Observation and Classification of Low-Altitude Haze Aerosols Using Fluorescence–Raman–Mie Polarization Lidar in Beijing during Spring 2024

Abstract

Haze aerosols have a profound impact on air quality and pose serious health risks to the public. Due to its geographical location, Beijing experienced haze events in the spring of 2024. Lidar is an active remote sensing technology with a high spatiotemporal resolution and the ability to classify aerosols, and it is essential for effective haze monitoring. This study utilizes fluorescence–Raman–Mie polarization lidar with an emission wavelength of 355 nm, employing the ${\delta }_p$-$G_f$ method based on the particle depolarization ratio at 355 nm (${\delta }_{p355}$) and the fluorescence capacity ($G_f$), and combines meteorological data and backward-trajectory analysis to observe and classify low-altitude haze aerosols in Beijing during the spring of 2024. Notably, a mining dust event with strong fluorescence backscatter was detected. The haze aerosols were categorized into three types: pollution aerosols, desert dust, and mining dust. Their optical properties were summarized and compared. Desert dust showed a particle depolarization ratio range of 0.23–0.39 and a fluorescence capacity range from 0.18 × 10⁻⁴ to 0.63 × 10⁻⁴. Pollution aerosols had a larger fluorescence capacity but a lower depolarization ratio compared to desert dust, with a fluorescence capacity ranging from 0.55 × 10⁻⁴ to 1.10 × 10⁻⁴ and a depolarization ratio ranging from 0.10 to 0.17. Mining dust shared similar depolarization characteristics with desert dust but had a larger fluorescence capacity, ranging from 0.71 × 10⁻⁴ to 1.23 × 10⁻⁴, with a depolarization ratio range of 0.30–0.39. This study validates the effectiveness of the ${\delta }_{p355}$-$G_f$ method in classifying low-altitude haze aerosols in Beijing. Additionally, it offers a new perspective for more detailed dust classification using lidar. Furthermore, utilizing the ${\delta }_{p355}$-$G_f$ classification method and the ${\delta }_{p355}$-$G_f$ distributions of three typical aerosol samples, we developed a set of equations for the analysis of mixed aerosols. This method facilitates the separation and fraction analysis of aerosol components under various mixing scenarios. It enables the characterization of variations in the three types of haze aerosols at different altitudes and times, offering valuable insights into the interactions between desert dust, mining dust, and pollution aerosols in Beijing.

1. Introduction

Aerosols are ubiquitous in the Earth’s atmosphere, significantly influencing climate systems and public health [1]. Haze, which severely affects air quality, is closely linked to aerosols. Given the complex compositions of haze aerosols, different types of aerosols contribute variably to the formation and development of haze [2]. Therefore, analyzing the physical and chemical properties and radiative effects of different types of haze aerosol particles is a key focus in haze research [3]. Currently, various methods are available for the monitoring of the properties of haze particles [4,5,6]. Traditional monitoring techniques use in situ instruments to collect measurement data at a single location, allowing for the analysis of individual particles but failing to capture the dynamic changes and spatiotemporal distribution of haze over a large area [7]. Passive remote sensing methods, such as those employing sun photometers, rely on the sun as a light source and can only study the properties of aerosols in the entire atmosphere. Meanwhile, they are unable to analyze localized aerosol types and their specific characteristics [8]. As an active remote sensing device capable of continuous long-term observations, lidar offers high temporal and spatial resolutions in aerosol detection. It provides spatiotemporal particle characteristic parameters, such as the lidar ratio [9], Ångström index [10], and depolarization ratio [11], which enable the classification of aerosol particles at different times and heights [12,13,14], facilitating the analysis of aerosol formation and development. Therefore, lidar is crucial in understanding the vertical structure, composition, and transport processes of haze [15].

Taking the low-altitude haze aerosols in Beijing during the spring of 2024 as an example, Beijing’s location at the center of the North China industrial area, surrounded by numerous industrial production regions, results in air pollution from local and surrounding emissions [16]. This contributes to the high frequency of haze events in the region. In addition to pollution, dust intrusion is a major cause of haze in Beijing during spring [17,18]. The vast deserts and mining areas in Northern China and Southern Mongolia [19,20], influenced by the East Asian tropospheric westerly jet, become sources of dust events in Beijing. Westerly winds transport dust particles, leading to frequent dust events in Beijing during the spring [21]. The mining areas in this region contain a diverse range of minerals, including abundant non-metallic minerals such as fluorite, which contains calcium ions [22]. These non-metallic minerals can produce strong fluorescence effects when exposed to ultraviolet light [23]. Due to human mining activities, these areas have generated significant amounts of mineral dust. In this study, we denote mineral dust originating from mining activities as mining dust, specifically focusing on minerals such as fluorite that exhibit strong fluorescence effects. In contrast, desert dust from the desert regions mainly consists of particles such as clay minerals and quartz [24], which are predominantly composed of silicon dioxide and do not produce strong fluorescence effects. Since both desert dust and mining dust are classified as mineral dust, they tend to exhibit high depolarization ratios [25,26], indicating their greater non-sphericity compared to other aerosols. Sugimoto et al. [27] observed East Asian dust using polarization lidar combined with a spectrometer and found that mineral dust had inherent fluorescence properties. Therefore, the fluorescence characteristics can be used to distinguish between desert dust and mining dust. Regarding the detection of pollution aerosols, Li et al. [28] observed haze in Beijing using fluorescence–Mie lidar and found a strong correlation between the local aerosol fluorescence efficiency and the pollutant concentration. Wang et al. [29] utilized dual-wavelength lidar combined with a spectrometer to study the optical properties of desert dust and pollution aerosols. Their results indicated that the volume depolarization ratio of pollution aerosols was significantly lower than that of desert dust, but their fluorescence efficiency was higher. Therefore, it is feasible to use lidar combined with particle fluorescence characteristics and non-spherical depolarization properties to distinguish between pollution aerosols, mining dust, and desert dust in low-altitude haze aerosols in Beijing during the spring.

The systematic analysis of European aerosols using particle fluorescence and depolarization characteristics has also been undertaken by researchers [30,31,32,33]. Veselovskii et al. [34] conducted long-term observations of local aerosols using multi-wavelength lidar combined with fluorescence channels. They calculated the fluorescence capacity ($G_f$) and particle depolarization ratio (${\delta }_{p532}$) using the backscatter coefficient at 532 nm and classified urban, dust, smoke, and pollen aerosols using the ${\delta }_{p532}$-$G_f$ method. In their latest research [35], they constructed a mixed-aerosol equation system based on these parameters to quantitatively analyze the proportions of mixed aerosols in the area. For the Beijing area, Zhang et al. [36] classified low-altitude aerosols using the volume depolarization ratio and fluorescence–Mie ratio. They also qualitatively analyzed mixed aerosols using backward-trajectory analysis. However, the absence of a Raman channel prevented the accurate quantification of the aerosol fluorescence characteristics, thereby preventing the quantitative analysis of mixed aerosols. This study utilized fluorescence–Raman–Mie polarization lidar with an emission wavelength of 355 nm. Using the backscatter coefficient at 355 nm, the particle depolarization ratio and fluorescence capacity were determined. The ${\delta }_p$-$G_f$ method was then employed to classify the low-altitude aerosols in Beijing during the spring haze events of 2024, including mining dust, desert dust, and pollution aerosols, as well as their mixtures.

This paper is organized as follows: Section 2 describes the lidar system setup and data processing methods, along with a detailed explanation of the ${\delta }_p$-$G_f$ mixed-aerosol analysis method. Section 3 presents the analysis of three typical cases of single aerosols and summarizes the classification parameter ranges for each aerosol component. Section 4 discusses the fractions of each component in two cases of mixed aerosols using the mixed-aerosol analysis method. Finally, the concluding section summarizes the findings of this study.

2. Materials and Methods

2.1. Lidar System

The fluorescence–Raman–Mie polarization lidar system used in this study was located at the Zhongguancun campus of the Beijing Institute of Technology (39.96°N, 116.31°E). Building on the system configuration described by Zhang et al. [36], which uses an excitation wavelength of 355 nm and includes horizontal and vertical polarization elastic scattering channels and a fluorescence channel, a nitrogen vibrational Raman scattering channel with a central wavelength of 387 nm was added. Details of the other system parameters can be found in the referenced paper. Due to the strong sunlight background affecting the fluorescence channel during the day, the system was operated only at night. The calibration methods for the polarization channel and the fluorescence–Raman channel were the same as those used by Veselovskii et al. [30]. The data integration time for each channel was 5 min, corresponding to an accumulated pulse count of 6000.

2.2. Methods

2.2.1. Data Processing Methods

With the addition of the nitrogen vibrational Raman channel, the aerosol elastic backscatter coefficient at 355 nm ${\beta }_{p355}$ can be determined using the Raman–Mie method [37]. This method requires the selection of a reference height and value, which can introduce uncertainty when analyzing long-term spatiotemporal variations in inversion parameters. Therefore, based on the current system configuration, this study modified the method similarly to Veselovskii et al. [34] and performed constant calibration using the vibrational Raman channel. For simplicity, ${\beta }_{355}$ will be used to represent ${\beta }_{p355}$ in the following text, which is

$${\beta }_{355}=K{\displaystyle \frac{P_{\mathrm{L}}}{P_{\mathrm{R}}}}{N}_{\mathrm{R}}{\sigma }_{\mathrm{R}}{\displaystyle \frac{T_{\mathrm{R}}}{T_{\mathrm{L}}}}-{\beta }_{m355},$$

(1)

where K is the calibration constant for the elastic scattering channel and Raman channel, $P_L$ is the signal strength of the elastic scattering channel, $P_R$ is the signal strength of the Raman channel, $N_R$ is the number density of nitrogen molecules per unit volume, ${\sigma }_R$ is the backscatter differential cross-section of nitrogen molecules, the transmittance ratio is calculated using an Ångström index of 1, and ${\beta }_{m355}$ is the molecular backscatter coefficient at 355 nm.

The particle depolarization ratio ${\delta }_{p355}$ is the ratio of the vertical to horizontal components of the aerosol backscatter coefficient at 355 nm. For simplicity, ${\delta }_p$ will be used to represent ${\delta }_{p355}$. The calculation method for the particle depolarization ratio ${\delta }_p$ is the same as described in Ref. [38], which is

$${\delta }_{\mathrm{p}}={\displaystyle \frac{(1+{\delta }_m){\delta }_v{R}_{\beta }-(1+{\delta }_v){\delta }_m}{(1+{\delta }_m)R_{\beta }-(1+{\delta }_v)}},$$

(2)

where ${\delta }_m$ represents the linear depolarization ratio of air molecules; ${\delta }_v$ is the volume depolarization ratio, which is the ratio of the signal intensities in the vertical and horizontal polarization channels; and $R_{\beta }$ is the aerosol backscatter ratio, which is the ratio of the total backscatter coefficient to the molecular backscatter coefficient.

The calculation of the fluorescence backscatter coefficient ${\beta }_F$ follows the method described in Ref. [30], which is

$${\beta }_{\mathrm{F}}=k{\displaystyle \frac{P_{\mathrm{F}}}{P_{\mathrm{R}}}}{N}_{\mathrm{R}}{\sigma }_{\mathrm{R}}{\displaystyle \frac{T_{\mathrm{R}}}{T_{\mathrm{F}}}},$$

(3)

where k is the calibration constant for the fluorescence and Raman channels, $P_F$ is the signal strength of the fluorescence channel, and $P_R$ is the signal strength of the Raman channel. The term $T_R/T_F$ simplifies to the ratio of atmospheric transmittance at the center wavelengths of the fluorescence and Raman channels.

The fluorescence capacity $G_f$ is the ratio of the fluorescence backscatter coefficient to the aerosol elastic backscatter coefficient [39]. Based on the system configuration adopted in this study, the fluorescence capacity is calculated as follows:

$$G_f={\displaystyle \frac{{\beta }_{\mathrm{F}}}{{\beta }_{355}}},$$

(4)

and the system is not affected by geometric factors above approximately 800 m, so 800 m is taken as the minimum height for valid data. Due to the weak fluorescence signal [40], the height where the actual fluorescence signal-to-noise ratio equals 10 is selected as the maximum height for valid data.

This study analyzes multiple aerosol cases to determine the ${\delta }_p$ and $G_f$ ranges specific to aerosol types that are sensitive to these parameters. When a particular type of aerosol appears alone, distinct stratification often emerges in the spatiotemporal distribution of these two parameters, identifying the aerosol layer. The boundaries of this layer are usually marked by significant changes in ${\delta }_p$ and $G_f$, allowing for the use of boundary enhancement methods to extract it. Within the aerosol layer, the parameters tend to change slowly or remain within a certain range. Burton et al. [25] used the Haar wavelet method to enhance the aerosol boundaries. To simplify this process, we applied a gradient method, as the gradients at the aerosol layer boundaries are significantly higher than those within the layer. By setting a specific threshold, we can identify the aerosol boundaries where significant changes occur. The regions between adjacent boundaries at the same time period are considered as a single aerosol layer, and the average of the parameters within this range is used as the characteristic parameter of the aerosol sample point for that layer, which also helps in reducing the data volume.

2.2.2. Mixed-Aerosol Analysis Method

For aerosol mixing, Veselovskii et al. [35] used the ${\delta }_{p532}$-$G_f$ method to establish a set of equations and applied the least-squares method to analyze the mixing of four types of aerosols. In this method, the variant forms of the mixed-aerosol particle depolarization ratio ${\delta }_p$, namely the depolarization potential ${\delta }_p^{\prime }={\delta }_p/(1+{\delta }_p)$ and the fluorescence capacity $G_f$, are linear combinations of the corresponding parameters of the individual particle types. This study adopts the same set of equations to discuss the mixing of three types of aerosols in the Beijing area: pollution aerosol, mining dust, and desert dust. The equations are expressed as follows:

$$\left\{\begin{array}{c}{\eta }_1+{\eta }_2+{\eta }_3=1\\ G_1{\eta }_1+G_2{\eta }_2+G_3{\eta }_3=G_f\\ {\delta }_1^{\prime }{\eta }_1+{\delta }_2^{\prime }{\eta }_2+{\delta }_3^{\prime }{\eta }_3={\delta }_p^{\prime }\end{array}\right.,$$

(5)

where the fraction of the i-th type of aerosol, denoted as ${\eta }_i={\beta }_i/\beta$, is the ratio of its backscatter coefficient to the total aerosol backscatter coefficient. The depolarization potential and fluorescence capacity of the i-th type of aerosol are represented as ${\delta }_i^{\prime }={\delta }_i/(1+{\delta }_i)$ and $G_i$, respectively, where i = 1,2,3. By substituting the ${\delta }_p$ and $G_f$ values of the mixed aerosol into the equations, the proportions of each aerosol component can be solved. This study uses the central points of the three aerosol samples as the values for ${\delta }_i$ and $G_i$. The determination of these central points will be discussed in Section 3.

Since the set of equations has reasonable solutions only within the triangle formed by the central points of the three types of aerosols, different methods must be used to solve the mixed states outside this triangle. In such cases, the third aerosol type is ignored, and the mixing state of the remaining two aerosols is directly analyzed. Let the particle depolarization ratios of the remaining two aerosols be ${\delta }_1$ and ${\delta }_2$, their fluorescence capacities be $G_1$ and $G_2$, their backscatter coefficients be ${\beta }_1$ and ${\beta }_2$, and the total backscatter coefficient of the mixed aerosols be $\beta ={\beta }_1+{\beta }_2$. The mixed aerosols ${\delta }_p$ and $G_f$ are then expressed as follows:

$${\delta }_p={\displaystyle \frac{\left(\frac{{\delta }_1}{1+{\delta }_1}\right){\beta }_1+\left(\frac{{\delta }_2}{1+{\delta }_2}\right){\beta }_2}{\left(\frac{{\beta }_1}{1+{\delta }_1}\right)+\left(\frac{{\beta }_2}{1+{\delta }_2}\right)}},$$

(6)

$$G_f={\displaystyle \frac{{\beta }_1{G}_1+{\beta }_2{G}_2}{\beta }},$$

(7)

and, by substituting ${\beta }_2=\beta -{\beta }_1$ into Equation (6), the fraction of the second type of aerosol is obtained as

$${\eta }_2={\displaystyle \frac{{\beta }_2}{\beta }}={\displaystyle \frac{({\delta }_p-{\delta }_1)(1+{\delta }_2)}{({\delta }_2-{\delta }_1)(1+{\delta }_p)}}.$$

(8)

Moreover, the aerosol particle depolarization ratios must satisfy ${\delta }_1\le {\delta }_p\le {\delta }_2$. Tesche et al. [41] and Mamouri et al. [42] used this method to separate high-depolarization-ratio dust from low-depolarization-ratio non-dust aerosols.

For aerosols with a low fluorescence capacity and high fluorescence capacity, substituting ${\beta }_2=\beta -{\beta }_1$ into Equation (7) allows for the determination of the fraction of the second type of aerosol:

$${\eta }_2={\displaystyle \frac{{\beta }_2}{\beta }}={\displaystyle \frac{G_f-G_1}{G_2-G_1}}.$$

(9)

Similarly, the aerosol fluorescence capacities must satisfy $G_1\le G_f\le G_2$. The fraction of the first type of aerosol is represented by ${\eta }_1=1-{\eta }_2$.

Considering the impact of the relative humidity on $G_f$, the above methods should be applied under conditions where the relative humidity is less than 60% to minimize the effects of hygroscopic growth on the particles [35].

2.3. Auxiliary Data

Air quality data (PM₁₀ and PM_2.5) were obtained from the Wanliu station (39.96°N, 116.30°E) of the Beijing Environmental Protection Monitoring Center, located 3 km from the lidar observation site. The daily average data were calculated using the mean values from 00:00 to 23:00 UTC.

Veselovskii et al. [43] emphasized the impact of the relative humidity on the aerosol fluorescence capacity. Under high-humidity conditions, aerosol particles undergo hygroscopic growth, increasing their backscatter and thus reducing their fluorescence capacity. In their discussion of aerosol mixing using the ${\delta }_{p532}$-$G_f$ method [35], the team set a relative humidity threshold of 60% to minimize the effects of hygroscopic growth. Similarly, this study uses relative humidity data obtained from radiosonde measurements at the Beijing 54511 Meteorological Observation Station (39.81°N, 116.47°E), located approximately 20 km from the lidar observation site. The same threshold is applied when analyzing the impact of the relative humidity.

The backward-trajectory data were analyzed using the Hybrid Single Particle Lagrangian Integrated Trajectory Model (HYSPLIT) [44]. This model uses meteorological data from the Global Data Assimilation System (GDAS) provided by the National Oceanic and Atmospheric Administration (NOAA) to derive the movement trajectories of aerosols at specific locations over a given time period. These trajectories can be accessed via the HYSPLIT READY website (https://www.ready.noaa.gov/HYSPLIT.php (accessed on 30 June 2024)).

3. Results

3.1. Overview

Haze events occurred in Beijing during the spring of 2024, including multiple pollution and dust intrusion incidents. When dust or pollution reaches the ground, it results in significant changes in the PM₁₀ and PM_2.5 mass concentrations [45]. Figure 1c shows the distribution of the PM₁₀ and PM_2.5 mass concentrations and their ratios during the spring of 2024. The number of days with high PM₁₀ and PM_2.5 mass concentrations in March and April was higher than in May, with a decreasing trend in PM₁₀ and PM_2.5 in the latter half of the spring. Severe pollution and dust events were primarily concentrated in March and April. Since the valid data from this system start at 800 m, there may be a time lag in which dust reaches the ground from this height; moreover, because lidar observations are conducted at night, the selected dates for typical cases may not correspond to the daily ground data. Thus, the daily PM₁₀ and PM_2.5 data are used for reference only. Reports on dust events can be found at (http://www.xinhuanet.com/ (accessed on 30 June 2024)). Figure 1a,b show the average results of ${\delta }_p$ and $G_f$ at 0.8–2 km during the daily observation period, with cloud interference within this height range excluded. By combining the actual weather conditions and lidar data, five typical cases are selected for detailed analysis, as shown in Figure 1. In the figure, light yellow represents a desert dust case (DD), light red represents a mining dust case (MD), green represents a pollution case, and gray represents a mixed-aerosol case; see Section 4. The central values of each type of aerosol in these typical cases will be used as ${\delta }_i$ and $G_i$ for the analysis of mixed aerosols in Section 4, with i taking values of 1, 2, and 3.

3.2. Desert Dust Case

During the spring of 2024, multiple dust events occurred in Beijing. These events varied in terms of the dust concentration, duration, and distribution height, leading to differences in the daily average PM_2.5 and PM₁₀ levels. However, the particle characteristics of dust, represented by ${\delta }_p$ and $G_f$, do not exhibit significant variations due to these differences. Therefore, the dust event on April 14 was selected for comparative analysis using both ground data and lidar data.

In this observation, the average PM_2.5 was 41 μg/m³, and the PM₁₀ was 73 μg/m³, with a PM_2.5/PM₁₀ ratio of approximately 0.56. The next day’s average PM_2.5 mass concentration was 33 μg/m³, and the PM₁₀ mass concentration was 121 μg/m³, with a PM_2.5/PM₁₀ ratio of approximately 0.27. The sharp drop in the PM concentration ratio indicates that the haze particles were primarily coarse particles. Considering the actual weather conditions at night, the dust event occurred late and continued until the next day. Figure 2 shows the spatiotemporal evolution of the aerosol backscatter coefficient ${\beta }_{355}$, the particle depolarization ratio ${\delta }_p$, and the fluorescence capacity $G_f$ from 19:00 to 21:00 UTC. The ${\delta }_p$ and $G_f$ values clearly indicate distinct aerosol layering. Chen et al. [46] reported that the backscatter Ångström index for 532 nm and 355 nm ${\mathrm{A}}_{355/532}^{\beta }$ for dust aerosols in Beijing is less than 0, indicating strong UV absorption by dust compared to other aerosols. This results in a lower backscatter coefficient for dust aerosols at 355 nm. In this case, the backscatter coefficient ${\beta }_{355}$ was less than 0.014 km⁻¹sr⁻¹ at 1~1.6 km at around 19:30 and at 0.9~2 km from 20:00 to 21:00, which are lower than the backscatter coefficients in other areas. Meanwhile, the particle depolarization ratio ${\delta }_p$ in this range was above 0.25, similar to the observations of Pan et al. [26] for dust at 355 nm in Beijing. The fluorescence capacity $G_f$ for this dust event ranged from 0.2 × 10⁻⁴ to 0.4 × 10⁻⁴. Since the classification relies on ${\delta }_p$ and $G_f$, only the vertical profiles of these two parameters are presented to illustrate their uncertainties. Figure 3b shows the average profiles of ${\delta }_p$ and $G_f$ for the period 20:55~21:10 UTC, with the error bars representing the uncertainties in the profiles. The spatiotemporal distribution indicates that the 0.8–2 km altitude range was entirely within the dust layer during this period. Radiosonde data showed that the relative humidity within 0.8~2 km at midnight was between 19% and 34%, indicating that low relative humidity does not affect $G_f$.

As shown in Figure 3a, the HYSPLIT backward-trajectory analysis indicates that at 21:00 UTC on that day, the aerosols originated from the Gobi Desert in Mongolia and passed through sandy regions in central and western North China. This is consistent with the dust source directions in East Asia reported by Sun et al. [47].

In the spring of 2024, Beijing experienced multiple dust events.

Table 1. All dust cases observed by lidar in Beijing during the spring of 2024 are included, along with their corresponding heights, times, and average $\overline{{\delta }_p}$ and $\overline{G_f}$ values.

Date	UTC Time	Height (km)	$\overline{{\mathit{\delta }}_{\mathit{p}}}$	$\overline{{\mathit{G}}_{\mathit{f}}}\times {10}^{-4}$
16 March	10:50–22:00	0.8~2.0	0.27 ± 0.02	0.46 ± 0.05
18 March	10:50–18:00	0.8~3.0	0.29 ± 0.01	0.48 ± 0.04
27 March	16:50–21:40	0.8~1.5	0.36 ± 0.02	0.48 ± 0.05
14 April	19:00–21:10	1.0~2.0	0.27 ± 0.01	0.29 ± 0.04
10 May	12:15–20:40	0.8~2.5	0.30 ± 0.02	0.31 ± 0.02

summarizes the average particle depolarization ratio $\overline{{\delta }_p}$ and average fluorescence capacity $\overline{G_f}$ during the five desert dust events observed by our lidar system. Desert dust exhibits the characteristics of a high depolarization ratio and low fluorescence capacity. The ranges of the ${\delta }_p$ and $G_f$ values in the table are used as the reference ranges for desert dust.

3.3. Mining Dust Case

In the spring of 2024, our lidar system successfully identified mining dust with strong fluorescence effects via the fluorescence channel during dust observations. For instance, on 28 March, a dust event occurred with an average PM₁₀ concentration of 404 μg/m³, the highest in Beijing during the spring of 2024, and a PM_2.5 concentration of 67 μg/m³, indicating moderate levels. The PM_2.5/PM₁₀ ratio was approximately 0.19, indicating that the haze consisted mainly of coarse particles and had a very high concentration. The average $G_f$ on that day was 0.85 × 10⁻⁴, the highest in March, and the average ${\delta }_p$ was 0.39, the highest in spring, indicating a high fluorescence capacity and high depolarization characteristics. Figure 4 shows the spatiotemporal distribution of three lidar inversion results during this observation period. At around 12:00 UTC, during the initial phase of the dust passage, the concentration of mining dust was relatively low, and, due to its strong UV absorption capacity, the aerosol backscatter coefficient ${\beta }_{355}$ was less than 0.01 km⁻¹sr⁻¹. After 14:00 UTC, the dust concentration increased, with the ${\beta }_{355}$ rising above 0.015 km⁻¹sr⁻¹. After 16:00, the concentration of dust decreased, and the ${\beta }_{355}$ was reduced. The particle depolarization ratio ${\delta }_p$ remained stable between 0.34 and 0.4 during the dust event, indicating high particle non-sphericity, similar to that of desert dust. The fluorescence capacity $G_f$ also remained stable above 0.8 × 10⁻⁴ during the dust passage, more than twice that of desert dust. Radiosonde data at 12:00 UTC indicated that the relative humidity below 3 km was less than 40%, within the low relative humidity range, suggesting that the observed $G_f$ for this dust event was close to the true value.

The vertical profiles of ${\delta }_p$ and $G_f$ at the observation point at 15:00 UTC are shown in Figure 5b. Due to the high dust concentration during this observation, the signal-to-noise ratio (SNR) of the lidar system was affected, resulting in higher parameter uncertainties for this case compared to others. The 48 h backward-trajectory analysis below 3 km at the same time, shown in Figure 5a, indicates that this dust event originated from the Gobi Desert in Mongolia. Within 36 h, the aerosol height relative to the ground remained below 1500 m, suggesting the possibility of mining dust being carried from open-pit mines.

The average particle depolarization ratio $\overline{{\delta }_p}$ for this mining dust sample was 0.35 ± 0.02, and the average fluorescence capacity $\overline{G_f}$ was (0.97 ± 0.12) × 10⁻⁴. Mining dust exhibits the characteristics of a high depolarization ratio and high fluorescence capacity. The ranges of the ${\delta }_p$ and $G_f$ values in this study are used as the reference ranges for mining dust.

3.4. Pollution Aerosol Case

Due to local and surrounding industrial pollution, air pollution aerosols in the Beijing area frequently increase the PM mass concentrations. For example, on 3 April, the average PM_2.5 concentration was 46 μg/m³ and the PM₁₀ concentration was 107 μg/m³. On this day, the average $G_f$ was 0.81 × 10⁻⁴, but the average ${\delta }_p$ was only 0.16. This haze event exhibited a high fluorescence capacity and low depolarization characteristics. Figure 6 shows the spatiotemporal distribution of three lidar inversion results during this observation period. Distinct layering can be seen in ${\delta }_p$ and $G_f$. Radiosonde data on this day showed that the relative humidity between 1 and 1.5 km was below 55%, which would not affect $G_f$. Within this height range, $G_f$ was above 0.9 × 10⁻⁴ and ${\delta }_p$ was below 0.15, indicating low particle depolarization and a high fluorescence capacity. These characteristics are similar to those of the air pollutants reported by Sugimoto et al. [48], suggesting that this layer consists of pollution aerosols. Between 1.5 and 2 km, the relative humidity was between 50% and 60%. After 12:00 UTC, ${\beta }_{355}$ decreased and ${\delta }_p$ increased to around 0.19. In this height range, $G_f$ remained stable between 0.7 × 10⁻⁴ and 0.8 × 10⁻⁴, which was lower compared to the pollution aerosols below 1.5 km.

Based on the 48 h backward-trajectory analysis at 13:00 UTC on that day, as shown in Figure 7a, the aerosols below 1.5 km traveled a short distance within 48 h and were concentrated in the heavily industrialized regions of Northeastern Hebei and Liaoning, China. This suggests that these aerosols likely contained pollution aerosols. The aerosols in the 1.5~2 km height range may have been a combination of dust from the Gobi Desert in Mongolia and pollution aerosols, resulting in polluted dust. The vertical profiles of ${\delta }_p$ and $G_f$ from 12:55 to 13:10 are shown in Figure 7b. Clear stratification is observed at 1.3 km, where, above 1.3 km, the mixing of dust and pollution aerosols leads to an increase in ${\delta }_p$ and a decrease in $G_f$.

Pollution events in the Beijing area often occur with high relative humidity [49,50]. Under this condition, ${\delta }_p$ and $G_f$ cannot be used to effectively characterize pollution aerosols’ properties. Therefore, we screened for air pollution events under low relative humidity conditions.

Table 2. All low-relative-humidity pollution cases observed by lidar in Beijing during the spring of 2024, including the corresponding heights, times, and average $\overline{{\delta }_p}$ and fluorescence capacity $\overline{G_f}$ values.

Date	UTC Time	Height (km)	$\overline{{\mathit{\delta }}_{\mathit{p}}}$	$\overline{{\mathit{G}}_{\mathit{f}}}\times {10}^{-4}$
6 March	16:00–20:00	1.0~2.0	0.13 ± 0.01	0.72 ± 0.05
3 April	11:00–14:30	0.8~1.4	0.14 ± 0.01	0.82 ± 0.09
25 April	11:50–21:00	1.0~1.5	0.14 ± 0.01	0.78 ± 0.05
7 May	12:00–17:00	0.8~1.5	0.15 ± 0.01	0.91 ± 0.06

summarizes the average particle depolarization ratio $\overline{{\delta }_p}$ and average fluorescence capacity $\overline{G_f}$ during the four air pollution events observed by our lidar system in the spring. Pollution aerosols exhibit low depolarization ratios and high fluorescence capacities. The ranges of the ${\delta }_p$ and $G_f$ values in the table are used as the reference ranges for pollution aerosols.

4. Discussion

Based on the above analysis, we identify three typical aerosol types contributing to low-altitude haze events in Beijing during the spring of 2024: desert dust, mining dust, and pollution aerosols. Figure 8 shows the distribution of ${\delta }_p$ and $G_f$ for all sample points for these three typical aerosols. Red points represent mining dust samples, yellow points represent desert dust samples, and green points represent pollution aerosol samples. The error bars indicate one standard deviation. The ranges of the ${\delta }_p$ and $G_f$ values for each aerosol type, including all sample points and their standard deviations, are used as the reference ranges for these aerosol types. In Figure 8, the light-red box represents the mining dust region, the light-yellow box represents the desert dust region, and the light-green box represents the pollution aerosol region. The remaining white areas represent the mixed-aerosol range. The discussion of the mixing of the three aerosol types will be confined to the area enclosed by the blue dashed lines in the figure.

Table 3. The ranges of the particle depolarization ratio ${\delta }_p$ and fluorescence capacity $G_f$ for the three typical aerosol types, along with the values of the central points for each type.

Aerosol Type	${\mathit{\delta }}_{\mathit{p}}$	${\mathit{G}}_{\mathit{f}}\times {10}^{-4}$	Center ${\mathit{\delta }}_{\mathit{p}}$	Center ${\mathit{G}}_{\mathit{f}}\times {10}^{-4}$
DD	0.23~0.39	0.18~0.63	0.29	0.40
MD	0.30~0.39	0.71~1.23	0.35	0.97
Pollution	0.10~0.17	0.55~1.10	0.14	0.82

summarizes the value ranges of the two parameters for the three aerosol types, with the central points for each aerosol type being the mean values of the $\overline{{\delta }_p}$ and $\overline{G_f}$ for the sample points of each type.

By substituting the central points of the three types of aerosols into Equations (6) and (7) and using an aerosol fraction step size of 0.1, we simulated the ${\delta }_p$ and $G_f$ characteristics of the mixtures of two different aerosols. The black lines in Figure 8 represent the simulated characteristic curves. The pentagon, diamond, and circle symbols indicate different aerosol mixtures. The error bars of the sample center points represent one standard deviation.

The proportions of each component in the mixed aerosols were calculated using the corresponding ${\delta }_p$ and $G_f$ values in different regions of the ${\delta }_p$-$G_f$ figure. Within the triangle enclosed by the three mixed-aerosol characteristic curves, the proportions of the three aerosol components were solved using the system of equations in Equation (5). To the right of this triangular region, where high-fluorescence-capacity mining dust and low-fluorescence-capacity desert dust are primarily mixed, the proportions of the two aerosols were calculated using Equation (9) based on $G_f$. In the upper part of the mixing triangle, where high-depolarization mining dust and low-depolarization pollution aerosols are mixed, the proportions were calculated using Equation (8) based on ${\delta }_p$. For the lower-right region of the mixing triangle, the proportions were calculated using both Equations (8) and (9) and then averaged.

For a specific aerosol type, the optical parameters of its particles are not fixed constants. Therefore, using average values as the optical parameters of a specific aerosol to solve the aforementioned equations can lead to uncertainties in the solutions. Veselovskii et al. [35] employed the Monte Carlo method to estimate these uncertainties. Similarly, in this section, we use the same method to estimate the uncertainties in the proportions of various aerosol types. In the Monte Carlo experiment, random values within a certain range of ${\delta }_p$ and $G_f$ for the three aerosol types are selected as coefficients to solve the equations. Given the limited data available in this study, to maximize the inclusion of all parameter values while ensuring that these values fell within the ranges listed in Table 3, we selected a range within twice the standard deviation of each central point for the Monte Carlo analysis. We tested multiple iteration counts, and, following the methodology in Ref. [35], we used the standard deviation of 100 experimental results to represent the uncertainty in the calculated fractions.

4.1. Mixture of Mining Dust and Desert Dust

The dust event from 27 March to 29 March resulted in high concentrations of PM₁₀ and PM_2.5. Observations obtained using the ${\delta }_p$-$G_f$ method indicated that the dust on the 27th was classified as desert dust, and the dust on the 28th was classified as mining dust. Applying this method to the dust observation on the 29th, the spatiotemporal evolution of the aerosol backscatter coefficient ${\beta }_{355}$, particle depolarization ratio ${\delta }_p$, and fluorescence capacity $G_f$ was obtained, as shown in Figure 9. The results illustrate the gradual increase in ${\delta }_p$ and $G_f$ for dust. In this observation, the average PM₁₀ concentration was 146 μg/m³, remaining at a high level, but the PM_2.5 concentration decreased to 26 μg/m³. Tian et al. [51] reported that the morphology of dust changes significantly due to the coating of hygroscopic soluble compounds, and dust particles settle quickly, which means that dust can clear pollution aerosols [26]. This explains the decrease in the PM_2.5 concentration on the 29th after the passage of the mining dust on the 28th. During the lidar observation period, the backscatter coefficient ${\beta }_{355}$ between 1 and 3 km at 12:00 UTC ranged from 0.003 to 0.004 km⁻¹sr⁻¹, and ${\delta }_p$ ranged from 0.2 to 0.35. The $G_f$ within the same region ranged from 0.3 × 10⁻⁴ to 0.4 × 10⁻⁴, characteristic of desert dust. After 12:30, the ${\beta }_{355}$ in the same height range decreased to around 0.002 km⁻¹sr⁻¹, while ${\delta }_p$ increased overall to between 0.25 and 0.4. After 14:00, $G_f$ rose to above 0.7 × 10⁻⁴ between 1 and 3 km. Radiosonde data showed that at 12:00 UTC, the relative humidity increased from 22% at 1 km to 50% at 3 km. This relative humidity range would not affect ${\delta }_p$ and $G_f$.

The backward trajectory at 14:00 UTC (Figure 10a) on this day indicates that the dust originated from the Gobi Desert in Mongolia, with a path similar to the mining dust path on March 28. Therefore, it is possible that mining dust particles were carried within the desert dust. However, the aerosol transport height within the previous 12 h was relatively low, which could have led to contamination by pollution aerosols.

The aerosol sample points for this case are plotted in Figure 10b. It can be observed that most of the sample points fall within the dust region, with some points distributed in the mining dust region and the desert dust–mining dust mixed region. A few samples are a mixture of pollution and dust. Figure 11 shows the fractions of the three types of aerosols. Similar to the previous analysis, this dust event is primarily composed of mining and desert dust. Before 12:30, desert dust predominated, accounting for over 60% of the mixture. After this, the fraction of desert dust decreased to 45%~50%, while the fraction of mining dust increased. The fraction of pollution aerosols was relatively small, possibly due to the dust’s scavenging effect on pollution aerosols. Figure 12 presents the average fractions of the three aerosol types between 13:00 and 14:00 UTC (shown as dark lines). The standard deviations were calculated using the Monte Carlo method and are depicted as shaded areas in the figure. Pollution aerosols show a relatively low fraction across the observed altitude range, with higher uncertainties. In contrast, the dust fraction is higher above 1.5 km, where the uncertainty is less than 30%. Mining dust exhibits a larger fraction below 1.5 km, with uncertainties within 50%, but above 1.5 km, the fraction decreases, and the uncertainty increases to over 50%. It is important to note that the fraction calculation algorithm ensures the non-negativity of the solution. However, when the fractions of mining dust and desert dust are large, the algorithm may calculate the fraction of pollution aerosols as close to zero, resulting in 100% uncertainty. In reality, the pollution aerosol fraction would not be zero.

4.2. Mixture of Dust and Pollution Aerosols

During transport, dust often combines with other aerosols, altering its optical properties [52]. At the center of the North China industrial zone, Beijing’s dust may mix with pollution aerosols along its path before reaching the observation point; we refer to this dust as polluted dust.

Figure 13 shows the spatiotemporal distribution of three lidar inversion results during the observation on March 9. The average PM_2.5 concentration was 38 μg/m³, and the PM₁₀ concentration was 80 μg/m³. Radiosonde data at 12:00 UTC indicated that the relative humidity between 1 and 3 km remained below 30%, which was also the case at midnight. Thus, the relative humidity was not expected to affect the ${\delta }_p$ and $G_f$ values during this period. Between 12:00 and 13:00 UTC, the backscatter coefficient ${\beta }_{355}$ at 2–3 km was approximately 0.004 km⁻¹sr⁻¹. Within the same region, ${\delta }_p$ was below 0.15 and $G_f$ exceeded 1 × 10⁻⁴, indicating the presence of pollution aerosols due to their low particle depolarization ratio and high fluorescence capacity. Below 2 km, ${\delta }_p$ ranged between 0.2 and 0.22, higher than that of pollution aerosols but lower than that of dust. ${\beta }_{355}$ was below 0.004 km⁻¹sr⁻¹, and $G_f$ was lower than that above 2 km, suggesting the presence of polluted dust in this height range. After 16:00 UTC, dust particles with a high depolarization ratio ${\delta }_p$ around 0.25 appeared above 2.5 km. In this region, ${\beta }_{355}$ decreased compared to that at 12:00 UTC. As the dust began to settle and mix with polluted dust, the $G_f$ below 3 km gradually decreased.

Based on the 48 h backward-trajectory analysis at 13:00 UTC, shown in Figure 14a, aerosols below 1 km at the observation point remained within the pollution source region reported by Zhao et al. [19] during the 48 h period. These aerosols also passed through the heavily industrially polluted regions of Northern Shanxi and Hebei, as reported by Xu et al. [53]. The overall short backward trajectory in North China indicates a slow aerosol transport speed, allowing dust from the Gobi Desert and Inner Mongolia deserts to fully combine with the pollution aerosols along the transport path.

The distribution of the aerosol samples’ ${\delta }_p$-$G_f$ for this case is shown in Figure 14b. Most of the sample points are located within the mixing triangle. The fractions of aerosol components, calculated using the mixed-aerosol algorithm from Section 2, are shown in Figure 15. During the observation period, the fraction of pollution aerosols below 1.5 km was consistently above 50%. Before 16:00 UTC, the polluted dust mainly consisted of a mixture of pollution aerosols and mining dust, with pollution aerosols being predominant and the mining dust fraction being below 40%. After 16:00 UTC, as the desert dust settled from higher altitudes, its fraction gradually increased to above 90%, while the fraction of pollution aerosols decreased, forming an aerosol layer where pollution aerosols and dust were uniformly mixed. Figure 16 shows the average fractions of the three aerosol types between 12:30 and 13:30 UTC (indicated by dark lines), with the standard deviations represented by the shaded areas. During this period, pollution aerosols were dominant, with fractions exceeding 60% at most altitudes and estimated uncertainties below 20%. A desert dust layer emerged at around 1.5 km, where the desert dust fraction slightly increased to 20% and its uncertainty decreased to 60%. At other altitudes, the desert dust fraction remained low, with uncertainties reaching up to 100%. Mining dust accounted for 20%–40% during this period, with uncertainties ranging between 10% and 40%.

5. Conclusions

This study utilized a fluorescence–Raman–Mie polarization lidar system with an emission wavelength of 355 nm to observe and analyze the low-altitude aerosols in Beijing during the spring of 2024. This lidar system can detect the fluorescence of atmospheric particles excited at a wavelength of 355 nm while simultaneously performing polarization detection at 355 nm. It successfully identified mining dust with strong fluorescence effects during dust observations. Using the ${\delta }_p$-$G_f$ method, three types of aerosol particles contributing to the haze events in Beijing were identified: desert dust, mining dust, and pollution aerosols. Their optical properties were summarized and compared. The particle depolarization ratio for desert dust ranged from 0.23 to 0.39, with a fluorescence capacity ranging from 0.18 × 10⁻⁴ to 0.63 × 10⁻⁴. Pollution aerosols had a higher fluorescence capacity and a lower depolarization ratio compared to desert dust, with a fluorescence capacity ranging from 0.60 × 10⁻⁴ to 1.10 × 10⁻⁴ and a depolarization ratio ranging from 0.10 to 0.16. The fluorescence capacity of mining dust particles ranged from 0.71 × 10⁻⁴ to 1.23 × 10⁻⁴, with a particle depolarization ratio between 0.30 and 0.39. While these particles exhibit depolarization characteristics similar to those of desert dust, their higher fluorescence capacity distinguishes them from desert dust, validating the effectiveness of the ${\delta }_p$-$G_f$ method in classifying low-altitude haze aerosols in Beijing. This also offers a new perspective for more detailed dust classification using lidar. We have developed a method for the separation of components of mixed aerosols under different conditions to discuss the fractions of each component, characterizing changes in the three types of haze aerosols at different heights and times. This provides a reference for the analysis of the interactions between dust, mining dust, and pollution aerosols.

However, haze events in Beijing often occur under high relative humidity, where the hygroscopic nature of aerosol particles can affect their fluorescence capacity and depolarization ratio, limiting the ${\delta }_p$-$G_f$ method’s ability to characterize haze aerosols. Therefore, we plan to integrate temperature and humidity lidar with a different emission wavelength for synchronous observations, adding new particle characteristic parameters while achieving high-temporal-resolution and high-spatial-resolution detection of temperature and humidity. This will allow for the study of aerosols’ hygroscopicity and the characterization of more aerosol types.