Study on the Parameters of Ice Clouds Based on 1.5 µm Micropulse Polarization Lidar

Abstract

Dust aerosols can participate in the heterogeneous nucleation process as effective ice nucleation particles, thus changing the physical properties of clouds. In this paper, we used an eye-safe 1550 nm micropulse polarization single photon lidar combined with meteorological stations, HYSPLIT backward trajectory analysis, ERA5 reanalysis data, CALIPSO, Himawari-8 and Terra-MODIS satellite data to compare the difference in cloud characteristics between dust and clean cirrus cases in Jinan from 26–29 March 2022. The study found that the aerosol affected the cloud effective radius, and the cloud top temperature impacted the properties of depolarization of dust ice clouds. According to the statistical results of the upper and lower quartiles, the depolarization ratio (DPR) range of dust cirrus on 26 March was 0.46–0.49, a similar range to the clean cirrus, while that of dust cirrus on 27 March was 0.54–0.59, which seemed much larger. Different height and temperature conditions lead to differences in the habits of ice crystals in clouds, thus changing the DPR. However, the range of the DPR between clean cirrus and dust cirrus showed no obvious difference, as the former was 0.43–0.53 and the latter was 0.46–0.59. Under the condition of higher aerosol loading, the lidar range-corrected signal (RCS) of cirrus clouds was stronger, and the cloud effective radius was 48 μm, larger than that of clean cirrus (32 μm). This may be the effect of dust on the microphysical properties of clouds. This study discusses the indirect effects of dust aerosols on cirrus clouds and the underlying mechanisms from the perspectives of microphysics and optics, which can provide more references for urban air pollution processes and aerosol-cloud interactions.

1. Introduction

Aerosols are small solid or liquid particles suspended in the atmosphere, typically represented by smoke, dust and pollen, and they play an important role in the formation of clouds [1,2]. Aerosols can directly modify the radiation balance and climate of the Earth by absorbing and scattering solar radiation [3,4] and indirectly by acting on clouds [5,6]. Furthermore, the heating effect of absorbing aerosols in the atmosphere may lead to the evaporation of cloud droplets, thereby reducing cloud cover (i.e., the semi-direct effect of aerosols) [7]. In particular, aerosols can act as cloud condensation nuclei (CCN) and ice nuclei (IN), affecting the microphysical characteristics of clouds [8]. The largest uncertainty in current climate estimates lies in aerosol-cloud interactions [9] due to the difficulty in separating the effect of aerosols on cloud radiative forcing from meteorological effects in both observations and numerical simulations [10].

The nucleation processes in clouds are usually defined as homogeneous nucleation and heterogeneous nucleation [11]. In the former case, cloud droplets are formed from activated CCN and then freeze homogeneously when the temperature decreases below −38 °C [12,13]. In the latter case, heterogeneous nucleation can proceed through aerosol particles called ice nucleation particles (INPs) [14]. INPs act directly on the ice nucleation process that determines the initial number concentration and size distribution of ice crystals [10]. The ice cloud crystal shape and size also naturally vary with cloud top temperature and atmospheric conditions [15]. There have been many parametric schemes to explain the interaction between different types of INP and clouds [16,17,18], especially dust [19,20,21]. Even with moderate concentrations of INP, heterogeneous nucleation is an important ice-production process for cirrus [22].

There are abundant observations and simulations for changes in the macroscopic and microphysical properties of clouds associated with increased aerosol loading. Using MODIS data, Meskhidze et al. [23] found that during the rainy season in the Amazon region, under polluted conditions, the convection from water clouds to ice clouds was stronger from morning to afternoon. The results of combining satellite and ground measurements with numerical simulations by Ten et al. [24] showed that, for a low AOD environment, the cloud optical depth would be higher with increasing aerosols. Some observational experiments indicated that the size distribution of ice particles near cloud tops is a function of aerosol loading [25], and studies in Asia have shown [26] that contaminated environments are associated with smaller ice particles near cloud tops [27,28]. However, there are also some findings that show larger ice particles in contaminated environments [29] or that aerosols have no significant effect on ice crystals [30]. We used lidar in combination with other observation methods to compare the differences in optical and microphysical properties of ice clouds formed under two conditions of high and low aerosol loads in Jinan, North China Plain, eastern China, and analyzed the impact of temperature on the properties of cirrus in similar dusty environments.

The North China Plain in eastern China is one of the most populous and rapidly developing economic regions, but excessive emissions of aerosol particles lead to high pollutant loads with significant radiative effects [31,32,33,34]. The terrain of Jinan is high in the south and low in the north. The weak cold air from the north and the warm air from the south often form an inversion temperature layer in Jinan [35], which blocks the updraft of the ground and makes the particles easy to stay, resulting in more serious air pollution at the low level [36]. Jinan is located in the warm temperate continental monsoon climate zone, which is often dry in spring with little rain. The dust weather in the North China Plain mainly comes from the Midwest of Inner Mongolia and the northwest of Hebei, and the surface dust principally originates from the area of origin and along the way [37]. Sand dust, soil dust and coal ash produced by coal burning constitute the main sources of local particulate pollution [38].

Lidar is widely used in meteorological and environmental monitoring due to its obvious advantages, such as high spatial and temporal resolution, high precision and large-scale profile observation capability [39]. It is often applied to measure the vertical profile of aerosol optical properties [40,41], analyze the long-range transport characteristics of aerosols [42,43], observe wind and turbulence [44,45] and measure the height of the atmospheric boundary layer [46]. Because of its multichannel and multiparameter observation capability, polarization lidar can be used to deduce the macroscopic properties of clouds [47], such as retrieving the height of the cloud base [48] and analyzing the optical characteristics of clouds [49]. Based on the above capabilities, lidar provides more possibilities for studying the influence of aerosol particles on cloud characteristics [50]. In addition, a variety of ground-based and space-borne remote sensing devices have the ability to observe aerosols, clouds and their interactions. The Aerosol Robotic Network (AERONET) can provide sparse but more accurate measurements of total aerosol optical characteristics [51] and is used to classify typical aerosols in China [52]. The comparative study of micropulse polarization lidar and Weather Research and Forecasting (WRF) models has been applied to the temporal variation in aerosol loadings and the climatological characteristics of the ocean boundary layer over the Great Barrier Reef in Australia [53]. The lidar dataset has already been shown to improve the accuracy of the WRF-Chem model in fine particulate matter (PM2.5) simulation [54]. The American High-altitude Lidar Observatory (HALO) system is an effective means for a wide range of aerosol and boundary layer height measurements under different atmospheric and surface conditions [55]. Active sensors (such as CALIOP carried by CALIPSO) can provide a variety of information regarding aerosols and clouds and are important devices for analyzing the interaction between aerosols and clouds. Wu et al. [56] evaluated CALIPSO cloud and aerosol products by combining ground-based lidar and a solar photometer and found that they showed a strong correlation and moderate relative error, which proved the reliability of using CALIPSO to analyze clouds and aerosols. Using the CALIOP profile, Costantino et al. [57] found that when aerosols mix with clouds, there is a strong correlation between aerosol concentration and the effective radius of cloud droplets. Chen et al. [58] analyzed the Japanese Himawari-8 satellite and reported that the relationship between aerosols and convective clouds was relatively stable.

Wavelengths of 532 nm and 1064 nm are common in current high-output energy lidar systems, but they are limited in urban areas due to the lack of eye safety, and the cloud and fog penetration ability is not as good as 1550 nm under the same optical power [59]. Firstly, we applied 1550 nm micropulse polarization single-photon lidar in the Jinan Quantum Lidar Network, combined with ground stations and ERA5 reanalysis data, CALIPSO, Japanese Himawari-8 Satellite and Moderate Resolution Imaging Spectroradiometer (MODIS) by Terra Satellite to observe the formation process of three ice cloud cases under different aerosol loads from 26 March to 29 March 2022. In a short period of time and relatively stable weather conditions, the physical characteristics of dust and clean cirrus clouds were compared, and the aerosol indirect effects in cirrus clouds were discussed.

The organization of this work is as follows: Section 2 introduces the equipment, data and retrieval methods used in this study. In Section 3, the distribution characteristics of aerosols and clouds are quantified, and aerosol-cloud interactions are discussed. The conclusions of this study are presented in Section 4.

2. Instruments and Methods

2.1. Instruments

The lidar supplies the vertical backscattering photon counts with spatial and temporal resolutions of 30 m and 1 s per light of sight, respectively. The diagram of the 1.5 µm wavelength polarization lidar is shown in Figure 1b. The whole lidar laser system utilizes a master oscillator power amplifier architecture. A continuous wave (CW) laser (Beogold Technology) from a distributed feedback diode (DFB, 1550.12 nm) is chopped into a pulse train using duo acousto-optic modulators (AOM, CETC-26) with a high extinction ratio (60 dB). The AOM is driven by a pulse generator, which controls the shape and pulse repetition frequency (PRF) of the laser. In the experiment, the PRF of the pulses is set to 10 kHz, which indicates that the maximum unambiguous detection range is 15 km. The weak laser pulses are fed into an erbium-doped fiber amplifier, which emits a pulse train with 70 μJ pulse energy and 100 ns pulse width. A large-mode-area fiber with a numerical aperture of 0.08 is used to increase the threshold of stimulated Brillouin scattering and to avoid the self-saturation of amplified spontaneous emission (ASE). The laser is transmitted into the atmosphere via a collimator. Atmospheric backscattering is collected by a telescope, adopting a double “D” configuration to reduce the blind zone of the telescope. As shown in Figure 1b, the two aspheric lenses are glued together with parallel optical axes to obtain a small blind zone. The absolute overlap distance is <1 km. The backscattering is collected, and a filter with 0.3 nm bandwidth (FWHM) is used to filter out the noise. After the filter, a polarization beam splitter with a 20 dB polarization ratio is used to divide backscattering into two polarization states, and two specially designed InGaAs/InP NFAD-based free-running single photon detectors (CETC-44) are used to record the backscattering photon counts [60]. The detector has a quantum efficiency of 13% and dark noise of 2500 counts per second (cps), which are much better than those of traditional commercial InGaAs single photon detectors. The multichannel scaler (MCS) converts the electric pulse from detectors to a digital signal. All the control instruction and measurement data are transmitted by the 4G module on the ARM-based circuit board. The depolarization ratio of the lidar is calibrated before the experiment [61], and the detailed parameters of the lidar are listed in

Table 1. The parameters of the 1.5 µm polarization lidar.

Parameter	Value
Wavelength	1.5 µm
Pulse duration	100 ns
Pulse energy	70 µJ
Diameter of collimator	80 mm
Diameter of telescope	70 mm
Spatial resolution	30 m
Temporal resolution	1 s
Typical detection distance	15 km
Detector quantum efficiency	13%
Dark noise counts	2500 cps
Fiber filter bandwidth (FWHM)	0.3 nm
Polarization ratio of the PBS	20 dB

.

The CALIPSO satellite was launched in April 2006, and the atmospheric measurements provided by this satellite are almost consistent with those from MODIS and the Polarization and Anisotropy of Reflectances for Atmospheric Sciences coupled with Observations from a Lidar (PARASOL) [62]. The main instrument carried on the CALIPSO satellite is a Cloud–Aerosol Lidar with Orthogonal Polarization (CALIOP), which is a dual-wavelength polarization-sensitive lidar for near-sky bottom observation [63]. The determination of aerosol types is mainly based on factors such as the attenuated backscatter coefficient, depolarization ratio, surface topography and aerosol observation height [64]. In this study, cloud and aerosol classification are obtained from CALIPSO Level 2 lidar vertical feature mask data products. They combine data from all three CALIOP channels to identify layers as clouds or aerosols and determine aerosol types and cloud ice/water phases. Its vertical resolution varies with height, with a vertical resolution of 60 m in a region below 15 km. A vertical profile of the attenuated 532 nm backscattering coefficient from the level 1 product is also used. However, polar-orbiting satellites pass over the same area only twice a day at fixed local times, making it impossible for CALIPSO to cover a sufficient time span.

The Japanese Himawari-8 satellite is a geostationary satellite that can provide continuous observations of research areas with high temporal resolution. It carries the Advanced Himawari Imager (AHI) on board, a 16-channel multispectral imager for capturing visible and infrared images of the Asia-Pacific region [65]. It is an important instrument for understanding aerosol pollution in China [66], providing comprehensive observations every 10 min. We use aerosol optical thickness (AOT) at 500 nm with a ten-minute resolution on 26 March 2022 from the Himawari-8 satellite L3 products. The official website (https://www.eorc.jaxa.jp/ptree/ (accessed on 5 July 2022)) can directly generate the image.

The hourly average meteorological data used in the study are provided by the China Meteorological Administration (CMA) (http://data.cma.cn (accessed on 15 April 2022)). The observation site is located at 36.360°N, 117.000°E, 170 m altitude. Meteorological data include hourly temperature (Temp), relative humidity (RH), wind direction (WD) and wind speed (WS). The CMA also provides hourly concentrations of PM2.5 and PM10 from 26 to 29 March 2022.

“Atmospheric reanalysis” is a process of fusing observations of various types and sources with short-term numerical weather forecasting products using a well-developed data assimilation system [67], and its role in climate monitoring applications has been widely recognized. ERA5 is ECMWF global climate reanalysis data (http://climate.copernicus.eu/products/ (accessed on 5 July 2022)), which provides a regular grid of latitude and longitude data covering the globe from 1959 to the present. Its horizontal resolution is 0.25° × 0.25°, and it is vertically divided into 37 layers with a top height of approximately 47 km. Compared to previous versions, ERA5 provides more output parameters (e.g., 100 m wind product), high-resolution output per hour and three times per hour uncertainty information [68]. Its vertical stratification data include air pressure, potential, temperature and specific humidity; surface data include air pressure, 2 m dew point temperature, potential and specific humidity. We use the vertical stratification of temperature, relative humidity and wind from 26 to 29 March 2022.

The first Moderate Resolution Imaging Spectroradiometer (MODIS) instrument was carried by the Terra satellite in 1999, which provided the possibility to study global environmental processes in the lower atmosphere. We choose the MODIS Level 2 cloud product (1 km × 1 km, MOD06_L2, Version 6), which consists of cloud optical and physical parameters [69]. In this study, the cloud parameters derived from MODIS include cloud effective radius, cloud optical depth, cloud phase and cloud-top temperature. We focus our study on investigating the changes in cloud properties caused by dust, so the parameters of two ice cloud cases on 26 March and 29 March are selected for comparison.

The Hybrid Single-Particle Lagrangian Integrated Trajectory (HYSPLIT) model of the National Oceanic and Atmospheric Administration (NOAA) is a complete system for calculating trajectories, complex dispersions and deposition simulations using the powder or particle method [70]. It is calculated by a combination of Euler and Lagrangian methods [71]. We used the Web version of the HYSPLIT model to calculate the 24 h backward trajectory of air masses arriving at Jinan to trace the source of polluting aerosols.

2.2. Method

The main lidar data products in this paper are the range-corrected signal (RCS), boundary layer height (BLH), extinction coefficient and depolarization ratio (DPR), which are common parameters used to analyze atmospheric processes for lidar [72]. We retrieve the aerosol extinction coefficient and atmospheric boundary layer height from the basic elastic scattering lidar equation. The equation of the elastic scattering micropulse lidar is as follows:

$$P(r)=P_0\frac{c\tau }{2}A\frac{\beta (r)}{r^2}\exp \left[-2{\displaystyle {\int }_0^r\sigma }\left(r^{\prime }\right){dr}^{\prime }\right].$$

(1)

$P(r)$ is the backscattering-signal power received at distance r, $P_0$ is the emitted laser power at time $t_0$, $c$ is the speed of light, τ is the pulse width of the laser, A is the effective receiver area of the telescope, $r=c(t-t_0)/2$ is the propagation distance of the received signal in the atmosphere $\beta (r)$ and $\sigma (r)$ are the backscatter coefficient and extinction coefficient in the atmosphere. In the process of lidar data analysis, the range-corrected signal (RCS) is defined as the lidar signal power multiplied by the square of the distance ($P(r)\times r^2$). This means that, except for the near field, the value of RCS is much larger than that of the lidar backscattering signal. It is easy to find the fine structure of the lidar signal. According to the shape of the RCS, an uneven atmosphere, clear atmosphere or uniform aerosol layer can be easily found. Therefore, RCS is commonly used to identify the aerosol boundary layer [73].

The laser signal is attenuated by both atmospheric aerosols and air molecules on the transmission path. The Fernald method [74] is the most common method used to retrieve the aerosol extinction coefficient. It assumes that the particle backscattering coefficient is proportional to the extinction coefficient, and it distinguishes atmospheric molecules from atmospheric aerosol particles, which overcomes the drawback that the Klet retrieval method [75] can only provide the total extinction coefficient. Existing extinction coefficient inversion algorithms are flawed, especially at the bottom of a cloud area. They present distortion, causing the extinction coefficient to be relatively small under the cloud. Considering that the optimization of the algorithm is beyond the scope of this paper, we only utilize the most common extinction coefficient inversion method. The improved algorithm can refer to the piecewise iterative inversion method [76].

The depolarization ratio (DPR) is an important parameter for identifying non-spherical particles [77,78]. It is defined as the ratio of backscattering photon counts in vertical and parallel channels ($D=B_{\perp }/B_{\parallel }$). $B_{\parallel }$ is the backscattering signal parallel to the plane of polarization of the emitted laser beam and $B_{\perp }$ is the backscattering signal perpendicular to the plane of polarization.

The boundary layer height (BLH) is often used to determine the vertical diffusion, deposition and transport of pollutants [79]. In recent years, some serious air pollution events have occurred in China, and heavy pollution is often related to low BLH. The increase in aerosol concentration will attenuate the solar radiation reaching the surface, strengthen the inversion of temperature and then weaken the turbulent diffusion through the feedback of the atmospheric boundary layer. Generally, a decrease in BLH will be conducive to the further accumulation of pollutants [80,81]. In this study, the two-dimensional matrix method is used to retrieve the aerosol boundary layer height from RCS [82]:

$$A_n(m,n)=\left[\begin{array}{cccccc}-m^{-2n}& \cdots & -m^{-n-1}& m^{-n}& \cdots & m^{-2n}\\ ⋮& \cdots & ⋮& ⋮& \cdots & ⋮\\ -m^{-n}& \cdots & -m^{-1}& m^0& \cdots & m^{-n}\\ -m^{-n-1}& \cdots & -m^{-2}& -m^{-1}& \cdots & m^{-n-1}\\ ⋮& \cdots & ⋮& ⋮& \cdots & ⋮\\ -m^{-2n}& \cdots & -m^{-n-1}& -m^{-n}& \cdots & m^{-2n}\end{array}\right],n\ge 1,$$

(2)

$$Q_n(t,h,n)={\left[\begin{array}{cccccc}{N}_{t-n\Delta t}(h+n\Delta h)& \cdots & N_{t-\Delta t}(h+n\Delta h)& N_t(h+n\Delta h)& \cdots & N_{t+\Delta t}(h+n\Delta h)\\ ⋮& & ⋮& ⋮& & ⋮\\ N_{t-n\Delta t}(h+\Delta h)& \cdots & N_{t-\Delta t}(h+\Delta h)& N_t(h+\Delta h)& & N_{t+n\Delta t}(h+\Delta h)\\ N_{t-n\Delta t}(h)& \cdots & N_{t-\Delta t}(h)& N_t(h)& & N_{t+n\Delta t}(h)\\ ⋮& & ⋮& ⋮& & ⋮\\ N_{t-n\Delta t}(h-n\Delta h)& \cdots & N_{t-\Delta t}(h-n\Delta h)& N_t(h-n\Delta h)& \cdots & N_{t+\Delta t}(h-n\Delta h)\end{array}\right]}_{(2n+1)\times (2n+1)},$$

(3)

$$M(h)={\displaystyle \sum [Q_n(t,h)\times A_n]}.$$

(4)

In these equations, m is the weight size (m > 1), and N is the gradient determined by time and height, which constitutes the gradient matrix Q. After determining the dimension size of the matrix (2n + 1) and the weight of the elements in the matrix, the two-dimensional matrix $A_n$ can be obtained, and then the new contour $M(h)$ can be calculated. The position where $M(h)$ reaches the minimum is defined as the BLH.

3. Results and Discussion

The micropulse polarization single-photon lidar at the Jinan Quantum Research Institute (36.67°N, 117.13°E) observed the formation process of three ice cloud cases from 26 to 29 March 2022. PM10, PM2.5, wind speed and wind direction over a 4-day period monitored by ground stations and RCS, DPR and extinction coefficient retrieved from lidar are shown in Figure 2. We chose the grid point (36.75°N, 117.25°E) of ERA5 data closest to the lidar station to display the local vertical profile of temperature and relative humidity with respect to water, as shown in Figure 3.

At noon on 26 March, the ground stations showed that PM10 reached a maximum value of 178 µg/m³. Meanwhile, according to the DPR observed by the polarization lidar, a high DPR value area (Area 1 in Figure 2d) appeared at an altitude of 7.5–9.8 km, from 10 to 12 o’clock, with a DPR range of 0.40–0.55 and an extinction coefficient range of 0.05–0.107 km⁻¹. The RCS was very prominent and had obvious upper and lower boundaries. These features indicated that it was an ice cloud. ERA5 data showed that in this altitude range, the average relative humidity with respect to water was 100.2%, and the mean cloud-top temperature was 230.4 K (Figure 3). Before and after this period, although the changes in the RCS and extinction coefficient were obvious, the values of DPR were between 0.3 and 0.4. According to the temperature, they were also ice clouds. From 22:00 on 26 March to 03:30 on 27 March, an ice cloud area (Area 2 in Figure 2d) with a DPR of 0.5–0.7 appeared at an altitude of approximately 10 km, with a cloud-top height of 10.8 km and cloud thickness of approximately 1.1 km. The relative humidity in this region was similar to that in Area 1, but the temperature was lower, which was 220.2 K. During this period, the surface PM10 was always above 120 µg/m³, and an abnormal aerosol boundary layer appeared near the surface, with the top height of the boundary layer reaching up to 4 km (Figure 2c). From 18:00 on 28 March to 15:00 on 29 March, several areas with DPR between 0.45 and 0.7 and extinction coefficients between 0.06 km⁻¹ and 0.24 km⁻¹ appeared at altitudes of 8–10.7 km. This was the third ice cloud case (Area 3 in Figure 2d). ERA5 showed that the relative humidity of the ice cloud was 99.7% on 29 March but low on 28 March and that the cloud-top temperature was an average of 215.0 K. During this time, the ground PM10 ranged from 80 µg/m³ to 120 µg/m³. Figure 2d shows that the DPR of the three cloud cases may have different features. We also used the cloud/aerosol data of CALIPSO, Himawari-8 and Terra-MODIS to investigate the characteristics of ice clouds or aerosols.

The three ice cloud cases were divided into two categories: one was the case of high aerosol load, and the other was the case of low aerosol load. Figure 2a illustrates the PM10, PM2.5, temperature and humidity measured by the ground station for 4 days from 26 March to 29 March 2022. It was cloudy on 26 March, and the temperature was 5–17 °C during the day. The increase in particulate matter occurred from 07:00 to 12:00 on 26 March. Before the increase in particulate matter, the surface temperature was low, the relative humidity was above 80% and the surface wind speed was close to 0. From 07: 00, PM10 rose from 81 µg/m³ to 178 µg/m³, PM2.5 from 35 µg/m³ to 54 µg/m³ and the ratio of PM10 to PM2.5 continued to grow, indicating that the pollutants were dominated by large particles. As shown in Figure 2b, at the time the pollution occurred, the wind direction started to change from southeast to southwest, shifting approximately 90 degrees, and the wind speed increased rapidly to 1.5 m/s, but it was still a breeze. It is presumed that the southwest breeze brought the particles. The appearance of pollution accompanies the rise of the sun, causing a gradual increase in temperature from 10 °C to 15 °C and a gradual decrease in relative humidity. There were some ice clouds at 9 km during 11:00–12:00, while PM10 nearly reached the peak of the day. After that, PM10 remained above 120 µg/m³ until 04:00 on the 27th, when it dropped to the level before pollution. The change in wind direction on the ground might bring in clean air and carry away pollutants. This means that there were large particles on the ground when the first and second ice clouds formed from 10:00 to 12:00 on 26 March and from 22:00 on 26 March to 03:30 on 27 March, respectively. Notably, in Figure 2e, the extinction coefficient near the surface weakened after 09:00 on 26 March, which seemed to be inconsistent with the observation of the ground weather station, and this was related to the inversion algorithm. We used the Fernald backward integral inversion method. Due to the multiple scattering of lasers in clouds and the instability of the lidar ratio [83], this algorithm resulted in an inversion error in the backward data of the mutation signal [76], which is still a common problem in lidar inversion at present but has less impact on this study. According to the RCS in Figure 2c, the signal on the ground was strong, which was consistent with the meteorological stations.

The lidar continuously monitored the range of 15 km above Jinan. After averaging the data, the time resolution was 5 min, and the distance resolution was still 30 m. Figure 2c–e represent the parameters of the vertical distribution, which are RCS, DPR and extinction coefficient. The black point in Figure 2c indicates the aerosol boundary layer height retrieved by the two-dimensional matrix method. The boundary layer has a strong diurnal cycle [84,85]. In general, daytime solar heating will lead to convective instability and increase the boundary layer [86]. Over the course of pollution on the 26th, the height of the boundary layer fluctuated, although the response of the boundary layer lagged behind; in general, the height of the boundary layer decreased, which we believe was related to the increase in pollution. Many studies have reported that the change in the boundary layer has a significantly negative correlation with the particle concentration [59,87], and the increase in particle concentration and the decrease in boundary layer height are positive feedbacks of mutual promotion [88]. However, that would require heavy pollution conditions, and Jinan saw only a slight increase in particulate matter that day. The extinction coefficient of aerosols near the ground remained above 0.013 km⁻¹ throughout the day of 26 March. The lower threshold of DPR of dust particles $\left({\mathsf{\delta }}_{\mathrm{d}1}=0.35\right)$ was introduced [89], and according to CALIOP data analysis, the DPR of dust particles was usually less than ${\mathsf{\delta }}_{\mathrm{d}2}=0.4$ [90]. Therefore, in our study, the particles in the range of ${\mathsf{\delta }}_{\mathrm{d}1}<\mathsf{\delta }<{\mathsf{\delta }}_{\mathrm{d}2}$ are considered to be pure dust particles, which indicates that the near-surface was dust aerosol. The black spots in Figure 2c show that from 18:00 on the 26th, a thick aerosol layer appeared at low altitudes, and the maximum aerosol boundary layer height identified by the matrix method reached 4 km, while the heights of the turbulent boundary layer and temperature inversion boundary layer rarely reached 4 km. Researchers once observed an ultrathick convective boundary layer with a thickness of more than 4 km in an extremely arid desert area [91], which indicated that in our study, it was likely to be related to the invasion of exogenous particles. The 24 h backward trajectory of hysplit (Figure 4c) showed that the air masses at 3 km altitude at 19:00 and 22:00 local time 26 March and 01:00 local time 27 March were all from Mongolia. The anomalous boundary layer began to decline at 0 o’clock on the 27th, and multiple layers appeared at 3 o’clock.

The RCS of Figure 2c shows that from 06:00 on 28 March, the aerosol layer at 4–8 km began to settle slowly and fell below 4 km after 18:00. According to the extinction coefficient range, it could be considered that only a small amount of aerosol remained above 4 km. Affected by this settlement, the extinction coefficient and DPR below 2 km were enhanced, meaning that the aerosol increased. Figure 2a also illustrates a slight increase in PM10 and PM2.5 on the morning of the 29th. From 18:00 on 28 March to 15:00 on 29 March, ice clouds appeared at altitudes of 8–10.7 km. Because aerosols settled in the previous period, ice clouds formed in this period under the condition of a low aerosol load.

Figure 5 shows the total attenuation backward scattering coefficient provided by CALIOP Level-1 products, cloud/aerosol classification and aerosol subtypes provided by CALIOP Level-2 VFM products. The satellite passed over Jinan at UTC 19:15 on 26 March (LT 3:15 on 27 March) and was able to observe the formation of the second ice cloud case. It obtained the vertical distribution of aerosols and clouds. Figure 5b shows the aerosol classification diagram, and Figure 5c shows the cloud/aerosol classification diagram. The red lines in the figure represent 105 km away from Jinan, and the nearest place was Zibo (36.39°N, 118.08°E), Shandong Province, which was approximately 100 km away. It was found that there were pure dust aerosols with a thickness of 1.5 km and a wide range of clouds at altitudes of 8–10.5 km. The aerosols and clouds were in contact with each other, meaning that they were likely to interact. According to the hysplit 24 h backward trajectory (Figure 4b), the aerosol at an altitude of 10 km came from the long-distance transport of the air mass in southern Xinjiang. Additionally, the Taklimakan Desert, which is known to be one of the main sources of dust in East Asia, lies in southern Xinjiang. On 25 March, many areas in southern Xinjiang experienced dusty weather (Table S1). CALIPSO also observed large areas of dust in the upper air as it passed through southern Xinjiang (Figure S1). The polluted air mass from the high altitude of southern Xinjiang passed through long-distance transport for about 20 h to the east and reached 10 km above Jinan. In Figure 5b, a large number of aerosols were distributed from the surface to 3 km, consistent with observations at ground stations. The aerosol was mainly composed of dust aerosols and a small amount of polluting dust aerosols, and the attenuated backscattering coefficient reached 0.004 km⁻¹ sr⁻¹ (Figure 5a). Although the retrieval results of CALIPSO are uncertain due to the assumption of the CALIOP lidar ratio, cloud aerosol overlap, the difference in instrument observation angle and signal noise [90,92,93,94], most studies found that the aerosol and cloud observations of CALIPSO are consistent with other observation data [95,96,97]. In our study, we believe that the observation of CALIPSO was reliable because the hysplit model could trace the pollutant sources.

The 550 nm AOT of the Himawari-8 satellite could quantify the aerosol content over Jinan. On 26 March, the AOT at LT 10:00–10:09 was greater than 0.8, and some areas reached 1.8 (Figure 6a). Figure 4a shows that the air mass passed through Afghanistan and Uzbekistan at LT 11:00 on 25 March and then the Kashgar and Hotan areas in southern Xinjiang at LT 14:00 on 25 March. From CALIPSO data (Figure S2), there were plenty of dust aerosols over Afghanistan and Uzbekistan from 3 km to 7 km. The dust aerosol over 8 km in Jinan might be the air mass carrying pollutants over Afghanistan and Uzbekistan. The AOT of 11:00–11:09 was approximately 1 (Figure 6b), while there was almost no aerosol coverage over Jinan at 16:00–16:09 (Figure 6c) on 26 March. This proved that during the first ice cloud formation period, exogenous pollution input caused an increase in aerosols above the ground. At LT 11:00, 13:00 and 14:00 on 29 March, the Himawari-8 satellite observed only a small amount of aerosol over Jinan and its surrounding areas, which was consistent with the joint observation of lidar and the ground station. This proved that the third ice cloud was formed under a low aerosol load.

The horizontal and vertical winds of ERA5 are shown in Figure 7. At the height of the first ice cloud, the average horizontal wind speed was approximately 30 m/s, mainly eastward wind. The horizontal wind at the height of the second ice cloud averaged nearly 60 m/s, and it was also eastward wind. Combined with the backward trajectory (Figure 4a,b), the eastward wind brought the polluted air mass high above Jinan. From Figure 7c, during the first ice cloud, the wind direction was upward at altitudes below 4.5 km and downward between 4.5 km and 10 km. Compared with horizontal transport, vertical activity was weak, indicating that aerosols were mainly transported from distant regions. Furthermore, there was wind convection at a low altitude from 18:00 on 26 March, and the abnormal uplift of the aerosol boundary layer may be related to this.

The cloud-top temperature of these three ice cloud cases was less than 236 K, the cloud-top height was higher than 7 km, the thickness of the cloud layer was 2 ± 1.1 km and the clouds were likely to be cirrus [98,99,100]. In our study, ice crystal detection was based on depolarization ratio measurements. The backward scattering of spherical droplets does not produce any depolarization, and non-spherical crystals introduce considerable depolarization through many internal reflection processes [101]. To investigate the depolarization ratio characteristics of dust cirrus (Area 1 and Area 2 in Figure 2d) and clean cirrus (Area 3), we selected a region from 10:00 to 12:00 in Area 1, a region from 21:00 to 03:00 in Area 2 and four regions in Area 3 for statistical comparison. The approximate T-test results showed that the mean values of the two clouds were significantly different at the 95% confidence level. The statistical results are shown in Figure 8; by selecting the upper and lower quartiles, the range of DPR of cirrus clouds with low aerosol load was 0.43–0.53, and that of dust-related cirrus clouds was 0.46–0.59, which showed no obvious difference. The DPR range of dust cirrus on 26 March was 0.46–0.49, which was similar to that of clean cirrus, and the DPR range of dust cirrus on 27 March was 0.54–0.59. However, the two dust cirrus ranges were distinctly not coincident. The cloud top temperature of DC1 was 230.4 K, that of DC2 was 220.2 K and the height of DC2 was higher. Laboratory measurements showed that temperature is the principal factor determining crystal habit [102]. Different height and temperature conditions lead to differences in the habits of ice crystals in the cloud. The shape, type and orientation of ice crystals will dominantly determine the depolarization characteristics [103], causing the difference in the DPR of two dust cirrus. The combined effect of temperature and dust may cause the difference in DPR between dust cirrus and clean cirrus not obvious. Using a multi-wavelength lidar [104,105,106,107,108,109] or single-wavelength polarization lidar combined with other data [110,111], we can obtain the number concentration, shape and size of ice crystals and dust aerosols and thus analyze the possible mechanism for the difference in DPR in detail, and more accurately and quantitatively determine the influence of dust on the microphysical and optical properties of ice clouds.

Many studies have reported that dust aerosols in the upper troposphere at cirrus height can act as effective ice nucleation particles [112] to participate in the heterogeneous nucleation process [113], and this heterogeneous ice nucleation of dust can affect the concentration and shape of ice crystals [114,115]. To further analyze the characteristics of these two types of cirrus clouds, we used MODIS data to compare the average cloud effective radius and cloud optical thickness. The MODIS carried by TERRA passed at UTC 03:00 (LT 11:00) on 26 March 2022, which just observed the first dust cirrus case. From Figure 9b,c, the cloud effective radius was 48 μm, and the cloud optical thickness was less than 5. Terra-MODIS observed the third cirrus case at UTC 03:30 (LT 11:30) on 29 March 2022. Figure 9f,g show that the cloud’s effective radius was 32 μm and the optical thickness of the cloud was less than 5. From Figure 2c, the average RCS of the first dust cirrus region was 3.49, while that of the clean cirrus was 2.54. Under the condition of high aerosol loading, the lidar echo signal of the cirrus cloud was stronger. This difference in intensity of the echo signal is related to ice particle size and number concentration. As shown in Figure 9b,f, the two effective radii of the ice cloud were observed differently by MODIS. This indicates that dust aerosols interact with clouds, thus changing the microphysical properties of clouds, which is also known as the aerosol indirect effect. Observations in the middle latitudes of the Northern Hemisphere also showed that mineral dust increased the crystal size of cirrus [113], the same as we observed, and they also found that the ice crystal number concentration decreased.

Our observation found that dust could lead to an increase in the size of cirrus crystals. However, the DPR value obtained by lidar measurement showed no clear impact. It is worth noting that the DPR depends not only on the shape and size of particles [116] but also on the horizontal direction of particles [117]. Generally, in the mid-latitude ice cloud model, the ice crystal habit is mainly composed of solid columns and plates of small particles and plate aggregates of large particles [118,119]. The mirror reflection of the horizontally oriented ice column and ice sheet makes the DPR value almost not depolarized [120]. The angle of remote sensing observation will impact the ratio of horizontally oriented ice crystals to total particles and finally affect the DPR measured by lidar.

4. Conclusions

In this paper, the formation process of three cirrus cases in the Jinan Development Zone from 26 March to 29 March 2022 was observed by near-infrared 1.5 μm aerosol polarization lidar, which has strong cloud penetration ability. The spatial and temporal variations in aerosols and clouds were analyzed using lidar and meteorological materials. With the additions of the CALIPSO, Himawari-8 and MODIS data, the interaction between dust aerosols and cirrus clouds was analyzed. We observed that the aerosol affected the cloud’s effective radius, and the cloud-top temperature impacted the properties of depolarization of dust ice clouds. The results are as follows:

1. Under the condition of higher aerosol loading, the lidar range-corrected signal (RCS) of cirrus clouds was stronger. This is because the intensity of the lidar echo signal is related to ice particle size and number concentration. We observed that the cloud effective radius (CER) of the dust cirrus is larger than that of the clean cirrus for the CER of the dust cirrus is 48 μm and the CER of the clean cirrus is 32 μm. The dust shows the aerosol indirect effect.

2. Under higher altitude and lower temperature, the second dust cirrus had a higher value of DPR than that of the first dust cirrus. We think the different temperature conditions led to differences in the habit and number concentration of ice crystals in the cloud, thus causing the difference in properties of depolarization. However, under high and low aerosol loading conditions, the DPR of two types of cirrus did not show obvious characteristics. The combined effect of temperature and dust may make the difference in DPR between dust cirrus and clean cirrus clouds not distinct.

In the future, we will combine 1550 nm lidar with 1064 nm and 532 nm lidar to obtain the lidar color ratio and particle size distribution and then determine more detailed differences in parameters between dust cirrus and clean cirrus. It is worth noting that the current work is only based on individual cases. Since the interaction between aerosols and clouds has an important impact on human production and life and global climate change, it is necessary to analyze more cases, which will help us to further understand the process of aerosol-cloud interaction and its influence mechanisms on weather and climate to reduce the uncertainty of climate estimation.