A Comparison of the Different Stages of Dust Events over Beijing in March 2021: The Effects of the Vertical Structure on Near-Surface Particle Concentration

: Mineral dust is of great importance to climate change, air quality, and human health. In this study, multisource data, including the reanalysis data and remote sensing data, were used to compare the three dust events that occurred in the March of 2021 over Beijing and reveal the effects of atmospheric vertical structure on near-surface dust concentration. The combined effect of the Mongolian cyclone and a wide persistent cold-front induced two events (E1: from March 15 to 16 and E3: from March 28 to 29). E1 was more intense, more extensive, and longer-lasting than E3 due to the combination of the stronger Mongolian cyclone, slower high/cold surface pressure, and the low-level jet. However, under the appropriate conﬁgurations of temperature and pressure ﬁelds between high and low altitudes, weak updrafts were still induced and could elevate dust up to 850 hPa, as occurred during E2 on March 22 and 23. The dust emission was inferior to E1 and E3, which contributes to the low dust concentration near the surface in E2. On the other hand, the downdraft strength directly affected both the vertical distribution of dust and the concentration of surface particles. There was a strong temporal consistency between the occurrence of the downdraft and the dust touchdown. In E1, the continuous strong downdraft caused the maximum dust concentration to be above 4000 µ g/m 3 at around 200 m. In contrast, the maximum height of the dust mass concentration in E3 occurred at about 800 m due to the transient downdraft, which weakened its effect on surface visibility. Besides, the weak vertical motion in E2 caused most of the dust to become suspended in the air. Overall, the large dust emission resulted from active updrafts in the source region, and the lengthy strong downdrafts led to the ultrahigh particle concentration near the surface.


Introduction
Mineral dust is a key component of the Earth system. By absorbing and scattering the solar and terrestrial radiation, dust aerosols alter the radiation balance and energy  [46,47]. The detailed parameters are presented in Table 1. The dual-polarization 532 nm receiver makes the depolarization ratio possible. If the lidar ratio is given, the extinction and backscatter coefficients are obtainable with the Fernald method [48]. Considering the atmosphere during the study period was dominated by dust aerosols, we determined the lidar ratio to be a constant of 50 sr in this paper, which is the typical value for the lidar ratio of dust, not only from Asia but also from the Sahara Desert [49,50]. Before the retrievals, the raw lidar signal was corrected by background subtraction, geometrical factor correction, and range square correction. The Rayleigh signal profile in the middle to upper troposphere, obtained from the U.S. Standard Atmosphere model, was used to calibrate the measured 532 nm signal. Besides, we focused on the lidar signal above 150 m in order to avoid the effect of incomplete overlap in our study. Particle depolarization ratios of 0.3-0. 35 have often been observed in field experiments involving dust from East Asia [51,52] and other regions [53,54], although the value of anthropogenic aerosols is lower at less than 0.1-0.2 [49,55]. Due to the considerable difference between dust and spherical aerosols in the particle depolarization ratio, the available depolarization ratio at 532 nm of the polarization lidar facilitates the quantification of dust profiles, which is also the principle of POLIPHON. The details about one-step POLIPHON are described by Mamouri and Ansmann [34]. In this paper, the thresholds of dust and spherical aerosols are 0.35 and 0.05, respectively, according to the characteristic of Asian dust over Beijing [56].

Reanalysis Datasets
The large-scale circulation data from the European Centre for Medium-Range Weather Forecast (ECMWF) Reanalysis v5 (ERA5) were used to study regional atmospheric circulations during the dust events. This study combined the pressure field at sea level and the geopotential height field at 500 hPa to describe the horizontal circulation. It used divergence and vertical velocity data at 27 levels from 1000-100 hPa for the vertical condition. The ERA5 reanalysis data, with an hourly temporal resolution and a spatial resolution of 0.25 •

Trajectory Model
The Hybrid Single-Particle Lagrangian Integrated Trajectory Model (HYSPLIT) (available online at https://www.ready.noaa.gov/index.php, accessed on 18 April 2021) is a comprehensive modeling system for calculating and analyzing the transport and diffusion trajectories of air pollutants. The model was operated using meteorological data from the Global Data Assimilation System. In this study, the dust source region and transport path were tracked using the HYSPLIT model.

FY-4A Satellite
The full disk data of the dust detection (DSD) level-2 product used in this paper was provided by the FY-4A satellite from the China National Meteorological Satellite Center (http://www.nsmc.org.cn, accessed on 25 May 2021). It characterizes dust density at spatial and temporal resolutions of 4 km and 1 h. The quality control of dust intensity (IDDI-BK) was conducted using the dust fraction (DST). Only the IDDI-BK data with a DST value higher than 10 were valid [57].

Pollutant Data
Nationwide hourly surface particulate matter (PM 10

Overview of Dust Events
In March of 2021, three dust events of differing intensities were captured. Figure 1a-c shows the coverage of influence by the dust events. The dust storm on March 15 mainly affected northern China and also impacted the central and eastern cities in the provinces of Henan and Shandong. The weak-intensity dust event on March 22 only involved the Beijing-Tianjin-Hebei region. As shown in Figure 1c, the area affected by the dust on March 28 was less extensive than that on March 15, but the impact of the dust was still significant. An overview of the surface pollutants over Beijing is presented in Figure 1d, as the city was affected continuously by dust. Two strong dust storms are evident (E1: from March 15 to 16 and E3: from March 28 to 29), along with a minor floating dust event (E2: from March 22 to 23). The events were characterized by a ratio of PM 2.5 to PM 10 being less than 0.25. Statistically, E1 was the largest and most widespread dust storm over the past decade, which lasted for over 12 h in Beijing with the maximum hourly PM 10 of 8916 µg/m 3 . Both E1 and E3 were followed by a decreasing-increasing trend, which indicated the clean-up and backflow stages. Subsequently, the increasing ratio of PM 2.5 to PM 10 signaled the end of the dust event and the dominance of PM 2.5 . During E2 (the floating dust event), the PM 10 concentration remained close to 300 µg/m 3 .
According to the traditional conceptual model, a complete dust process in a non-dustsource region is divisible into three distinct stages: dust emission, long-range transport, and outbreak. The following sections present a detailed comparative analysis of the three dust events in horizontal and vertical directions based on multisource data. This plane-to-point analysis reveals the effects of vertical dust distribution on near-surface particle concentration.

Dust Emission Stage
The two major systems that influence dust events in northern China are the Mong lian cyclone and cold fronts [58]. Figure 2 clearly shows that the Mongolian cyclone w located in Mongolia at 500 hPa on March 14 (E1) and March 27 (E3), although the cyclon central intensity and influence were stronger in E1. In the outer area of the cyclone, wind speed reached 40 m/s, while a straight westerly flow dominated most of northe China. Figure 3a shows that these conditions corresponded well to the dust density d tribution on March 27, and there was good consistency between the center of the cyclo and the dust intensity. At sea level, the low-pressure center occurred near northeast Ch on March 14 and 27, where there was a cold front. The dense isobars and the large pressu gradient before and after the cold front produced gale conditions. According to Lei a Wang [59], cold-front-induced dust storms are characterized by rapid and strong d emissions. Conducive to dust emission was the confluence of warm and cold air, wh rendered the lower atmosphere unstable. However, in the E3 event, the temperature d ference between the cold and warm air was less significant than in E1. According to Ta 2, which quantitatively gives the maximum wind speed of the dust emission stage, smaller maximum horizontal wind speed in E3 was reasonable. Besides, as shown in F ure 2b, E2 had no typical synoptic system of dust emission unlike the strong dust storm However, the configuration of the temperature and pressure fields between the high a

Dust Emission Stage
The two major systems that influence dust events in northern China are the Mongolian cyclone and cold fronts [58]. Figure 2 clearly shows that the Mongolian cyclone was located in Mongolia at 500 hPa on March 14 (E1) and March 27 (E3), although the cyclone's central intensity and influence were stronger in E1. In the outer area of the cyclone, the wind speed reached 40 m/s, while a straight westerly flow dominated most of northern China. Figure 3a shows that these conditions corresponded well to the dust density distribution on March 27, and there was good consistency between the center of the cyclone and the dust intensity. At sea level, the low-pressure center occurred near northeast China on March 14 and 27, where there was a cold front. The dense isobars and the large pressure gradient before and after the cold front produced gale conditions. According to Lei and Wang [59], cold-front-induced dust storms are characterized by rapid and strong dust emissions. Conducive to dust emission was the confluence of warm and cold air, which rendered the lower atmosphere unstable. However, in the E3 event, the temperature difference between the cold and warm air was less significant than in E1. According to Table 2, which quantitatively gives the maximum wind speed of the dust emission stage, the smaller maximum horizontal wind speed in E3 was reasonable. Besides, as shown in Figure 2b, E2 had no typical synoptic system of dust emission unlike the strong dust storms. However, the configuration of the temperature and pressure fields between the wind speed in E1 was above 50 × 10 −3 hPa/s, greater than E3. Furthermore, near the su at 45°N, the strong convergence lasted longer in E1. Therefore, a larger amount of emission in E1 was predictable. Additionally, there was a much less intense momen exchange in the vertical direction of E2, and the updraft extended only to about 800 Consequently, the stronger the updraft, the more dust particles can be carried int atmosphere and transported downwind, which indicates the significant influence o vertical dynamic condition in the dust source on the near-surface dust concentr downwind.   at 45°N, the strong convergence lasted longer in E1. Therefore, a larger amount of emission in E1 was predictable. Additionally, there was a much less intense mome exchange in the vertical direction of E2, and the updraft extended only to about 800 Consequently, the stronger the updraft, the more dust particles can be carried int atmosphere and transported downwind, which indicates the significant influence o vertical dynamic condition in the dust source on the near-surface dust concentr downwind.    1 Airflow with the center vertical wind speed greater than 20 × 10 −3 hPa/s is considered as a strong downdraft. 2 The unit of wind speed: m/s. 3 The unit of wind speed: 10 −3 hPa/s. 4 The unit of vertical divergence: 10 −5 /s.
Generally, vertical motions are closely related to the amount of dust in the source region. Figure 4 presents the typical distribution of the zonal divergence and vertical velocity profiles in the three dust events. The clear trend of divergence-convergencedivergence from the ground to the upper air with the convergence center uplifting in the longitudinal direction is evident in Figure 4a,c. This pattern was beneficial to the formation of alternating downdrafts and updrafts, which resulted in multiple secondary circulations that further accelerated the formation of the surface gale. However, there was still some difference between the two events. As shown in Table 2, the maximum value of vertical wind speed in E1 was above 50 × 10 −3 hPa/s, greater than E3. Furthermore, near the surface at 45 • N, the strong convergence lasted longer in E1. Therefore, a larger amount of dust emission in E1 was predictable. Additionally, there was a much less intense momentum exchange in the vertical direction of E2, and the updraft extended only to about 800 hPa. Consequently, the stronger the updraft, the more dust particles can be carried into the atmosphere and transported downwind, which indicates the significant influence of the vertical dynamic condition in the dust source on the near-surface dust concentration downwind.

Long-Range Transport Stage
According to the 24 h backward trajectories in Figure 5, all three dust events originated in Mongolia at different altitudes and passed over the Gobi Desert, which corresponds to the dust product of the FY-4A satellite (Figure 3b). Over time, the center of the dust intensity moved in a southeasterly trajectory. Additionally, the trace data from different altitudes indicated that horizontal advection was the main transport mechanism for the dust in E2, and little vertical exchange occurred. However, the vertical movement was essential for the development of this dust event. As described in Section 3.2, the updraft in both E1 and E3 could reach 500 hPa, whereas it could only reach 850 hPa in E2. Figure 6 illustrates the distribution of geopotential height (500 hPa on March 15 and 28 and 850 hPa on March 21), indicating dust pathways. During E1 and E3, a strong northwesterly airflow dominated the Beijing-Tianjin-Hebei region; here, the center of the Mongolian cyclone at 500 hPa moved toward northeast China, and the denser isobars at 500 hPa in E1 were conducive to the long-range dust transportation. However, limited by the weak vertical movement of the system, the dust was only carried to 850 hPa, where the 850 hPa northwesterly wind transported it in E2.

Regional Circulation Situation
Under the domination of the downdraft, the dust transported to northern China w entrained to the lower atmosphere. Figure 7 presents the typical downdraft situation the dust events. It can be seen clearly that there was a strong downdraft center over 40

Regional Circulation Situation
Under the domination of the downdraft, the dust transported to northern China w entrained to the lower atmosphere. Figure 7 presents the typical downdraft situation the dust events. It can be seen clearly that there was a strong downdraft center over 40 116°E. As Figure 7a shows, at 10:00 (LT) on March 15, when the surface PM10 concentrat

Regional Circulation Situation
Under the domination of the downdraft, the dust transported to northern China was entrained to the lower atmosphere. Figure 7 presents the typical downdraft situation of the dust events. It can be seen clearly that there was a strong downdraft center over 40 • N, 116 • E. As Figure 7a shows, at 10:00 (LT) on March 15, when the surface PM 10 concentration peaked, the downdraft peaked at 28 × 10 −3 hPa/s at 850 hPa in E1. This was consistent with the presence of the low-level jet at 850 hPa (Figure 7b). The positive shear vorticity region in the left front of the low-level jet was conducive to convective activities [60]. The strong downdraft, with the maximum of 28.4 × 10 −3 hPa/s and the long duration reaching 16 h (Table 2), in E1 was attributable to the divergence-convergence-divergence trend in the vertical direction. Under such conditions, the stability of the atmospheric state promoted the endurance of the dust storm. In contrast, Figure 7d shows the vertical situation at 12:00 (LT) on March 28, when the surface PM 10 concentration comes to a peak. The maximum downward speed was only 14 × 10 −3 hPa/s, and the strong downdraft only lasted for 6 h. The shallower and moderate downdraft in E2, which is obvious in Figure 7c, prevented the complete delivery of the transported dust to the surface.

Lidar Continuous Observation
POLIPHON, referred to in Section 2.1, was applied to the polarization lidar data to study the vertical distribution of dust aerosols during the outbreak stage over Beijing. A cloud screening scheme was operated in advance with reference to Shimizu et al. [52]. Figure 8 reveals that, before the dust outbreaks of E1 and E3, the depolarization ratios The moving speed of the synoptic systems, i.e., Mongolian cyclone and high/cold surface pressure, determined the duration of the dust outbreak, which corresponds with the conclusion of Lei and Wang [59] that dust storms triggered by the Mongolian cyclone last the longest. In both E1 and E3, the Mongolian cyclone persisted over northeast China for over 24 h, leading to the continuous control of the northwesterly airflow in the Beijing-Tianjin-Hebei region (Figures S2 and S3). Under such conditions, the larger amount of dust from Mongolia, as discussed in Section 3.2, promoted the continuation of E1. However, the strong wind was also instrumental in dust removal, resulting in rapid decreases in nearsurface PM 10 concentration (as in E3). Due to the weaker suppression of the Mongolian cyclone in E3, as shown in Figure S3, the cold air circulated by the high/cold surface pressure reached northern China earlier. Conversely, the slower movement of the stronger Mongolian cyclone in E1 hindered the movement of the high/cold surface pressure to a certain extent.

Lidar Continuous Observation
POLIPHON, referred to in Section 2.1, was applied to the polarization lidar data to study the vertical distribution of dust aerosols during the outbreak stage over Beijing. A cloud screening scheme was operated in advance with reference to Shimizu et al. [52]. Figure 8 reveals that, before the dust outbreaks of E1 and E3, the depolarization ratios above 1 km were close to 0.3 (as in the white box in Figure 8a), indicating the airborne dust. Unfortunately, due to cloud interference, the statistics for March 14 do not include data above 1 km; however, on March 27, the airborne dust was obvious, which reflected the long-distance transport mechanism before the dust outbreak.  As described in Section 3.4.1, the dust events commenced once the airborne dust reached the ground. There was high temporal consistency between the downdraft occurrence and dust touchdown. Clearly, in March of 2021, three pure dust events occurred over Beijing. The circumstances dominated by spherical aerosols agreed well with the conditions under which the ratio of PM 2.5 to PM 10 (Figure 1d) was greater than 0.75, which confirmed the applicability of the depolarization ratio in dust identification. The backflow process occurred to differing degrees during all three dust events and was closely related to the change in wind direction. Figure 9c presents the profiles of dust mass concentration. The mass concentration was obtained from the dust extinction coefficients using an extinction-to-mass conversion factor of 0.78 ± 0.14 × 10 −12 Mm and a dust density of 2.6 g/cm 3 [61]. The conversion factor used in this study was obtained from AERONET, as described in detail in the Supplementary Materials. Limited by the incomplete overlap zone, the dust mass concentration at 150 m was used to compare with the surface PM 10 concentration, in order to verify the effectiveness of the conversion factor. The comparison is presented in Figure 10. Generally speaking, the dust concentration from lidar was in good agreement with the surface observation, and the correlation coefficient of the two was about 0.8. Significantly, the results from lidar corresponded to the PM 10 concentration well when the dust broke out, and lidar was good at picking up the peaks of dust concentration. However, there was still an overestimation with the normalized mean bias of 0.76, which is an index to measure the degree of deviation [62]. Actually, the mismatch of target height is the primary error. Since the dust was transported from top to bottom, a higher dust concentration at 150 m is normal when there is uneven mixing in the boundary layer. In strong dust storms, a large amount of dust is transported to the surface by a strong downdraft, reducing the difference between ground and low level.

Conclusions
Mineral dust is of great importance to climate change, air quality, and human h The near-surface dust mass concentration directly determines the intensity of the event to some extent, because its rating is divided according to the ground visibil this study, multisource data, including the reanalysis data and remote sensing data, used to compare the three dust events that occurred in the March of 2021 over Beijing plane-to-point analysis reveals the effects of the atmospheric vertical structure on surface dust concentration.
In E1 and E3, the eastward movement of the Mongolian cyclone coupled high/cold surface pressure caused significant amounts of dust in the source region lifted and transported to northern China via a northwesterly airflow at different alti before being carried back to the ground by a downdraft. The stronger Mongolian cy the slower high/cold surface pressure, and the occurrence of the low-level jet ma more intense, more extensive, and longer-lasting than E3. Additionally, the appro configuration of the temperature and pressure fields at high and low altitudes can generate an airborne dust process, as in E2. However, the dust emission of E2 was in to E1 and E3 due to its close relationship with the vertical dynamic conditions in the s region, which contributed to the low dust concentration near the surface in E2 to degree.
Under the domination of the downdraft, the dust transported to northern Chin entrained to the boundary layer. At this point, the strength of the downdraft direc fected the amount of dust reaching the ground. In E1, the maximum dust concent was recorded at about 200 m due to the continuous and high-intensity downdraft. A no dust aerosols were recorded above 1.5 km; however, the maximum height of the mass concentration in E3 occurred at about 800 m due to the transient downdraft hough E3 demonstrated a higher maximum value, the average near-surface dust concentration in E1 was almost double that in E3. The transient subsidence move It can be determined from Figure 9 that there was a significant difference in the vertical distribution of the dust, which reflects the average state and the standard deviation of the dust outbreak stage during different events. As described in Section 3.4.1, the continuous and high-intensity downdraft in E1 produced a maximum dust concentration at about 200 m. Almost no dust aerosols were recorded above 1.5 km, while the transient downdraft in E3 caused a dust mass concentration at a maximum height of about 800 m. The higher depolarization ratio of E3 above 600 m also confirmed the dust in the upper atmosphere. Although with a higher maximum value, the average near-surface dust mass concentration in E1 was almost double that of E3. In E1, the downdraft was powerful enough to transport the dust to the ground, whereas the weaker and transient descending movement suspended the dust aloft in E3, weakening its effect on surface visibility. The weak vertical motion of the E2 event acted on the airborne dust to produce a maximum concentration of only 1000 µg/m 3 (due to the lower dust emissions, as analyzed in Section 3.2). In E1, E2, and E3, the near-surface average dust mass concentration was about 4000, 300, and 2400 µg/m 3 , respectively. These values correspond well with the surface observations of particle concentration, which confirms the effectiveness of lidar in monitoring dust events. Consequently, it is evident that the strength of the downdraft directly influences the dust concentration on the ground. It should be noted that, in all three events, the maximum depolarization ratio occurred near the ground, which was attributable to the easier sedimentation of irregular-shaped large particles.

Conclusions
Mineral dust is of great importance to climate change, air quality, and human health. The near-surface dust mass concentration directly determines the intensity of the dust event to some extent, because its rating is divided according to the ground visibility. In this study, multisource data, including the reanalysis data and remote sensing data, were used to compare the three dust events that occurred in the March of 2021 over Beijing. The plane-to-point analysis reveals the effects of the atmospheric vertical structure on near-surface dust concentration.
In E1 and E3, the eastward movement of the Mongolian cyclone coupled with high/cold surface pressure caused significant amounts of dust in the source region to be lifted and transported to northern China via a northwesterly airflow at different alti-tudes before being carried back to the ground by a downdraft. The stronger Mongolian cyclone, the slower high/cold surface pressure, and the occurrence of the low-level jet made E1 more intense, more extensive, and longer-lasting than E3. Additionally, the appropriate configuration of the temperature and pressure fields at high and low altitudes can also generate an airborne dust process, as in E2. However, the dust emission of E2 was inferior to E1 and E3 due to its close relationship with the vertical dynamic conditions in the source region, which contributed to the low dust concentration near the surface in E2 to some degree.
Under the domination of the downdraft, the dust transported to northern China was entrained to the boundary layer. At this point, the strength of the downdraft directly affected the amount of dust reaching the ground. In E1, the maximum dust concentration was recorded at about 200 m due to the continuous and high-intensity downdraft. Almost no dust aerosols were recorded above 1.5 km; however, the maximum height of the dust mass concentration in E3 occurred at about 800 m due to the transient downdraft. Although E3 demonstrated a higher maximum value, the average near-surface dust mass concentration in E1 was almost double that in E3. The transient subsidence movement weakened its effect on surface visibility.
Overall, the large dust emission resulted from active updrafts in the source region, and the lengthy strong downdrafts led to the ultrahigh particle concentration near the surface. Additionally, this study confirmed the effectiveness of using lidar to monitor dust storms. There was high temporal consistency between the occurrence of downdrafts and the touchdown of dust according to vertical circulations and lidar observations. In fact, focusing solely on surface dust is not conducive to a comprehensive understanding of dust storms; the consideration of movements in the vertical direction is still of great significance.
Supplementary Materials: The following are available online at https://www.mdpi.com/article/ 10.3390/rs13183580/s1, Text S1: The description of the conversion factor calculation over Beijing; Figure S1: Relationship between AOD and total volume concentration. The conversion factor c d , correlation coefficient, and effective sample size are also presented. Each point denotes a pair of values for the pure dust moment; Figure   Data Availability Statement: The data will be shared by the authors upon reasonable request.