 5(c) and (d)) and HDOS (figures 6(c) and (d)) began to penetrate downward simultaneously at the two sites, which has been confirmed by the horizontal wind direction in figures 3(c) and (d) and figures 4(c) and (d).But the similar sinking pattern led to the different influences in the nearsurface PM 2.5 concentration, shown as black circles in figure 2. We suppose there are two reasons.Firstly, the downdraft at HDOS is relatively more vigorous and penetrated down to a lower height at 150 m (red arrows in figures 6(c) and (d)) than at NJOS (figures 5(c) and (d)).Hence, it prevents the vertical upward diffusion of accumulated ground pollutants, whose effect is similar to the inversion layer.Secondly, the airflow has shifted to the southerly direction at the lower height, which could also bring more transported particles to the ground.This finding reveals that the downdraft with a notable impact of either easing or worsening on the near-surface air quality largely depends on its intensity, and the location above the ground.
The vertical velocity at NJOS returned to the weak motion on 20 November and maintained to 21 November (figures 5(e) and (f)).But HDOS witnessed a more robust convective process in the morning of 21 November (red rectangular in figure 6

Spatial-temporal distribution of aerosols
Figure 7 exhibits the time-height contour plots of the lidar-derived aerosol extinction coefficient (AEC) at NJOS (figure 7(a)) and HDOS (figure 7(b)), as well as the MWR-derived PBLH (figure 7(c)) through the haze period.It provides a distinct haze evolution process of the transported aerosols in the sinking, accumulation, and dissipation stages (also shown in figure S2).At NJOS (figure 7(a)), aerosols (AEC > 0.2 km −1 ) initially appeared at ∼15:00 LST on 16 November, with a pronounced haze layer extending from the ground to 1500 m height.The arrival time of the haze pollution was nearly the same as the increase of surface PM 2.5 observations at NJOS (figure 2(a)), suggesting instantaneous sedimentation of haze particles to the surface in southern Beijing.The aerosols accumulated and were lofted to 1475 m at 15:00 LST on 18 November.Then it sank gradually and aggravated the surface aerosol mass concentration until 22:00 LST on 20 November.The haze plume disappeared at noon on 21 November.For HDOS, the intrusion of the haze appeared one day later than that in NJOS as expected, with an AEC greater than 0.2 km −1 appearing on 17 November (figure 7(b)).The aerosol layer increased to the maximum height of 1230 m at 16:00 LST on 18 November, then declined, and dissipated in the early morning of 21 November.
Figure 7(c) presented the MWR-derived PBLH at HDOS.Though no distinct inversion layer through the haze period was observed, the PBL exhibited a clear diurnal pattern, with a reduction in the peak height of PBLH observed at 1050 m, 864 m, and 722 m from 16 to 18 November, respectively.Because of the horizontal topographic blockage and weak vertical motion (figure 6), the pollutants gradually build up with RH fluctuating from 20% to 70%.The RH reached high values of >85% on 18 November.Hence, a significant number of pollutants were restricted in the upper layer and swiftly amassed (figure 7(b)), accompanied by the high RH (figure 7(c)) on that day.The PBLH kept lowering and reached its minimum average value of 242 m with the maximum RH of 95% on 20 November.The PBL returned to its diurnal feature and rose to the peak height of 1441 m at noon on 21 November, due to the considerable velocity of convection on that day (figure 6(f)).This also led to the vertical dilution of pollutants and rapidly mitigated air pollution at HDOS.

Vertical profiles of aerosol extinction coefficient (AEC)
To examine the vertical evolution of the aerosols, we extract the AEC profiles at NJOS and HDOS, with the standard deviation indicated by the error bars in figure 8. On 16 November, a pronounced haze layer was observed extending from 180 m to 1500 m height with an AEC peak of 0.84 ± 0.17 km −1 at around 690 m height at NJOS (figure 8(a)).During the following days, the haze layer descended, and the greatest AEC peak (1.77 ± 0.38 km −1 ) through this haze event was observed at 300 m on the night of 20 November.Though the haze signature is weaker than the previous days, the AEC peak value of 0.74 ± 0.15 km −1 was still observed at 900 m on the morning of 21 November.
The AEC at HDOS intensified from 300 m to 1260 m with an average AEC value of 0.25 km −1 on 17 November, one day later than NJOS.Then the AEC peak of the haze layer is detected at lower altitudes with a descending trend from 18 to 20 November, with 780 m, 480 m, and 390 m, respectively.Like NJOS, the AEC reached its maximum value of 1.39 ± 0.27 km −1 at 420 m on 20 November.Nevertheless, the haze significantly decreased on 21 November, being extremely low AEC values smaller than 0.05 km −1 , signifying the end of the haze over Beijing's urban area.
Combining the AEC contour plots of figure 7 and the vertical profiles of figure 8, we find that the haze loading in Beijing was mainly confined to 1500 m and dropped sharply with the height.The development of AEC heights reflects the same trend of decreasingfrom approximately 1200 m on 16 November to 300-400 m on 20 November.In addition, both aerosol lidars observed the AEC peak on 20 November, when surface PM 2.5 concentration reached their maximum values at two sites (figure 2) and PBLH the minimum height of 242 m (figure 7(c)) at HDOS.Because the primary source of the haze intrusion is traced from the southern region of BTH, it led to a time-lag for the arrival of the haze at two sites.And the southerly winds directly influenced the mixing and diffusion of pollutants, resulting in the higher AEC and cyclical feature of PM 2.5 concentration at NJOS.While HDOS experienced variable wind direction and dynamic wind shear due to multiple airflow circulations, the AEC results with less intensity and slower response can be taken as an extension of what is observed at NJOS.But when the solid northwesterly flow penetrated near the surface and shifted to the southerly flow, the PM 2.5 concentration at HDOS showed a different feature than that at NJOS, maintaining the stable, increasing trend.1.There is a time lag of influences of the winds on pollution development between NJOS and HDOS: For the near-surface PM 2.5 concentration, NJOS reached the value of >75 µg m −3 2 d earlier than HDOS.It is reasonable that NJOS is located in the aerosol transportation path and is earlier influenced by the polluted sources from the south direction (both southwesterly and southeasterly channels).

Summary and conclusions
2. From a single observation perspective, the NJOS witnessed the aerosol surge in the vertical scale and the increase of surface PM 2.5 nearly at the same time.It suggests the rapid sedimentation of haze particles to the surface in southern Beijing.By comparison, HDOS observed vertically high AEC values one day later than NJOS.And it takes longer time for the pollution to transport from high altitudes to the ground air quality, implying a more complicated circulation pattern in urban Beijing.3.For the vertical wind velocity at two sites, the intense downdraft is more frequently observed than the updraft at HDOS.The influences of downdraft over the surface PM 2.5 concentration in Beijing depends on its source, strength and distribution.For example, the generally stronger intensity helps faster pollutants dissipation at HDOS through the haze period.But when the sinking westerly flow penetrated to the lower height (i.e. 150 m in our study) and shifted to the southerly flow, it could prevent the vertical upward diffusion of accumulated ground pollutants, whose effect is similar to the inversion layer.
This study focuses on the local winds in the BTH, particularly in Beijing, which has significant implications for the local levels of air pollution.The study of wind vertical velocity could also help enhance the accuracy of pollution event prediction.However, more impacts still need to be researched, such as the temperature and humidity features in a variety of locations.In addition, the investigation should not just focus on the optical properties of the aerosol; rather, it should look into its chemical components in the future.

Figure 1 .
Figure 1.The location of the Beijing-Tianjin-Hebei (BTH) region is shown in figure (a).Two observing stations in Beijing-Nanjiao Observation Station (NJOS) and Haidian Observation Station (HDOS) are labeled as red circles in figure (b).
(a)).Considering the blocking effect of topography on synoptic winds, the local wind circulations become one of the main meteorological factors affecting aerosol pollution in BTH (Li et al 2020, Jiang et al 2021, Gao et al 2022).For example, high correlations between PM 2.5 and the meridional flow at 850 hPa are found in Li's study (Li et al 2022).The majority of severe pollution episodes tend to arise primarily when southerly winds are present (Zhang et al 2012a, 2016).Ren et al (2005) also found that the local winds near the ground are prone to occur under the equalized pressure field and result in the pollution convergence zone.Hu et al (2022) statistically analyzed the different weather patterns and confirmed the existence of the wind-related convergence zone in northern China.In addition, the southwesterly and southeasterly wind belts, characterized by the prevailing winds from the southwesterly to the northeast along Taihang Mountain and the sea breeze from Bohai Sea, also act as two crucial passageways for pollution transportation (Su et al 2004, Lin et al 2009, Chang et al 2018).In other cases, researchers have observed that the transportation of air pollutants from cities located to the south of Beijing had a substantial impact on the air quality of Beijing (Ma et al 2021, Xiang et al 2021).This phenomenon was found to cause a rapid surge in PM 2.5 concentrations, which could reach hundred µg m −3 within a short period of several hours (Zheng et al 2019, Chen et al 2020).Furthermore, a two-way feedback mechanism is formed between adverse meteorological conditions and the accumulation of aerosols (Zhong et al 2018, Zhang et al 2019).The accumulation of aerosol pollution within the planetary boundary layer (PBL) can substantially alter the conditions of the layer (Miao et al 2019, Su et al 2020).It can result in temperature inversion, increased humidity in the lower PBL, and a significant decrease in the PBL height (PBLH) (Miao et al 2009, Coen et al 2014, Xiang et al 2019, Li et al 2021).

Figure 2 .
Figure 2. Hourly surface PM2.5 and PM2.5/PM10 ratio at (a) NJOS and (b) HDOS from 16 to 21 November 2022.Black circles indicate the different PM2.5 concentration variations from the late evening of 18 November to the early morning of 19 November 2022 at two sites.

Figure 3 .
Figure 3. Time-height contour plots of horizontal wind feature (including wind speed and wind direction) by wind lidar at NJOS from 16 to 21 November 2022.
(f)), which helped the pollutants dissipation at HDOS through increased ventilation and turbulent mixing (figure 2(b)).It confirmed that the wind drives daily

Figure 5 .
Figure 5. Time-height contour plots of vertical wind speed by wind lidar at NJOS from 16 to 21 November 2022.The positive value represents the downdraft while the negative value represents the updraft.The red arrows in figures (c) and (d) indicate the downdraft from the noon of 18 November to the early morning of 19 November.

Figure 6 .
Figure 6.Same as figure 5 but at HDOS.Compared to figure 5, the red arrows in figures (c) and (d) indicate stronger and lower-height downdraft at HDOS during the same period.The red rectangular in figure (f) suggests the convection in the morning of 21 November.

Figure 7 .
Figure 7. Time-height contour plots of aerosol extinction coefficient (AEC) by 532 nm lidar at (a) NJOS, (b) HDOS and (c) relative humidity (RH, green contour), temperature (black line, red line indicates the zero temperature) and PBLH (dark black line) at HDOS from 16 to 21 November 2022.

Figure 8 .
Figure 8. Vertical profiles of aerosol extinction coefficient (AEC) at (a) NJOS and (b) HDOS from 16 to 21 November 2022.Errors bars are also presented.
This study evaluates a haze occurrence in winter 2022 over BTH in northern China.Particularly the development of haze properties and wind influences on 16-21 November at two Beijing stations have been examined.The haze plume was first noticed at NJOS (southern Beijing; 39.80 • N, 116.47 • E) on 16 November.Later it arrived at HDOS (urban Beijing