Investigation of the Vertical Distribution Characteristics and Microphysical Properties of Summer Mineral Dust Masses over the Taklimakan Desert Using an Unmanned Aerial Vehicle

Abstract

Investigating the vertical distribution of mineral dust masses and their microphysical properties is crucial for accurately assessing the climate effects of dust. However, there are limited studies related to relevant in situ observations over dust source areas. In this study, the near-surface vertical characteristics (within 500 m a.g.l) of dust mass concentrations in five size fractions (PMs: TSP, PM₁₀, PM₄, PM_2.5, and PM₁) were investigated using an unmanned aerial vehicle (UAV) in Tazhong (TZ) in the Taklimakan Desert (TD) in July 2021. To the best of our knowledge, the vertical profiles of particle number concentration (PNC), effective radius (R_eff), and volume concentration (Cv) were obtained for the first time by UAV over the TD. Four scenarios of clear sky, floating dust, blowing sand, and dust storm were selected based on the classification criteria for PMs. The PMs, PNC, R_eff, and Cv decreased with height for all scenarios. From clear-sky to dust-storm scenarios PMs, PNC, R_eff, and Cv in the column gradually increased. R_eff (Cv) increased from 1.15 μm (0.08 μm³/μm²) to 4.53 μm (0.74 μm³/μm²). The diurnal variations of PMs, PNC, and R_eff (Cv) revealed a unimodal pattern, with the peak occurring between 13:00 and 16:00, due to the evolution of wind speed and the atmospheric boundary layer in TZ. Unexpectedly, among the three postprecipitation scenarios (P1, P2, and P3), the PNC of P2 was smaller than those of P1 and P3. The R_eff (Cv) for P2 was similar to or greater than that for dust storms, which may be associated with moist dust particles on the ground surface being carried into the air by wind. These investigations add to our understanding of the mineral dust vertical characteristics over the source area, and provide a meaningful reference for colocated lidar inversion and dust simulations.

1. Introduction

Mineral dust is a major aerosol type in the Earth’s system and plays an important role in global climate, human health, and air quality. Dust particles regulate the radiation energy redistribution in the atmosphere by directly scattering and absorbing solar short- or terrestrial long-wave radiation [1,2]. However, the large temporal variability and characterization of the spatial distributions, aerosol mixtures, and absorptive properties make it difficult to evaluate aerosol direct radiative forcing [3]. Dust aerosols are one of the most common types of ice nuclei (IN) and can modify cloud macrophysical and microphysical properties in different ways [4]. As a result, cloud radiative properties and precipitation processes can be modulated by aerosol–cloud interactions [5]. Dust aerosols can also accelerate snow melting by altering the surface albedo of the snow and ice sheets [6,7,8]. From a public health perspective, mineral dust can transport pathogens and anthropogenic pollutants that trigger a variety of respiratory and cardiovascular diseases, as well as additional systemic pathologies [9,10].

Mineral dust’s vertical characteristics are essential to accurately evaluate its climate forcing and transboundary transportation to downstream areas [11,12,13]. Aerosol–radiation interactions mainly depend on dust particle size, aerosol vertical distribution, total optical depth, aerosol refractive index, temperature vertical profiles, and surface albedo. Among these parameters, dust vertical distribution is currently poorly characterized, resulting in large uncertainties when improving and validating models related to dust emission and transport, ultimately hindering effective evaluation of the impact of dust on climate processes [14,15]. In addition to its direct radiative effect, the indirect effect mainly depends on the particle number concentration (PNC) [16,17] and the relative vertical location of the cloud and aerosol layers [18,19]. In particular, a better characterization of the vertical distribution over dust source areas is required [15]. However, the existing models can effectively reproduce the onset and decline times of the dust process. The differences in the maximum near-surface dust concentrations, reflected by different models, are very extreme when they fail to accurately simulate the vertical distribution [20,21]. Desert areas are sparsely populated, making it difficult to obtain continuous vertical observations. This means that adequate constraints cannot be provided for the model simulations [22,23,24]. Thus, it is essential to better characterize the vertical distribution of dust aerosols in the atmosphere.

Usually, the dust vertical distribution is investigated using tethered balloons, meteorological towers, active remote sensing, and aircraft. Tethered balloons can carry a variety of instruments and are capable of reaching heights of 100–2000 m, with high resolution and lengthy duration [25]. However, it is more difficult to obtain continuous vertical profiles of atmospheric particles because of the low horizontal maneuverability of tethered balloons, the need for strong financial support, complex technical requirements, and balloon release frequency being limited by weather conditions [26,27]. Meteorological towers have long been used because of their high load capacity and stability; however, they are characterized by low vertical resolution, poor monitoring height, and low flexibility [27,28,29]. A space-based lidar (e.g., CALIOP) can capture the vertical structure of clouds and aerosols on a global scale. However, the spatiotemporal resolution is limited for a fixed location [30]. While ground-based lidar measurements range from hundreds of meters to tens of kilometers, providing information on particle distribution throughout the atmosphere, high spatiotemporal resolution, and high-precision extinction coefficient profiles, they are more restrictive when attempting to provide vertical resolution of aerosol microphysical properties [30,31,32]. Aircraft observations can overcome some of the shortcomings of the above methods by providing the microphysical, scattering, and absorption properties of aerosols in addition to the vertical distribution of these parameters. For example, [33] used aircraft and tethered balloons to investigate the vertical structure of the aerosol size distribution showing particles characterized by a large size within a dust plume during the Charmex campaign in 2013. However, aircraft require an airport to initiate frequent deployment of exhaust engines, which requires us to consider the limitations of the site [34,35,36,37,38]. Currently, the unique advantages attributed to unmanned aerial vehicles (UAVs), namely, having strong mobility and stability, and being exhaust-free, low-risk, and cost-effective, help detect lower tropospheric atmospheres and are considered one of the most suitable platforms for observing the vertical distribution of aerosols and measuring their physical and optical properties. UAV observations can be nicely coordinated with other observations, such as vertical data observed in the lower atmosphere, which can provide excellent constraints for the algorithm of lidar inversion of the microphysical properties of aerosols [39].

The Taklimakan Desert (TD), which is the largest dust source in the East Asian region, is located in the Tarim Basin of China and is surrounded by the Tianshan Mountains, Pamir Plateau, and Kunlun Mountains, which range over 5000 m above sea level. So far, there are only a few studies that have been conducted on the vertical distribution of dust over the TD. Ref [40] seeks to analyze the characteristics of the horizontal wind field and vertical velocity of a breaking dust weather in a desert hinterland and calculate the radar echo intensity and vertical distribution of a dust storm, blowing sand, and floating dust weather. Observational analysis of dust storm samples collected at seven vertical heights from an 80m high-throughput tower in the TD hinterland analyzed the vertical distribution of dust storm particle sizes and horizontal dust storm sediment fluxes near the surface [28]. Ref [41] explored the vertical distribution of dust with different particle size over the TD based on UAV observations, environmental particulate matter analyzer, and radiosonde. These previous studies provided us the preliminary insight on dust vertical distribution, but the information on the key microphysical parameters of the aerosol particles (e.g., effective radius (R_eff) and volume concentration (Cv)) in the vertical distributions is still extremely scarce under harsh experimental conditions. Moreover, the dust particles can be transported from Tarim Basin to Tibet Plateau in summer [42], while the understanding on the correlation between the temporal–spatial evolution of the dust layer in the Tarim Basin and the dust transported to the Tibetan Plateau is still poor, and the exploration of this issue requires sufficient observation on mineral dust vertical distribution over this area.

In order to understand the vertical distribution of mineral dust in the Tarim Basin and its relationship with dust transported to the Tibetan Plateau, an intensive field campaign was carried out in Tazhong (TZ) in July 2021. This study is the first part of our work frame and mainly clarifies the vertical distribution of near-surface mineral dust mass and microphysical properties with UAV observation. The remainder of this study is organized as follows. Section 2 introduces the experimental station, instrument information, and retrieval methods; Section 3 and Section 4 present the results and discussion, respectively; and Section 5 concludes the study.

2. Materials and Methods

2.1. Field Campaign

As shown in Figure 1a, the TD, located in the Xinjiang Province, China, is a typical arid zone. TZ (39 00′N, 83 63′E, 1140 m) is located in the hinterland of the TD, where dust activity is frequent and intense, and flowing sand dunes account for more than 80%. The temperature varies greatly, with a maximum temperature of 44.2 °C and a minimum temperature of 18.6 °C during our observation period. Precipitation in TZ is extremely scarce, with an average annual precipitation of 26 mm in recent years, which is 37.3% lower than the average precipitation in surrounding regions [43]. Our intensive field experiment was conducted in TZ during July 2021. The experimental location was selected near a desert road at a distance of approximately 50 m, surrounded by several restaurants. Sand dunes (height of 30–150 m) surrounded the observation location from which the UAV was operated (Figure 1b). TZ is primarily crossed by winds traveling from the northeast, east, and north in July, with a 14.9% frequency of easterly winds and an average wind speed of 2.09 m/s near the surface (Figure 1c).

2.2. UAV

An eight-rotor UAV was manually operated on a concrete surface during the period of 9–31 July 2021. The maximum load of the UAV was approximately 7.3 kg, and the duration of the sustained flight time was 39.5 min (without a load). In our observations, the portable instruments equipped on the UAV included a microAeth (model: MA200), a personal ozone monitor (POM_ 2 B), a dust monitor spectrometer (model: Grimm11_D), and a temperature/humidity/pressure logger. To validate the data, all instruments were calibrated to a uniform time and sampled 10–15 min in advance to allow the instruments to warm up before each flight. Once the UAV reached 10 m in height, it was switched to a 1.2–1.5 m/s steady ascent and descent to achieve an optimized vertical resolution. As a result, a typical flight required approximately 13 min. A total of 200 flights were successfully conducted every 3 h at 10:00 (local morning), 13:00, 16:00 (noon), 19:00 (afternoon), and 22:00 (evening) (Beijing Time) to capture diurnal variations. To reduce the disturbance of dust from the surface due to the rotation of the paddles during takeoff and landing, data were only used after an ascent of 10 m above the ground in this study. Observations of the portable instrument at the fixed sites of Lanzhou University and TZ were conducted concurrently before the start of the experiment, and the observations of the portable instrument were in excellent agreement.

2.3. Dust Monitor Spectrometer

The dust monitoring spectrometer model Grimm11_D (Grimm Aerosol Technik (Ainring, Germany), hereinafter: 11D) was used to determine the particle number size distribution (PNSD), which subsequently displayed output related to nine dust mass fractions: total suspended particles (TSP), PM₁₀, PM₄, PM_2.5, PM₁, PM_Coarse, inhalable, thoracic, respirable, as well as the PNC distribution in 31 channels ranging from 0.253 to 35.15 μm. The instrument works on the principle of scattered light measurement pertaining to individual particles and utilizes a polytetrafluoroethylene (PTFE: a plastic made of fluorine and carbon, also known as Teflon) filter for gravimetric validation of the calculated dust mass. The time resolution of 11D sampling time resolution can be varied between 6 or more (31 channels) and 1 s (only 16 channels). A time resolution of six seconds was selected for this experiment. The volume flow rate was set to 1.2 L/min.

In this study, the dust mass fractions were calculated based on the PNSD measured using 11D. The data from each observation were smoothed to remove outliers. The PNC summed the particle concentrations of the 31 channels for each observation. The 15 μm radius is the R_eff for the largest particle size considered in the inversion of AERONET optical data into a particle volume size distribution [44,45]. Furthermore, the microphysics parameters (particle volume size distribution ($dV\left(r\right)/dlnr$), particle R_eff, and Cv) were calculated using Equations (1)–(3) based on PNSD ($dN\left(r\right)/dlnr)$. R_eff is the quotient of the sum of the cubic and quadratic radii of all particles, and represents the average distribution of the aerosol radius in the atmosphere, and Cv denotes the total volume of aerosol particles per unit volume. All observation data were refined to remove outliers, and the observed spectral distributions were compared with the results obtained from AERONET [46] to show that the data processing and its calculation process were reasonable. The maximum and minimum values of the 11D particle size are represented by r_max and r_min, respectively.

In this study, the sensor used was for dry aerosols, and the instantaneous relative humidity observed at the time of the precipitation event was greater than 60%, which can affect the mass concentration; thus, the observed mass concentration in three different size fractions postprecipitation were corrected using Equation (4) [13,47]. PMs represent the particle mass concentrations of five size fractions (TSP, PM₁₀, PM₄, PM_2.5, and PM₁). PM₁₀, PM₄, PM_2.5, and PM₁ refer to particle diameters less than or equal to 10, 4, 2.5, and 1 μm, respectively. TSP is a general term for solid and liquid particulate matter floating in the air with a particle size range of about 0.1–100 μm.

$$R_{eff}=\frac{{\int }_{r_{min}}^{r_{max}}{r}^3\frac{dN\left(r\right)}{\mathit{dln}r}\mathit{dln}r}{{\int }_{r_{min}}^{r_{max}}{r}^2\frac{dN\left(r\right)}{\mathit{dln}r}\mathit{dln}r}$$

(1)

$$\frac{dV\left(r\right)}{\mathit{dln}r}=V\left(r\right)\frac{dN\left(r\right)}{\mathit{dln}r}=\frac{4}{3}\pi r^3\frac{dN\left(r\right)}{\mathit{dln}r}$$

(2)

$$C_v=\underset{r_{min}}{\overset{r_{max}}{\int }}\frac{dN\left(r\right)}{\mathit{dln}r}\mathit{dln}r$$

(3)

$$\mathrm{P}\mathrm{M}\mathrm{s}=\frac{\mathrm{P}\mathrm{M}\mathrm{s}}{1+0.25\frac{{\mathrm{R}\mathrm{H}}^2}{1-\mathrm{R}\mathrm{H}}}$$

(4)

2.4. Radiosonde

The radiosonde system used to provide vertical information of atmospheric temperature, relative humidity, pressure, and horizontal wind speed/direction in this experiment was composed of GPS sounding and a ground antenna receiving system (Beijing Changfeng Microelectronics Technology Co., Ltd., Beijing, China). The data vertical resolution was 10 m and the average speed of the balloon was 300 m/min. Sounding observations were made every 6 h during 9–31 July (01:15, 07:15, 13:15, and 19:15 Beijing Time). In addition, intensive observations were carried out at 04:15 and 22:15 from 6 to 15 July and at 10:15 and 16:15 from 12 to 21 July. We analyzed the meteorological elements during three different precipitation processes and dust storms using sound data from 0 to 500 m altitude.

2.5. Meteorology Elements

A portable micro-automatic weather instrument (NK5500 Kestrel, Chicago, US) was deployed to collect near-surface temperature, relative humidity, pressure, and wind vector on a bare leaky sand dune, at a distance of 50 m from the UAV observation location and 1.2 m above the ground. This experiment was performed using a new factory-calibrated NK5500 instrument with a time resolution of 10 min. Compared to the same type of product, the NK5500 has the characteristics of high precision, high performance, and portability.

3. Results

3.1. Measurement Overview

As shown in Figure 2d–f, during the observation period, the mean values of temperature and pressure in TZ were 30.77 °C and 882.5 hPa, respectively, and were mainly influenced by the northeast and east traveling winds with an average wind speed (WS) of 3.13 m/s. The PMs mass concentration and PNC within 10–30 m above ground level were averaged to represent particle variation over the entire observation period (Figure 2a–c). On 12, 17, and 31 July (19:00–22:00), the values of PM₁₀, PM_2.5, PNC, and TSP increased rapidly, and the maximum values of PM₁₀ and TSP exceeded 500 μg/m³ and 6000 μg/m³, respectively, considering that a dust-storm event occurred during these three days. The overall trend of PM mass concentration and PNC increased in the days before the occurrence of dust storms on 12 and 31 July, when the pressure was lower than the average pressure and daily maximum temperature gradually decreased. The constant high temperatures and high ground WS, as well as the ample dust sources in the surrounding environment, contributed to the high dust concentrations in TZ. The largest dust storm during the experiment occurred on the night of 17 July, in which TSP, PNC, PM₁₀, and PM_2.5 were all higher than those on 12 and 31 July. Near-ground TSP, PM₁₀, and PM_2.5 were 6000–12000, ~1100, and 380 μg/m³, respectively. Before the dust storm, the PM mass concentration and PNC values were low, and there was no significant trend in the barometric pressure or daily maximum temperature; however, the WS was at a high level throughout the day. As shown in the observation records (yellow shading), there were continuous dusty weather occurrences from 19:00 to 10:00 on 10, 17, and 30–31 July, with a significant reduction in visibility in the afternoon and evening. The dust storms on 12 and 31 July were mainly influenced by larger weather systems that lasted longer, whereas 17 July was influenced more by near-surface winds and experienced a slightly shorter duration of dust storms driven by small-scale weather systems.

As shown in Figure 2d, the maximum relatie humidity (RH) on nonprecipitation days during the observation period was 52.7%. We determined that an RH greater than this value represents the occurrence of precipitation. Three precipitation events occurred during the study period. Precipitation occurs mostly at night and early in the morning. The first precipitation event (P1) lasted for 4 h from 6:30 on 9 July, when the ground temperature was high and evaporation was faster. The second precipitation event (P2) occurred on 13 July, lasting for 8.5 h, during which high-intensity precipitation lasted for 1.5 h (RH = 100%) and had a significant wetting effect on the dust layer (observation records indicated that P2 had the largest precipitation volume). The third precipitation event (P3) had the longest duration, from 23:00 on 19 July to 10:30 on 20 July; however, the intensity of precipitation was the lowest of the three events, during which RH failed to exceed 80%.

Typically, PM₁₀ concentrations are used to classify different pollution levels. The PM₁₀ evaluation criteria in the ambient air quality standards have been implemented in China since 1 January 2016. The yellow sand intensity forecasting criteria were developed by the Ministry of Environment and the Meteorological Agency of Korea in April 2002, which delineate the criteria for dusty weather levels [48,49]. In general, the trend of the mass concentration was more obvious in the range of 10–100 m, and the influence of sand dunes at the observation points was effectively avoided. The PM₁₀ maximum value, within 10–100 m in our observations, was selected to distinguish dusty weather. As shown in

Table 1. The criteria for identifying four scenarios using PM₁₀.

	Clear Sky	Floating Dust	Blowing Sand	Dust Storm
PM10 (μg/m3)	0–150	150–350	350–500	>500

, the four scenarios can be divided into clear sky, floating dust, blowing dust, and dust storm. According to the criteria, 149 flights observed clear skies, 33 flights observed floating dust, 4 flights observed blowing dust, 8 flights observed dust storms, and 6 flights occurred during postprecipitation, of which approximately 75% were in clear-sky conditions.

3.2. Vertical Characteristics under Clear and Dust Scenarios

According to the criteria given in Table 1, four flights were selected to present the vertical characteristics of mass concentration changes under different situations. As shown in Figure 3, the PM mass concentrations in the different size fractions increased significantly from clear sky, floating dust, and blowing dust to dust storms. For example, the mean TSP during a dust storm is 5348.8 µg/m³, which is approximately 6 and 15 times higher than that of blowing dust (938.1 µg/m³) and floating dust (361.1 µg/m³) scenarios, respectively, and 318 times higher than that of the clear sky (16.8 µg/m³). The mass concentration ratio decreased for the four scenarios when small particle size fractions were considered. For instance, the average value of PM₁ during a dust storm was 44.4 µg/m³, which is 1.2, 1.8, and 13.4 times higher than those during blowing sand, floating dust, and clear sky, respectively. This indicates that the contribution of large particles to the mass concentration had a significant effect on this ratio under different scenarios. During clear-sky periods, the mass concentrations in different size fractions revealed a decreasing trend, with heights below approximately 80 m near the ground, remaining essentially constant above that height. In particular, the profiles for dust storms had a greater decreasing trend than those for clear sky at altitudes below approximately 40 m. As pointed out by [50], the mass concentration distribution varied slightly above 32 m using an 80 m meteorological tower over TZ. The observations reported in the current study are consistent with these results.

As mentioned previously, the four scenarios can be considered as almost instantaneous vertical distributions of dust particles. Compared to the clear-sky and dust-storm scenarios, the TSP profiles for the floating-dust and blowing-dust scenarios exhibited more pronounced swellings in Figure 3b,c. Considering the high spatial resolution sampling characteristics of 11D, we attribute these pronounced swellings to the sudden increase in large dust particles with radii greater than 10 µm on the fine scale of tens to hundreds of meters in the dust layer. This finding is consistent with the results obtained by [35] during the Charmex campaign in 2013. As shown in Figure 3e–h, the clear-sky and dust-storm scenarios provide a reference range for the PM₁₀/TSP ratio. During the dust storm, the ratio remained almost constant at approximately 0.12 with height, which means that the percentage of particles larger than 10 µm was close to 90%. Under clear-sky conditions, the ratio increased with height below approximately 100 m and then remained almost constant at 1. This indicates that there were almost no particles larger than 10 µm above approximately 100 m. In particular, the maximum and minimum ratios under the floating-dust scenario were comparable to those under the clear-sky and dust-storm scenarios. In contrast, the PM_2.5/PM₁₀ ratios for the four scenarios were concentrated in the range of 0.3–0.6. The ratios for the three dust scenarios did not show a significant trend with height, with corresponding mean values of 0.48, 0.43, and 0.33, respectively, showing a trend of increasing and then decreasing with height under a clear sky, with peak values occurring at approximately 220–320 m. This again confirms the influence of particles above 10 µm on the mass concentration as described above.

PNC, R_eff, and Cv showed consistent changes with respect to mass concentration in the four scenarios (Figure 4a–c). The mean PNCs for the clear-sky, floating-dust, blowing-sand, and dust-storm scenarios were 7246, 58,765, 88,452, and 111,780 P/L, respectively. Compared with the PNC in a clear sky, the dust scenario increases the PNC by 8–16 times. In particular, the vertical decreasing rate of PNC under the dust-storm scenario was 2294 P/L per m (height below approximately 40 m), which was approximately 14 times that of the clear-sky scenario decreasing rate (161.35 P/L per m). In desert areas, aerodynamic factors significantly influence the vertical distribution of dust aerosols in the lower troposphere, particularly during the dust processes. Combined with the WS in Figure 2e, this may be because the hourly average WS exceeded 10 m/s at 19:00 on 17 July and the sand particles swept by strong winds rapidly changed the PNC below 45 m. The PNSD of the four scenarios shown in Figure 5a,b show that regardless of the scenario, the PNC of fine particles is invariably dominant (e.g., the PNC with a radius smaller than 1 μm accounts for 80%), whereas the PNC of large particles in the dust-storm scenario is only approximately 5–25 P/L. Therefore, it is clear that the PNC in Figure 4a exhibits virtually the same vertical characteristics as PM1 in Figure 3a–d. R_eff and Cv exhibit almost synchronous trends in the vertical profile. The profile-averaged R_eff (Cv) increased from 0.70 µm (0.03 µm³/µm²) in the clear-sky scenario to 4.55 µm (0.73 µm³/µm²) in the dust-storm scenario. Some studies have pointed out that the average R_eff was less than 5 μm during clear-sky, floating-dust (1.17 μm), and blowing-sand scenarios (2.12 μm). In contrast, R_eff was greater than 5 μm at 300–500 m during dust storms [41].

Combined with the PNSD in Figure 5a,b, it is clear that although the number of large particles is small, their effect on R_eff and Cv is significant. This can be understood in two ways: (1) For the two fine-dust layers at approximately 130 m and 350 m in the floating-dust scenario in Figure 3b, the TSP peak is in good agreement with the R_eff peak at the same altitude. Under the floating-dust scenario, it can be found that only the TSP profile in these two fine-sand layers shows a clear peak at the same height, excluding other mass fractions, which means that the increase in particles larger than 10 µm is more obvious in the two fine-dust layers. (2) Excluding TSP, the mass concentration distributions in other size fractions under all dust scenarios were relatively similar (Figure 3b–d); however, the R_eff and Cv distributions under the floating-dust scenario were closer to those under the clear-sky scenario (Figure 4b,c). In particular, the R_eff and Cv values in the floating-dust scenario were very close to those in the clear-sky scenario, except for two peaks, where R_eff and Cv were higher than those in the clear-sky scenario. This variation was mainly due to the presence of particles with radii larger than 10 µm in the dust-storm scenario.

3.3. Diurnal Evolution

Only the 149 clear-sky flights were considered to represent the diurnal evolution of the background mass and microphysical properties over TZ in July. As shown in Figure 6a–e, the average profiles of the mass concentrations in different size fractions at five time points (10:00, 13:00, 16:00, 19:00, and 22:00) showed a decreasing trend with height, especially below 100 m. The reason for this pattern is that wind-driven sand dust was transported from nearby natural sand dunes. For example, the rate of decrease in the TSP below 100 m reached 3.8 μg/m³ per m at 13:00. PMs in July 2021 in TZ showed a single-peak distribution, with the peak occurring at approximately 13:00. Combined with WS, the distributions of WS and PMs showed the same daily variation characteristics. The average WS near the ground reached a maximum of 4.4 m/s at approximately 13:00 on that day (Figure 2); however, the WS persisted at this level throughout the afternoon. The high WS near the surface of the sand source was more likely to roll up dust particles into the atmosphere and add to the increase in PMs. It has been shown that the daily variation in aerosol mass concentration is positively correlated with RH and pressure and negatively correlated with temperature [26,40], although these effects are not as significant as those of WS in this study. A possible reason is associated with the particle emission mechanism being closely related to WS over dust source areas, unlike the emissions in urban areas. The maximum atmosphere boundary layer height (ABLH) in TZ occurred at approximately 16:00 h in the day [51,52]. This strong thermal mixing within the atmosphere boundary layer (ABL) uplifted the near-surface particles to higher altitudes, resulting in a decrease in the mass concentration and PNC near the surface. As shown in Figure 6f,g, the average PM₁₀/TSP ratio at 13:00 was smallest at 0.43 and increased gradually thereafter. The average PM_2.5/PM₁₀ ratio was 0.46, which decreased until 16:00 and then gradually increased. This also confirms that WS is a key factor in the formation of coarse airborne dust particles.

As shown in Figure 7a, PNC also showed a decreasing trend with increasing height, particularly at 10:00. This is because in the experimental region, this stage occurs when the convective boundary layer starts to develop; here, the PBLH is usually less than 500 m. Furthermore, R_eff and Cv varied almost constantly with height. The R_eff remained approximately in the range of 1–2 μm, while the Cv remained mainly between 0.06 and 1.08 μm³/μm². Among them, R_eff and Cv at 13:00 and 16:00 were considered to be at a maximum level at different heights. The diurnal variation of PNC, R_eff, and Cv were similar to those of mass concentration showing a single-peaked distribution, with the peaks occurring at 13:00–16:00, when their values were 15,467 P/L, 1.4 μm, and 0.13 μm³/μm², respectively.

4. Discussion

4.1. Comparison with Previous Studies

According to the differentiation criteria in Table 1, the mean values of the mineral dust microphysical properties under different scenarios are listed in

Table 2. The mean and standard deviations of PNC, R_eff, and Cv under different scenarios.

Parameter	Flights	PNC P/L	Reff µm	Cv µm3/µm2
ALL	200	21,290 ± 3046	1.47 ± 1.48	0.16 ± 0.34
Clear sky	149	14,712 ± 1877	1.22 ± 1.12	0.11 ± 0.17
Floating dust	33	31,034 ± 4391	1.81 ± 1.86	0.28 ± 0.60
Blowing sand	4	64,447 ± 15,400	2.43 ± 2.13	0.35 ± 0.47
Dust storm	8	83,297 ± 12,881	3.86 ± 1.44	0.54 ± 0.30
Postprecipitation	6	16,867 ± 2885	1.86± 2.45	0.33 ± 0.70

. The mean PNC, R_eff, and Cv in TZ were 21,290 P/L, 1.47 μm, and 0.16 μm³/μm², respectively, during the experimental period. The microphysical properties gradually increased from clear sky, floating dust, and blowing dust to dust storm. Some relevant studies were reviewed for comparison with previous observations. As listed in

Table 3. The list of aerosol microphysical properties by UAV.

	Land Type	Size Range µm	PNC P/L	Reff µm	Cv µm3/µm2
She et al. (2018) [52]	Urban	---	---	1–1.8	---
Crumeyrolle et al. (2013) [53]	Urban	---	---	---	0.005–0.47
unpublished results	Urban		229,720	0.87	0.01
Kezoudi et al. (2021) [37]	Forest	0.5–50.0	1000–30,000	2.4	---
Weinzierl et al. (2011) [54]	Desert	0.1–30	83,300	2.47	---

, only a few studies have addressed these parameters based on UAV, and large differences exist between the values obtained from observations of different land types [34,37,52,53]. The particle size range of the instrument carried by Weinzierl during UAV observations in the Sahara Desert was similar to that in this study. R_eff was similar to the observations during the blowing sand scenario, whereas the PNCs were comparable to those under the dust-storm scenario [54]. In urban areas, PNC data were mostly obtained from mobile monitoring campaigns, e.g., 229,720 P/L, 0.87 μm, and 0.01 μm³/μm² for PNC, R_eff, and Cv, respectively, monitored in Lanzhou, northwest China, in January 2020 (unpublished results). The PNC in the desert area was approximately 1/10th of that in the urban area, whereas R_eff and Cv were larger than the values observed by urban UAV and mobile monitoring. In addition, R_eff was similar to observations in desert and forest land; however, the PNC varied greatly [37,52,53].

4.2. Vertical Characteristics of Postprecipitation Scenarios

During the campaign period, fortuitous precipitation conditions were captured as rain extremely rarely in the desert hinterland. In the three postprecipitation scenes shown in Figure 8 (P1, P2, and P3), the mean mass concentrations in different size fractions, other than the TSP for P1 and P3, were similarly located between P2 and the dust-storm scenario. The associated mean PM₁₀ values were approximately three times higher than those of P2. In contrast, the mean TSP for P2 was seven to nine times greater than those for P1 and P3. For P2, the PM₁₀/TSP ratios were lower than those of P1 and P3. The TSP decreased with height and revealed two peaks at 80 m and 240 m below 320 m, and its peak value was approximately 5–22 times those of P1 and P3. In Figure 8f, most of the PM₁₀/TSP ratios below 320 m were less than those of the dust-storm scenario, and the values above 320 m lay between those of the dust-storm scenario and P1. Therefore, it can be observed that there are a large number of dust particles with sizes greater than 10 µm below 320 m in P2. Overall, PM_2.5/PM₁₀ decreased with increasing height.

As shown in Figure 9, due to the scavenging effect of rain on airborne particles, it can be seen that the profile-averaged PNC for P1–P3 accounts for 69–89% of the dust-storm scenario (73329 P/L), with the lowest PNC in P2. The profile-averaged R_eff and Cv of P2 were 3.6 μm and 0.64 µm³/µm², respectively, which were significantly 5 and 14 times higher than those of P1 and P3, respectively, and the P2 values were comparable to those of the dust-storm scenario (3.9 μm and 0.52 µm³/µm²). The PNSD in Figure 10a shows that particles with a radius larger than 10 μm were observed in P2, similar to the dust-storm scenario. A possible reason for this is that the WS corresponding to P2 was much greater than those of P1 and P2, and even higher than that during the dust-storm scenario near the ground surface. This high WS accelerates the evaporation of water from the wet-sand layer, and the moist dust particles are able to break free of the sticky wet-sand layer and are then carried into the air from the surrounding sand dunes. In addition, a high-RH environment after precipitation is more favorable for water vapor to wrap around dust particles, thus adsorbing smaller dust particles onto the ground surface [55].

4.3. The Importance of UAV Observations for Lidar Inversion

Although the microphysical properties of aerosols can be inverted well using lidar, the impact of RH on the aerosol mass and microphysical properties is non-negligible, as shown in Figure 9 and Figure 10. Dionisi reported that lidar-inversed Cv and in situ observations agreed well when the RH was less than 60%, while the lidar-inversed Cv was 3–3.5 times higher than that in situ when the RH was greater than 90% [56]. Ziemba showed that the extinction coefficient was underestimated by 32%, indicating that the water in the particles had a significant effect on lidar inversion [13]. Therefore, it is necessary to introduce in situ observations with UAV as independent constraints for lidar data inversion to improve the performance of lidar inversion algorithms under multiple scenarios.

5. Conclusions

In this study, the vertical characteristics and diurnal evolution of near-surface dust mass concentrations of different PMs, PNC, R_eff, and Cv were determined using a UAV observation platform during field campaigns in July 2021 in the TZ region of the TD hinterland. To the best of our knowledge, this is the first report on the vertical profiles of PNC, R_eff, and Cv using a UAV over the TD.

From clear-sky, floating-dust, and blowing-sand to dust-storm scenarios, PMs, PNC, R_eff, and Cv gradually increased, and the values of PNC in the four scenarios were 14650, 58,765, 88,452, and 111,779 P/L, respectively. Compared with that under clear-sky scenarios, PNC increased by four to eight times under dust scenarios, while R_eff and Cv increased from 1.15 μm and 0.08 μm³/μm² under the floating-dust scenario to 4.53 μm and 0.74 μm³/μm² under the dust-storm scenario, respectively. The diurnal variations in PMs, PNC, R_eff, and Cv showed a unimodal distribution, with the peak occurring at approximately 13:00–16:00. Wind speed was the key factor causing peaks in mass concentrations and microphysical parameters in the desert hinterland, which in turn decreased near the surface as the convective boundary layer developed. PMs and PNC below the boundary layer height were approximately 1.5 times higher than those above, whereas the TSP/PM₁₀ ratio, PM₁₀/PM_2.5 ratio, R_eff, and Cv remained almost unchanged.

Only a few studies have covered PNC, R_eff, and Cv based on UAVs, and there are large differences among the observations of different land types. The R_eff in the Sahara Desert was similar to that observed in the TD during the blowing-sand scenario, whereas the PNCs was comparable to those under the dust-storm scenario. The PNC in the desert area was approximately 1/10th that in the urban area, whereas R_eff and Cv were larger than those in the urban region. For the P2 postprecipitation scenario, the PNC, R_eff, and Cv were 8363 P/L, 3.57 μm, and 0.64 μm³/μm², respectively. The PNC of P2 was smaller than that under the clear-sky scenario, whereas R_eff and Cv were similar to or greater than those for the dust-storm scenario. This increase in particle radius postprecipitation was associated with high WS and an increase in RH.

Here, we neglected the effect of the propeller-disturbance airflow on the sampled air characteristics during the rotorcraft flight because of the lack of other similar sampling instruments for auxiliary correction. In future studies, this effect will be further evaluated and applied to lidar inversion.