Detection of a Dust Storm in 2020 by a Multi-Observation Platform over the Northwest China

: Dust storms have occurred frequently in northwest China and can dramatically reduce visibility and exacerbate air quality in downwind regions through long-range transport. In order to study the distribution characteristics of dust particles sizes, structures and concentrations in the process of dust storm, especially for the vertical distributions, the multi-observation platform composed of six Lidars and nine aerosol analytical instruments is ﬁrst used to detect a severe dust storm event, which occurred in Northwest China on 3 May 2020. As a strong weather system process, the dust storm has achieved high intensity and wide range. When the intensity of a dust storm is at its strongest, the ratios of PM 2.5 (particulate matter with diameter < 2.5 µ m) and PM 10 (particulate matter with diameter < 10 µ m) (PM 2.5 /PM 10 ) in cities examined were less than 0.2 and the extinction coefﬁcients became greater than 1 km − 1 based on Lidar observations. In addition, the growth rates of PM 2.5 were higher than that of PM 10 . The dust particles mainly concentrated at heights of 2 km, after being transported about 200–300 km, vertical height increased by 1–2 km. Meanwhile, the dust concentration decreased markedly. Furthermore, the depolarization ratio showed that dust in the Tengger Desert was dominated by spherical particles. The linear relationships between 532 nm extinction coefﬁcient and the concentration of PM 2.5 and PM 10 were found ﬁrstly and their R 2 were 0.706 to 0.987. Our results could give more information for the physical schemes to simulate dust storms in speciﬁc models, which could improve the forecast of dust storms.


Introduction
In recent decades, dust storms have occurred frequently, along with climate change, on a global scale [1,2]. Global dust storms mainly come from the Sahara desert, the Middle East, Central Asia and East Asia [3][4][5]. With atmospheric circulation, dust particles mixed with bacteria, viruses, minerals and other chemical components are transported thousands of kilometers to downwind regions [6][7][8], which may affect the climate [9][10][11][12][13], environment [8,14], biology [15], and human activities and health [16,17]. Therefore, dust event monitoring is not only providing great help for dust forecast, early warning and ecological environment impact assessment, but also has a profound and lasting impact on global climate change and human life [13]. domination of coarse particles in the micron size range. Using an instrumented balloon and a ground-based Raman Lidar over Tsukuba, Japan, the influence of Asian mineral dust on an ice cloud formation in the upper troposphere was investigated, and the results pointed that the predominance of nonspherical particles were in the range of 6 to 12 km [34].
In order to make up for the ground monitoring and understand the temporal and spatial evolution characteristics of particulate matter, the relationship of mass concentration of PM 2.5 and extinction coefficient is studied. Also, the distribution and transport of fine particles by using mobile Lidar was discussed by Lv L.H. et al. [35], and they found that the southwestern Beijing was the main dust transport pathway. Xiang Y. et al. [36] used linear and exponential models to retrieve PM 2.5 concentration based on extinction coefficient from 355 nm Lidar, and the results showed that the PM 2.5 concentration and the measured values of the two models had a strong correlation coefficient. The surface PM mass distribution was obtained by synergy usage of satellite and Lidar measurements [37], and the results showed that the correlation coefficients between the estimated aerosol extinction coefficients and the surface PM mass were 0.57-0.86 for PM 2.5 and 0.59-0.78 for PM 10 , respectively.
Nevertheless, most of the above studies were based on a single Lidar, and a single Lidar is difficult to reveal the source, transmission and development trend of sand and dust. On the other hand, the fitting formula obtained by a single Lidar does not necessarily apply to all dust processes in all regions. Therefore, a multi-observation platform is need to research the dust transport processes. In order to obtain the temporal and spatial distribution information of dust, especially the vertical distribution, and solve the problem of single dust monitoring, it is useful to establish the multi-observation platform including six Lidars and nine aerosol analytical instruments. In addition, the relationship between PM 2.5 mass concentration and extinction coefficient was studied mostly; however, few studies researched the relationship between PM 10 mass concentration and extinction coefficient.
This paper analyzed the occurrence and development process of a severe dust event in Northwest China on 3 May 2020. It revealed the evolution characteristics of particles in the dust event, especially the dust vertical characteristics. Moreover, it estimated the particles concentration in vertical through fitting the extinction coefficient with particles concentration.
The data and methodology are introduced briefly in Section 2. The cause, transmission and development of the dust event and the extinction coefficient fitting the particulate concentration are shown in Section 3. Section 4 gives the conclusion and possible further improvements to this study.

Instrumentation
The Lidars (Light Detection and Ranging) in the study are all dual wavelength three channel Lidar, the main performance are shown in Table 1. The laser emits pulses of a specific wavelength (355 nm, 532 nm) and enters the atmosphere after collimation and expansion. The particles in the atmosphere produce the Mie scattering [38,39]. The scattering light with the direction of 180 • (backscattering) is received by the telescope system, and is divided into three channels, which are 355 nm, 532 nm parallel and 532 nm vertical light. The echo of the three channels are detected by the detection system, and then it is retrieved to the extinction coefficient and depolarization ratio to study the atmosphere. The backscattering echo power P(r) of atmospheric molecules and aerosol particles at distance r received by Lidar can be expressed as follows [40]: where P t is the laser emission power (W), K is the Lidar system constant (W·km 3 ·Sr), β a (r) and β m (r) are the backscattering coefficients (km −1 ·Sr −1 ) of aerosol particles and atmospheric molecules at distance r, respectively. α a (r) and α m (r) are the extinction coefficients (km −1 ) of aerosol particles and atmospheric molecules at distance r, respectively. The most frequently used retrieval methods for the extinction coefficient α(r) are Collis method, Klett method and Fernald method [41]. Most studies [35,36,42,43] have proved that the retrieval method of extinction coefficient by Fernald is not only mature and stable, but also has the smallest error. The formula for the aerosol extinction coefficient can be written as [44]: where S a = α a (r)/β a (r), S a depends on the incident laser wavelength, the size spectrum distribution of atmospheric aerosol particles and refractive index, and the numerical value is generally between 0~90. It is assumed that they are constants, which means that the size spectrum and chemical composition of atmospheric aerosol particles do not change with the altitude, and the extinction and scattering characteristics change only due to the change of their number density. For 532 nm wavelength, S a = 50 [45,46]. S m = α m (r)/β m (r), according to the Rayleigh scatter theory, S m is generally taken as 8π/3 [47].

Retrieval Method for the Depolarization Ratio
The depolarization ratio δ(r) of polarization Lidar can be expressed as follows [47]: where P rs (r) and P rp (r) are the backscattering echo power of the vertical and parallel channel at distance r, respectively. k s and K p are the gain constant ratio of the vertical and parallel channel, respectively. k = k p /k s . The depolarization ratio of atmospheric molecules is very small with the value of 0.0297 [48]. The depolarization ratio mainly comes from the nonspherical particles [47], and the dust event is mainly nonspherical particles. Due to the extinction coefficient reflecting the particle mass concentration, some researchers [35,36,49] have studied their relationship. In this study, the formula M p = aα + c is used for the PM 2.5 and PM 10 mass concentration. M p = aα b + c is used for the PM 10 -PM 2.5 mass concentration. M p is the particle mass concentration. α is the extinction coefficient. a, b, and c are all specific coefficients.

Lidar Network
Based on the dust transmission and the characteristics of dust particles in different regions, Gansu Environmental Monitoring Center of China has built a Lidar network in Gansu Province in 2020. In the dust storm event, six Lidars, which captured the process, were selected for further study. The six Lidars distribution are shown in Figure 1 and their coordinates are shown in Table 2. Due to the extinction coefficient reflecting the particle mass concentration, some researchers [35,36,49] have studied their relationship. In this study, the formula Mp = aα + c is used for the PM2.5 and PM10 mass concentration. Mp = aα b + c is used for the PM10-PM2.5 mass concentration. Mp is the particle mass concentration. α is the extinction coefficient. a, b, and c are all specific coefficients.

Lidar Network
Based on the dust transmission and the characteristics of dust particles in different regions, Gansu Environmental Monitoring Center of China has built a Lidar network in Gansu Province in 2020. In the dust storm event, six Lidars, which captured the process, were selected for further study. The six Lidars distribution are shown in Figure 1 and their coordinates are shown in Table 2.

Analysis of the Concentration of Particles Monitored on the Ground
A severe dust event occurred in Northwest China on 3 May 2020. In order to understand the transportation of the dust, Figure 2 shows the hourly variation of PM 10 and PM 2.5 mass concentrations in the cities examined. The red and blue dotted lines represent the PM 10 and PM 2.5 , respectively, and the green shadows represent the values of PM 2.5 /PM 10 . It can be seen that Bayan Nur which neared Ulan Buh Desert was affected firstly by the dust event, and the PM 10 concentration suddenly increased to 4171 µg·m −3 at 00:00 on 3 May. Then, the dust moved southwest under the influence of a surface northeast wind. At 04:00, the dust event reached Jiayuguan with a peak surface PM 10 concentration of 1224 µg·m −3 . The maximum PM 10 concentration in Ordos and Yinchuan occurred at 06:00 with a value of 657 µg·m −3 and 1741 µg·m −3 , respectively. Different from other cities, Yumen was affected by the dust storm in a long term from 09:00 to 18:00, and the maximum concentration of PM 10 was about 2215 µg·m −3 . When the dust was transported to Aksay, its intensity was obviously weakened with the maximum PM 10 concentration of 317 µg·m −3 . The concentration of PM 10 reached 4053 µg·m −3 at 08:00 in Jinchang and 4591 µg·m −3 at 09:00 in Wuwei. The dust intensity gradually weakened when it was transported to Baiyin and the peak concentration of PM 10 was 612 µg·m −3 . In addition, except for Yumen and Aksay, the PM 10 concentration in other cities reached the peak within two hours after being affected by the dust. Jinchang and Wuwei had been affected severely, and their hourly PM 10 concentration exceeded 4000 µg·m −3 , which were about 27 times of the daily average secondary standard of PM 10 (150 µg·m −3 ) [50], according to China's air pollutant concentration limit standard.
Due to the variation of PM 2.5 concentration being consistent with that of PM 10 in this dust event, the peak values of PM 2.5 and PM 10 concentration occured at the same time. The corresponding maximum concentrations of PM 2.5 in Bayan Nur, Ordos, Yinchuan, Yumen, Jinchang and Wuwei were 584 µg·m −3 , 89 µg·m −3 , 271 µg·m −3 , 261 µg·m −3 , 720 µg·m −3 and 750 µg·m −3 , respectively. Also, the corresponding values of PM 2.5 /PM 10 were 0.14, 0.14 0.16, 0.12, 0.18 and 0.16, respectively. The smaller values of PM 2.5 /PM 10 (all were less than 0.2) indicated that the proportions of fine particles in the dust event were very small. The peak concentration of PM 2.5 in Baiyin was 140 µg·m −3 and the value of PM 2.5 / PM 10 was 0.23. The proportion of fine particles was significantly larger than that in upstream cities, which indicated that the proportion of fine particles was gradually increasing when the dust was transported to downstream areas. It is worth noting that the PM 2.5 concentration reached 530 µg·m −3 , and the value of PM 2.5 /PM 10 was 0.43 in Jiayuguan at 04:00, which indicated that the proportion of the fine particles was similar to the coarse particles. Moreover, during the study period, the values of PM 2.5 /PM 10 of Jiayuguan were significantly higher than that in other cities, which indicated that Jiayuguan had more local dust than other cities. Due to the variation of PM2.5 concentration being consistent with that of PM10 in dust event, the peak values of PM2.5 and PM10 concentration occured at the same time. T corresponding maximum concentrations of PM2.5 in Bayan Nur, Ordos, Yinchuan, Yum Jinchang and Wuwei were 584 µg·m −3 , 89 µg·m −3 , 271 µg·m −3 , 261 µg·m −3 , 720 µg·m −3 a 750 µg·m −3 , respectively. Also, the corresponding values of PM2.5/PM10 were 0.14, 0.14 0 0.12, 0.18 and 0.16, respectively. The smaller values of PM2.5/PM10 (all were less than indicated that the proportions of fine particles in the dust event were very small. The p concentration of PM2.5 in Baiyin was 140 µg·m −3 and the value of PM2.5/ PM10 was 0.23. proportion of fine particles was significantly larger than that in upstream cities, wh indicated that the proportion of fine particles was gradually increasing when the dust w transported to downstream areas. It is worth noting that the PM2.5 concentration reach 530 µg·m −3 , and the value of PM2.5/PM10 was 0.43 in Jiayuguan at 04:00, which indica that the proportion of the fine particles was similar to the coarse particles. Moreover, d ing the study period, the values of PM2.5/PM10 of Jiayuguan were significantly higher th that in other cities, which indicated that Jiayuguan had more local dust than other citi Significantly, the dust storm event disappeared in the early morning on 4 May (F ure 2). In order to understand the impact of the dust over the whole China, the daily erage concentrations of PM10 on 4 May are analyzed (as shown in Figure 3). The dust sto event had a significant impact on most cities in Northern China. Although the durat was short, the intensity was very strong, especially in Inner Mongolia, Ningxia and Ga Province. The daily average PM10 concentrations of some stations were close to 1 µg·m −3 , which is 10 times of the daily average secondary standard of PM10. Significantly, the dust storm event disappeared in the early morning on 4 May ( Figure 2). In order to understand the impact of the dust over the whole China, the daily average concentrations of PM 10 on 4 May are analyzed (as shown in Figure 3). The dust storm event had a significant impact on most cities in Northern China. Although the duration was short, the intensity was very strong, especially in Inner Mongolia, Ningxia and Gansu Province. The daily average PM 10 concentrations of some stations were close to 1500 µg·m −3 , which is 10 times of the daily average secondary standard of PM 10 .

Dynamic Conditions
In order to analyze the dynamic conditions of the dust, Figure 4a,b show the 500 hPa synoptic situation at 20:00 on the 2nd and 08:00 on the 3rd, respectively. The blue solid lines represent the height, and the red dotted lines represent the temperature. Significantly, Remote Sens. 2021, 13, 1056 8 of 18 there were two troughs and one ridge at 50 • N. The cold trough was over the west of Baikal Lake and an upper-level trough was over Mongolia. Inner Mongolia was located in the trough area. The temperature trough lagged behind the height trough, and the wind fields were vertical to the temperature fields. That is to say, the cold advection behind the trough was strong. With the continuous accumulation of cold air, the height trough would be further strengthened and became the guiding system for the ground cold air movement. At 08:00 on the 3rd, the cold trough over west of Baikal Lake moved eastward to Baikal Lake. Meanwhile, its tail was located in the northwest of Hexi, which belonged to Northwest Gansu. In addition, at 20:00 on the 2nd, the high level of westerly jet stream appeared in Inner Mongolia and Gansu, and strengthened at 08:00 on the 3rd. The sandstorm was likely to occur at the right rear, where the high level of westerly jet entered. The relationship between the upper jet stream and the dust was consistent with the result of Cheng et al. [51]. According to the analysis of the surface synoptic situation, Hexi and Inner Mong were located in the front of the surface high pressure at 20:00 on the 2nd (Figure 4c), the cold front at the junction of Mongolia and Inner Mongolia moved slowly to the sou east. The strong wind after the cold front caused sand in the Badain Jaran Desert Tengger Desert to rise, and then the dust particles were transported to the southeast 08:00 on the 3rd (Figure 4d), the cold front arrived near Jinchang and Wuwei, and brou the dust storm event. With the cold front pressing southward, the gale behind the fr was mainly located in the Midwest of Inner Mongolia and Hexi. Since then, the high l tude vortex weakened, and the northwest jet stream continuously transported cold ai the southeast.
The northern section of the high trough had reached the mountain area in the no west of Greater Khingan Range in China, and the southern part was still in the middl Hexi. At 14:00 on the 3rd, the cold front moved out of Gansu, the southward movem of the cold center and high pressure center weaked. In combination with the surface c center and high pressure center, the direction of cold front gradually turned to the e west. Due to the dominant airflow of upper-level being mainly the westerly jet stream the bottom of the upper-level trough, the ground cold front on the 3rd moved ma eastward. Meanwhile, the dust particles gradually moved to the downstream cities. H ever, due to the maintenance of a weak east wind, the settlement was relatively slow. center of cold and high pressure moved eastward obviously, and Jinchang and Wu According to the analysis of the surface synoptic situation, Hexi and Inner Mongolia were located in the front of the surface high pressure at 20:00 on the 2nd (Figure 4c), and the cold front at the junction of Mongolia and Inner Mongolia moved slowly to the southeast. The strong wind after the cold front caused sand in the Badain Jaran Desert and Tengger Desert to rise, and then the dust particles were transported to the southeast. At 08:00 on the 3rd (Figure 4d), the cold front arrived near Jinchang and Wuwei, and brought the dust storm event. With the cold front pressing southward, the gale behind the front was mainly located in the Midwest of Inner Mongolia and Hexi. Since then, the high latitude vortex weakened, and the northwest jet stream continuously transported cold air to the southeast.
The northern section of the high trough had reached the mountain area in the northwest of Greater Khingan Range in China, and the southern part was still in the middle of Hexi. At 14:00 on the 3rd, the cold front moved out of Gansu, the southward movement of the cold center and high pressure center weaked. In combination with the surface cold center and high pressure center, the direction of cold front gradually turned to the east-west. Due to the dominant airflow of upper-level being mainly the westerly jet stream at the bottom of the upper-level trough, the ground cold front on the 3rd moved mainly eastward. Meanwhile, the dust particles gradually moved to the downstream cities. However, due to the maintenance of a weak east wind, the settlement was relatively slow. The center of cold and high pressure moved eastward obviously, and Jinchang and Wuwei were at the bottom of high pressure. The strong wind behind the front was northerly, bringing the dust particles to these areas.

Thermal Conditions
In order to further analyze the thermal conditions that caused the sandstorm, the daily average temperatures of the cities examined from 1 April to 10 May were analyzed in Figure 5. Obviously, the temperatures of cities examined continued to rise from 23 April to 2 May, and the rising range was about 10 • C. After the occurrence of sand dust and the invasion of cold air, the temperature of each city gradually decreased. In the early stage, the rising temperature and lack of precipitation led to the dust process. to 2 May, and the rising range was about 10 °C. After the occurrence of sand dust and the invasion of cold air, the temperature of each city gradually decreased. In the early stage, the rising temperature and lack of precipitation led to the dust process. Due to continuous heating over many days, the dust storm occurred over the desert regions. In addition, the cold air guided by high trough and high level of westerly jet stream gave rise to the dust emission, and more dust particles were brought into the atmosphere. In the atmosphere, the dust particles were transported westward firstly following the northeast wind, then transported southeastward with the surface cold front.

Extinction Coefficient and Depolarization Ratio
In order to understand the source, transmission, evolution trend and impact on urban air quality of the sandstorm further, six Lidars were used to study the temporal and spatial distribution characteristics of the sandstorm in the process of long-distance and cross-border transportation, especially in regards to the vertical distribution characteristics of dust. In this paper, only the retrieval products with wavelength of 532 nm are studied. Due to the Lidar blind area and the vertical development height of dust, the extinction coefficient and depolarization ratio of 0.15-5 km height retrieved by Lidars are shown in Figure 6. Due to continuous heating over many days, the dust storm occurred over the desert regions. In addition, the cold air guided by high trough and high level of westerly jet stream gave rise to the dust emission, and more dust particles were brought into the atmosphere. In the atmosphere, the dust particles were transported westward firstly following the northeast wind, then transported southeastward with the surface cold front.

Extinction Coefficient and Depolarization Ratio
In order to understand the source, transmission, evolution trend and impact on urban air quality of the sandstorm further, six Lidars were used to study the temporal and spatial distribution characteristics of the sandstorm in the process of long-distance and crossborder transportation, especially in regards to the vertical distribution characteristics of dust. In this paper, only the retrieval products with wavelength of 532 nm are studied. Due to the Lidar blind area and the vertical development height of dust, the extinction coefficient and depolarization ratio of 0.15-5 km height retrieved by Lidars are shown in Figure 6. Extinction coefficient of Jiayuguan Lidar (Figure 6 (a)) show that the dust storm gan at 02:00 on the 3rd, and the lower layer extinction coefficient suddenly increased fro less than 0.1 km −1 to more than 0.6 km −1 . The extinction coefficient was greater than 1 km from 02:00 to 04:00. There was a pollution belt at the height of 1-2 km from 02:00 to 12 on the 3rd, in which the extinction coefficient was obviously small. In addition, there w a fault phenomenon with the pollution lower layer, which indicates that the dust in t height was mainly external transmission. At the duration, the sudden increase of extin tion coefficient at the height of 1 km indicated that the dust was mainly concentrated in km height. The dust mainly concentrated in the height of 1-2 km from 18:00 on the 3rd 00:00 on the 4th, and then subsided and dissipated gradually. The extinction coefficie was still closed to 1 km −1 at 1.0 km height from 00:00 to 06:00 on the 4th, and it led to t ground particulate concentration increasing again in the future. Corresponding to Figu 2, it is found that the ground particulate concentration in Jiayuguan showed a slight u ward trend at the same time, and the PM10 concentration increased to 228 µ g·m −3 at 13:   Extinction coefficient of Jiayuguan Lidar (Figure 6a) show that the dust storm began at 02:00 on the 3rd, and the lower layer extinction coefficient suddenly increased from less than 0.1 km −1 to more than 0.6 km −1 . The extinction coefficient was greater than 1 km −1 from 02:00 to 04:00. There was a pollution belt at the height of 1-2 km from 02:00 to 12:00 on the 3rd, in which the extinction coefficient was obviously small. In addition, there was a fault phenomenon with the pollution lower layer, which indicates that the dust in the height was mainly external transmission. At the duration, the sudden increase of extinction coefficient at the height of 1 km indicated that the dust was mainly concentrated in 1 km height. The dust mainly concentrated in the height of 1-2 km from 18:00 on the 3rd to 00:00 on the 4th, and then subsided and dissipated gradually. The extinction coefficient was still closed to 1 km −1 at 1.0 km height from 00:00 to 06:00 on the 4th, and it led to the ground particulate concentration increasing again in the future. Corresponding to Figure 2, it is found that the ground particulate concentration in Jiayuguan showed a slight upward trend at the same time, and the PM 10 concentration increased to 228 µg·m −3 at 13:00. The extinction coefficient and depolarization ratio (Figure 6b) of Jiayuguan Lidar were lager, but the depolarization ratio did not change significantly with the increasing of extinction coefficient. It is attributed that the change of aerosol morphology was stable in the dust process.
The extinction coefficient of Yumen Lidar (Figure 6c) suddenly increased to 0.8 km −1 from the lower layer to 1.5 km height at about 06:00 on the 3rd, but the duration was very short. The dust mainly concentrated in the height of 0.6 km from 06:00 to 13:00 with the extinction coefficient in the range of 0.2-0.3 km −1 , and the corresponding ground PM 10 concentration was maintained at 500-1000 µg·m −3 . From 21:00 on the 3rd to 12:00 on the 4th, the lower layer extinction coefficient decreased obviously, but there was obvious dust pollution zone in the height of 0.5-1.5 km, which implied that the ground particulate concentration would increase in the future. Corresponding to Figure 2, the ground particulate concentration in Yumen showed a small upward trend, and PM 10 concentration increased again to 457 µg·m −3 at 13:00. Compared with the depolarization ratio of Yumen Lidar (Figure 6d), the results were similar to those obtained from Jiayuguan Lidar. When the extinction coefficient was large, the corresponding depolarization ratio was also large. However, with the increasing of extinction coefficient, the depolarization ratio did not change significantly.
When the dust was transported to Aksay, which is about 330 km away from Jiayuguan, the dust intensity had been obviously weakened, the extinction coefficient of Aksay (Figure 6e) was significantly lower than that of Jiayuguan and Yumen. Since 18:00 on the 3rd, the extinction coefficient slightly increased to about 0.2 km −1 , and the corresponding ground particulate concentration also increased, but this range was small. PM 10 concentration reached the maximum of 317 µg·m −3 at 00:00 on the 4th. The extinction coefficient of Aksay Lidar was low, but the development height of dust was high, which was close to 4 km. It is expected that the particulate concentration of ground would increase slightly in the future. Compared with Jiayuguan and Yumen Lidars, the biggest difference existed in the depolarization ratio (Figure 6f). The extinction coefficient of Aksay Lidar was not high, but the depolarization ratio was obviously higher than that of Jiayuguan and Yumen, indicating that the proportion of irregular particles in Aksay was significantly higher than that in Jiayuguan and Yumen.
Due to Jinchang and Wuwei being very close to each other and adjacent to the Tengger Desert, the detection results of the two Lidars were very close. The extinction coefficients (Figure 6g,i) of Jinchang Lidar and Wuwei Lidar at the height of 0.6 km altitude suddenly increased to about 1 km −1 at 05:00 and 06:00, respectively, which shown that dust aerosols were mainly concentrated in the height of 0.6 km. The high extinction coefficient of Jinchang lasted for about 7 h and that of Wuwei lasted for about 9 h. The maximum PM 10 hourly concentration was more than 4000 µg·m −3 in Wuwei, which was more serious than Jinchang. At this time, the Lidar extinction coefficient of Jinchang and Wuwei were very small above the height of 0.8 km, which indicated that the height of dust was not high. In other words, the dust concentration near the ground was very strong, but the height was not high, of which it was expected that the future dust weather intensity would be large but the duration would be short. After 18:00, the extinction coefficient of the two Lidars decreased significantly, and the corresponding particulate concentration of ground decreased significantly. Although the particle concentrations in Jinchang and Wuwei were very high and the corresponding extinction coefficients were large, their Lidar depolarization ratios were reduced significantly compared with Jiayuguan, Yumen and Aksay. This indicates that the particles in Jinchang and Wuwei were more spherical than those in Jiayuguan, Yumen and Aksay, especially in Jinchang.
With the cold air moving southeast, the intensity had been obviously weakened when the dust was transported to Baiyin (about 280 km away from Jinchang), the extinction coefficient (Figure 6k) slightly increased at 9:00-12:00 on the 3rd, ranging from 0.1 km −1 to 0.3 km −1 , and then the extinction coefficient decreased significantly. From 18:00 on the 3rd to 00:00 on the 4th, the dust mainly concentrated in the height of 1 km, and the extinction coefficient increased again to 0.2 km −1 . Compared with Jinchang, the height of dust increased by 2-3 km. The precipitation process occurred at 00:00 on the 4th. Due to the high extinction of water vapor, the extinction coefficient increased abruptly from near ground to 3 km height. The dust concentration decreased gradually by wet sedimentation. During the dust period, the depolarization ratio of Baiyin Lidar (Figure 6l) was about 0.23, which was close to Wuwei and larger than that of Jinchang.
Combined with the weather synoptic, the dust in Jiayuguan, Yumen and Aksay were affected mainly by Badain Jaran Desert, while the dust in Jinchang, Wuwei and Baiyin mainly came from adjacent Tengger Desert. According to the Lidar extinction coefficient and depolarization ratio, it was found that when the extinction coefficient was large, the depolarization ratio was large, but with the increase of extinction coefficient, the depolarization ratio did not change obviously. The high concentration of dust mainly concentrated within 1 km and the height of dust was about 2 km near the sand source. The height of dust was not high, resulting in a short distance to the downstream. The vertical development increased by 1-2 km with the horizontal transmission was about 200-300 km, and the concentration decreased significantly. The extinction coefficient of Aksay Lidar was significantly lower than that of Jiayuguan and Yumen, but the depolarization ratio was very high, while similarly, the extinction coefficient of Baiyin Lidar was significantly lower than that of Jinchang and Wuwei, but the depolarization ratio was very high. These indicated that irregular particles were easier to transmit downstream than spherical particles in the dust process. It is also possible that the proportion of irregular particles increased because spherical particles were easily made into sedimentation. The extinction coefficients of Jinchang and Wuwei Lidars were significantly higher than those of Jiayuguan and Yumen were, but the depolarization ratios were smaller obviously, which indicated that the dust particles in Tengger Desert were closer to sphere than those in Badain Jaran Desert were. Figure 7 shows the fitting curve of the hourly extinction coefficient of 532 nm at 150 m height with the ground PM 2.5 and PM 10 concentration. In order to reduce the influence of water vapor on the fitting effect [35], the data were removed when the relative humidity of Lidar station was greater than 80%. It can be seen that the change of extinction coefficients of all Lidars were very consistent with the change of particulate concentrations, which indicated that 532 nm at 150 m height can reflect the concentration of ground particles. The fitting results are shown in Table 3. The extinction coefficients of 532 nm with PM 2.5 and PM 10 concentration were larger and showed a good relationship. Due to the weak intensity of dust in Yumen, Aksay and Baiyin, the square of the correlation coefficient (R 2 ) were slightly inferior, ranging from 0.706 to 0.879. The R 2 of other stations were greater than 0.9. Further, two periods of 07:00-09:00 and 17:00-20:00 on the 3rd were selected to analyze the reason that why the R 2 of Yumen was slightly inferior. It was found that the extinction coefficients of the two periods were similar, but the particulate concentration from 07:00 to 09:00 was significantly lower than that from 17:00 to 20:00, especially for the difference of PM 10 concentration. Compared with the relative humidity of the two periods, the relative humidity from 07:00 to 09:00 was 37%-40%, and that from 17:00 to 20:00 was 24%-28%. This shows that the influence of water vapor on extinction coefficient cannot be ignored.  The linear relationship between 532 nm extinction coefficient and particulate concentration was used to explore the vertical change of particulate concentration further during the dust period. The fitting formula for the PM2.5, PM10 mass concentration in the linear model can be written as [35]: y = ax + b. When the extinction coefficient was the same, it is considered that the error of particulate concentrations within the acceptable range. Therefore, according to the fitting formula in Table 3, the particulate concentration at the vertical 26 of each Lidar station can be obtained, as shown in Figure 8. In Figure 8, the left column represents the retrieved PM2.5 concentration of each Lidar station, and the right column represents the retrieved PM10 concentration. The particulate in Jiayuguan (Figure 8 (a) and (b)) was mainly concentrated in the height of 2 km, in which it was mainly concentrated within 0.8 km from 00:00 to 12:00 on the 3rd day. The maximum concentration of PM2.5 and PM10 were about 500 µ g·m −3 and 1000 µ g·m −3 , respec- The linear relationship between 532 nm extinction coefficient and particulate concentration was used to explore the vertical change of particulate concentration further during the dust period. The fitting formula for the PM 2.5 , PM 10 mass concentration in the linear model can be written as [35]: y = ax + b. When the extinction coefficient was the same, it is considered that the error of particulate concentrations within the acceptable range. Therefore, according to the fitting formula in Table 3, the particulate concentration at the vertical 26 of each Lidar station can be obtained, as shown in Figure 8.

Retrieval of Vertical Distribution of Particulate Concentration
In Figure 8, the left column represents the retrieved PM 2.5 concentration of each Lidar station, and the right column represents the retrieved PM 10 concentration. The particulate in Jiayuguan (Figure 8a,b) was mainly concentrated in the height of 2 km, in which it was mainly concentrated within 0.8 km from 00:00 to 12:00 on the 3rd day. The maximum concentration of PM 2.5 and PM 10 were about 500 µg·m −3 and 1000 µg·m −3 , respectively. The particulate concentration in the height of 0.6-0.8 km decreased obviously with a concentration of 100-200 µg·m −3 . The PM 2.5 and PM 10 were lower than 100 µg·m −3 in the height of 1-1.4 km. At the same time, there was external dust at 1-2 km height, PM 2.5 and PM 10 concentrations were 100-150 µg·m −3 and 200-300 µg·m −3 , respectively. The particulate concentration was significantly lower than that of the low level, and the values of PM 2.5 /PM 10 were estimated to be 0.5. After that, the main pollution belt gradually diffused to the upper level, and the PM 2.5 concentration was 360 µg·m −3 and the PM 10 concentration was 1000 µg·m −3 . The values of PM 2.5 /PM 10 decreased from 0.5 to 0.36 when compared with the transport dust, indicating that the contribution of external dust to local PM 2.5 concentration was greater. The particulate concentration above 2 km was significantly reduced to about 100 µg·m −3 .
PM2.5/PM10 were estimated to be 0.5. After that, the main pollution belt gradually diffused to the upper level, and the PM2.5 concentration was 360 µ g·m −3 and the PM10 concentration was 1000 µ g·m −3 . The values of PM2.5/PM10 decreased from 0.5 to 0.36 when compared with the transport dust, indicating that the contribution of external dust to local PM2.5 concentration was greater. The particulate concentration above 2 km was significantly reduced to about 100 µ g·m −3 . The vertical development of dust in Yumen (Figure 8 (c), (d)) was close to Jiayuguan, and it was mainly concentrated within 0.4 km height from 06:00 to 14:00 on the 3rd, and then gradually lifted to 1.6 km. During this period, the concentration of PM2.5 was about 150 µ g·m −3 , and the concentration of PM10 varied greatly from 400-1000 µ g·m −3 . From 21:00 on the 3rd to 12:00 on the 4th, the particulate concentration within 0.4 km was very small,  Table 3 (from top to bottom: Jiayuguan (a,b), Yumen (c,d), Aksay (e,f), Jinchang (g,h), Wuwei (i,j), Baiyin (k,l)).
The vertical development of dust in Yumen (Figure 8c,d) was close to Jiayuguan, and it was mainly concentrated within 0.4 km height from 06:00 to 14:00 on the 3rd, and then gradually lifted to 1.6 km. During this period, the concentration of PM 2.5 was about 150 µg·m −3 , and the concentration of PM 10 varied greatly from 400-1000 µg·m −3 . From 21:00 on the 3rd to 12:00 on the 4th, the particulate concentration within 0.4 km was very small, the concentration of PM 2.5 was less than 60 µg·m −3 , and the concentration of PM 10 was less than 250 µg·m −3 , which is consistent with the analysis in Figure 2. However, there was an obvious dust pollution at the height of 0.5-1.5 km. The concentrations of PM 2.5 and PM 10 reaches the maximum, which were about 400 µg·m −3 and 1600 µg·m −3 , respectively. There was dust transported within 1.5-3 km height, which was significantly higher than that of Jiayuguan. The PM 2.5 concentration was about 90 µg·m −3 , and the