Hygroscopicity of Different Types of Aerosol Particles: Case Studies Using Multi-Instrument Data in Megacity Beijing, China

Water uptake by aerosol particles alters its light-scattering characteristics significantly. However, the hygroscopicities of different aerosol particles are not the same due to their different chemical and physical properties. Such differences are explored by making use of extensive measurements concerning aerosol optical and microphysical properties made during a field experiment from December 2018 to March 2019 in Beijing. The aerosol hygroscopic growth was captured by the aerosol optical characteristics obtained from micropulse lidar, aerosol chemical composition, and aerosol particle size distribution information from ground monitoring, together with conventional meteorological measurements. Aerosol hygroscopicity behaves rather distinctly for mineral dust coarse-mode aerosol (Case I) and non-dust fine-mode aerosol (Case II) in terms of the hygroscopic enhancement factor, fβ(RH,λ532), calculated for the same humidity range. The two types of aerosols were identified by applying the polarization lidar photometer networking method (POLIPHON). The hygroscopicity for non-dust aerosol was much higher than that for dust conditions with the fβ(RH,λ532) being around 1.4 and 3.1, respectively, at the relative humidity of 86% for the two cases identified in this study. To study the effect of dust particles on the hygroscopicity of the overall atmospheric aerosol, the two types of aerosols were identified and separated by applying the polarization lidar photometer networking method in Case I. The hygroscopic enhancement factor of separated non-dust fine-mode particles in Case I had been significantly strengthened, getting closer to that of the total aerosol in Case II. These results were verified by the hygroscopicity parameter, κ (Case I non-dust particles: 0.357 ± 0.024; Case II total: 0.344 ± 0.026), based on the chemical components obtained by an aerosol chemical speciation instrument, both of which showed strong hygroscopicity. It was found that non-dust fine-mode aerosol contributes more during hygroscopic growth and that non-hygroscopic mineral dust aerosol may reduce the total hygroscopicity per unit volume in Beijing.


Introduction
The presence of aerosols has an important impact on the Earth's climate and environmental changes. It not only indirectly changes the optical and microphysical properties of clouds, but also affects global climate change by changing the radiation balance of the Earth by absorbing and scattering the total incident radiation [1]. However, aerosol loading and properties have high temporal and spatial variations, thus their interactions with the Earth's climate remain highly uncertain [2,3]. Understanding the mechanism of changes in aerosol characteristics in the atmosphere is of great significance for studying their effects on the Earth's climate and its changes. A key factor influencing aerosol properties is water vapor, which may be uptaken by aerosol particles [4,5] to alter their optical and microphysical properties, especially scattering and absorption [6].
The sensitivity of aerosol particles to water vapor is usually measured by the hygroscopic enhancement factor that is a function of the aerosol light scattering coefficient with respect to changes in relative humidity (RH) and wavelength (λ) [f β (RH, λ)]. Kuang (2016) first observed the deliquescence of aerosols in the North China Plain using the humidified nephelometers [7]. Aerosol hygroscopicity is also described by hygroscopic growth factor (GF), which is defined as the ratio of the wet particle diameter at a high RH to that of the dry particle [8] ( Meier et al., 2009). Wang (2017) used hygroscopic and volatile tandem differential mobility analyzer (H/V-TDMA) to study the moisture absorption and volatility characteristics of submeter aerosols under controlled conditions [9]. In addition, the hygroscopicity parameter (κ) is used to describe water uptake ability of different types of aerosols; the κ of mixed aerosol can be obtained from the volume fractions of different chemical components in the aerosol following the method of the Zdanovskii-Stokes-Robinson (ZSR) mixing rule [10][11][12]. Some researchers also use f β (RH, λ) and GF to derive κ [13,14].
Most of these studies focused on the hygroscopicity of fine-mode spherical aerosol particles. Coarse-mode aerosols of irregular shapes such as dust and sea salt aerosols are also subject to hygroscopic growth in a more complex manner [15,16]. A water-absorbing process has been observed to occur on the surface layer of mineral dust particles, changing their apparent morphological properties, thus impacting dust-cloud interactions [17]. Laboratory studies help gain a deep insight into heterogeneous chemistry processes taking place on dust particles as a result of water-induced swelling and the ensuing changes in their optical properties [18]. While there have existed many studies on aerosol hygroscopicity using in situ and laboratory observation data, there have been relatively few studies in open atmospheric environment conditions in China whose aerosols are heavy and complex with significant impact on regional climate and environment [3,19]. With the increasing amount of lidar measurements, attempts have been made to derive the hygroscopicity of atmospheric aerosols under certain atmospheric conditions [20][21][22][23][24].
In this study, we attempt to investigate differences in the hygroscopic growth between two common types of aerosols observed in Beijing, dust aerosol and non-dust aerosol, using lidar data acquired during a field experiment from December 2018 to March 2019 by choosing two somewhat representative cases. One case corresponds to background fine-mode aerosol, and the other, aerosol mixed with dust particles. Section 2 describes the instruments and methodology. Section 3 shows and discusses the results of the study. The last section is a summary of the whole study.

Experiment Site
The observation site (Nan Jiao, NJ) is situated near the fifth-ring beltway in southern Beijing (39.

Lidar Systems
This study used two kinds of lidar systems. One system was the micropulse lidar system. It can acquire aerosol vertical optical profiles. It works through a low-energy (6-8 μJ) Nd: YVO4 laser with high frequency (2500 Hz) at 532 nm [26]. It can provide continuous profile data with 30 s temporal resolution and 30 m vertical resolution. This system can make polarization measurements at the 532 nm channel. Based on the parallel and vertical corrected signals collected by the Micro Pulse Lidar (MPL), the linear volume depolarization ratio was obtained characterizing the shape of aerosol particles per unit volume. The other system used simultaneously was the Raman lidar system that has three channels of 355 nm, 532 nm, and 1064 nm. It works through a high-energy (~1.2 J) Nd: YAG laser with a 20 Hz frequency. The 7.5 vertical resolution and 15 min temporal resolution profile data can be obtained by our Raman system. Each observation used in this study was the mean of 5000 laser pulses emitted at a frequency of 20 Hz during the interval of 5 minutes [22]. The Raman lidar system receives 532 nm and 1064 nm atmospheric Mie-scattering signals and vibration-rotation Raman-scattering (355 nm, 386 nm, and 407 nm) signals. These signals are processed to infer water vapor, clouds, and aerosols. This study used extinction ( ), backscattering coefficient ( ), volume depolarization ratio profiles at 532 nm from this MPL and water vapor profiles (retrieved at 386 nm and 407 nm vibrating Raman channels) from the Raman lidar. The temporal and spatial resolutions

Lidar Systems
This study used two kinds of lidar systems. One system was the micropulse lidar system. It can acquire aerosol vertical optical profiles. It works through a low-energy (6-8 µJ) Nd: YVO4 laser with high frequency (2500 Hz) at 532 nm [26]. It can provide continuous profile data with 30 s temporal resolution and 30 m vertical resolution. This system can make polarization measurements at the 532 nm channel. Based on the parallel and vertical corrected signals collected by the Micro Pulse Lidar (MPL), the linear volume depolarization ratio was obtained characterizing the shape of aerosol particles per unit volume. The other system used simultaneously was the Raman lidar system that has three channels of 355 nm, 532 nm, and 1064 nm. It works through a high-energy (~1.2 J) Nd: YAG laser with a 20 Hz frequency. The 7.5 vertical resolution and 15 min temporal resolution profile data can be obtained by our Raman system. Each observation used in this study was the mean of 5000 laser pulses emitted at a frequency of 20 Hz during the interval of 5 min [22]. The Raman lidar system receives 532 nm and 1064 nm atmospheric Mie-scattering signals and vibration-rotation Raman-scattering (355 nm, 386 nm, and 407 nm) signals. These signals are processed to infer water vapor, clouds, and aerosols. This study used extinction (σ), backscattering coefficient (β), volume depolarization ratio profiles at 532 nm from this MPL and water vapor profiles (retrieved at 386 nm and 407 nm vibrating Raman channels) from the Raman lidar. The temporal and spatial resolutions of the data from both lidar systems were unified to 15 min and 30 m, respectively, to ensure data consistency.

Aerosol Chemical Speciation Monitor
The aerosol chemical speciation monitor (ACSM) equipped with a fine particulate matter (PM 2.5 ) inlet impactor was used to detect the chemical characteristics of the aerosols and obtain the mass concentrations of chemical components including sulfate, nitrate, ammonium, chloride, and organics in real time [27,28]. It collects samples at 15 min intervals. First, submicron aerosol particles enter into the aerodynamic particle-focusing lens. After that, the aerosol comes into the particle composition analysis chamber and the particle beam impacts a flash vaporizor with a temperature of about 600 • C and is then ionized through a 70 eV electron impact [29]. Ng (2011) and Sun (2018) provide more details about the operation of the ACSM and its applications [29,30].

Particle Sizers
A nano-scanning mobility particle sizer (Nano-SMPS, model 3756, TSI Inc., Shoreview, MN, USA), a scanning mobility particle sizer (SMPS, model 3938, TSI Inc.), and an aerodynamic particle sizer (APS, model 3321, TSI Inc.) monitored the particle size distributions of the aerosols. These instruments measure particle size distributions and number distributions in real time and in size ranges of 0.00198-0.0649 nm (Nano-SMPS), 0.01-600 nm (SMPS), and 500-20,000 nm (APS). The Nano-SMPS and SMPS use electrostatic classifiers to charge particles, classify particles, and count particles using a condensation particle counter [31]. The APS measures particle velocities, which are then related to particle size through a laser scan technique to calculate particle diameters [31]. Through the relationship between the aerodynamic particle size and Stokes' particle size [32], data from these three instruments were combined to get a full spectrum distribution of particles from ultrafine to coarse throughout the day. Therefore, the particle spectrum distribution of aerosols at any time of day can be obtained.

Radiosondes and PM Measurement Instruments
L-band GTS1 digital radiosondes are launched twice a day at the NJ site (at~1115 and~2315 UTC), collecting data with a 1 s temporal resolution [33]. The GTS1 detector, which takes~40 min to reach 10 km, provides atmospheric RH profiles with resolution of 1.0%, temperature profiles with resolution of 0.1 • C, and pressure profiles with resolution of 0.1 hPa.
The data used in this study were the calibrated and quality-controlled PM 2.5 and PM 10 data acquired from the China National Environmental Monitoring Center (CNEMC) [34,35] of the 1480 PM concentration measurement stations across China, from a station about 2 km from NJ. More details of the instruments we use are listed in Table 1. When the laser beam emitted by a lidar passes vertically through the atmosphere, the energy P(z) of the echo signal at height z is determined through the lidar equation: where P 0 is the total energy of the laser, C is the lidar system constant, β 1 (z) and β 2 (z) are, respectively, the aerosol particles and air molecules backscattering coefficients.
is the transmittance of air molecules, and α 1 (z) and α 2 (z) are, respectively, the σ of aerosol particles and air molecules at height z [36]. The Fernald inversion algorithm [36] retrieved the aerosol σ and β profiles in this study, which provided an analytic solution to Equation (1) for Mie scattering. Li (2015) provided more details about this algorithm [37].
The water vapor mixing ratio (W) is inverted using the ratio of Raman lidar echo signals of water vapor molecules (P wv ) to nitrogen molecules (P N ) [38]: where C wv is the calibration constant of the Raman lidar, calculated using radiosonde data collected at the same time as the lidar measurements. Two criteria need to be met to ensure the accuracy of the calibration constant calculation within a specified height range: there is no significant change in W, and RH is greater than 30% [39]. The term ∆ w q is the correction function for atmospheric transmittance. The parameters α aer λ N (z) and α aer λ wv (z) are the aerosol extinction coefficients of nitrogen channel and water vapor channel, respectively.
The RH profiles directly measured by radiosondes are converted to W profiles as follows:

The selection of Hygroscopic Growth Cases
The first step in selecting a hygroscopic growth case is to ensure that the increase in moisture in the aerosol is accompanied by increases in aerosol optical parameters (e.g., β) and microphysical parameters (e.g., aerosol particle size). Therefore, when studying hygroscopic growth, the aerosol β and RH must increase together in the same aerosol layer. The aerosol layer must also be uniformly mixed, otherwise the change in aerosol β may not be solely caused by a change in RH. In general, the radiosonde-determined potential temperature (θ) and W of the aerosol layer in question determine the atmospheric mixing conditions of that aerosol layer. The atmosphere is considered uniformly mixed in this study when the variation of θ is less than 2 • C and the variation of W is less than 2 g/kg [22,[40][41][42].
The hygroscopic property of the aerosol is described by f β (RH, λ), defined as follows: where β(RH, λ) and β(RH ref , λ) are the backscattering coefficients of each RH measurement and the reference RH, respectively. The range of RH selected in this study is 65-90%. Equation (5) is usually parameterized using the Hänel [43] and Kasten [44] models: where γ, a, and b are empirical parameters with γ and b indicating the hygroscopic strength. Both models were used in this study and selected was the parameterization with the best fit according to the least-squares method.

Aerosol Chemical Ion-Pairing Scheme
The hygroscopicity of aerosols is closely related to the various organic and inorganic salts in the atmosphere. The ACSM can provide the mass concentrations and volume fractions of organic and inorganic salts of non-refractory fine PM in the atmosphere [10]. To study the hygroscopic properties of aerosol particles with different chemical compositions and calculate the neutral salts from all ion molar numbers, the ion-pairing scheme was used in this study [45]. Because of its low concentration, the chlorine ion was not considered. The method is described as follows: where n refers to the mole number, the "max" and the "min" in the equations indicate maximum and minimum values, respectively. For multicomponent particles, κ is defined by the ZSR mixing rule [8]: where ε i is the volume fraction of the ith component, and κ i is the hygroscopicity parameter for the ith component [46]. The effects of organic compounds are complex, but according to previous studies, its density and κ i are assumed to be 1.4 g·cm −3 and 0.1, respectively [47]. The volume fraction of each component can be calculated using the parameters listed in Table 2. The polarization lidar photometer networking (POLIPHON) method can be used to separate the optical properties of dust from atmospheric aerosols via the depolarization ratio [48]. This method has been used successfully to separate dust aerosol from other types of aerosols such as separating dust from smoke aerosol produced by biomass burning [49,50] and separating aerosol particles from volcanic ash [51]. It was used here just to differentiate dust and non-dust optical property profiles. The process to separate dust and non-dust backscatter scattering profiles starts with the volume (δ V ) and particle (δ) linear depolarization ratios [18]: where P cr and P co represent signals measured from the parallel and vertical channels, respectively. The parameter R is the backscattering ratio (R = β m +β p β m ), δ mol is the atmospheric molecules depolarization ratio. The depolarization ratios of dust and non-dust are assumed to be relatively stable with height changes; the β profiles with dust contributions and with non-dust contributions, β d and β nd , respectively, can be calculated [46]: The dust and non-dust depolarization ratios are δ d and δ nd , respectively, and are assigned values of 0.16 and 0.05, respectively [49]. In this way, coarse-mode dust (non-spherical particles) and fine-mode non-dust particles can be separated and their properties studied separately. Figure 2 shows the time series of the mass concentrations of PM 2.5 and PM 10 , the Raman-lidar-measured W, and the 532 nm MPL-measured σ and depolarization ratio, measured in NJ, Beijing, from 4-13 February 2019. The shaded areas denote two processes of hygroscopic growth. Case I on around 1115 UTC 5 February 2019 corresponds to the hygroscopic growth with dust particles. Case II on around 1115 UTC 12 February 2019 represents the background aerosol conditions. Using lidar to study the hygroscopic growth of aerosols in the atmosphere must meet certain conditions described in Section 2.3.2 that are distinctly different.

Selection of Dust and Non-Dust Cases and Their General Properties
Remote Sens. 2020, 11, x; doi: FOR PEER REVIEW www.mdpi.com/journal/remotesensing measured W, and the 532 nm MPL-measured and depolarization ratio, measured in NJ, Beijing, from 4-13 February 2019. The shaded areas denote two processes of hygroscopic growth. Case I on around 1115 UTC 5 February 2019 corresponds to the hygroscopic growth with dust particles. Case II on around 1115 UTC 12 February 2019 represents the background aerosol conditions. Using lidar to study the hygroscopic growth of aerosols in the atmosphere must meet certain conditions described in section 2.3.2. that are distinctly different. Prior to the time of Case I, there was a transport of dust above 700 m. After the dust transmission process, a local dust event occurred during Case I, which indicated that the depolarization ratio of~0.2 ( Figure 2d) and the mass concentration difference of PM 10 and PM 2.5 were large (Figure 2a). The PM 10 mass concentration of Case I (~300 µg m −3 ) was greater than that of Case II (<100 µg m −3 ). Figure 3 shows the normalized particle number size distribution (PNSD) in the two cases. It shows there were three obvious particle modes which peaked at the sizes of 0.0018, 0.12, and 1 µm in Case I. In Case II, the normalized PNSD was generally bimodal, dominated by ultrafine particles. In light of the MPL depolarization ratio and PM results, Case I and II were identified as dust and non-dust (background condition) cases, respectively. From Figure 2b, the upper boundaries of the Case I and Case II layers were located near cloud bases, making the range of RH changes large enough [52]. The locations of the cloud bases can be determined using the 1064 nm range correction signal of the Raman lidar and radiosonde-measured RH values [53]. The weather in Beijing is generally dry, especially in spring and winter. During the entire time series, W in the two hygroscopic growth processes was about 2 g/kg higher than that of other time periods, favourable for hygroscopic growth to occur.
Remote Sens. 2020, 11, x; doi: FOR PEER REVIEW www.mdpi.com/journal/remotesensing Case II layers were located near cloud bases, making the range of RH changes large enough [52]. The locations of the cloud bases can be determined using the 1064 nm range correction signal of the Raman lidar and radiosonde-measured RH values [53]. The weather in Beijing is generally dry, especially in spring and winter. During the entire time series, in the two hygroscopic growth processes was about 2g/kg higher than that of other time periods, favourable for hygroscopic growth to occur. .002 μm to 20 μm obtained by combining aerodynamic particle sizer (APS), scanning mobility particle sizer (SMPS), and nano-scanning mobility particle sizer (Nano-SMPS) measurements. Figure 4 shows the profiles of , RH, , β532, the depolarization ratio, and the color ratio, which indicates aerosol particle size in each case. In order to ensure that the two cases had the same range of relative humidity, two atmospheric layers were selected to study the hygroscopic growth: 410-540 m for Case I and 580-1000 m for Case II and the relative humidity range of both cases was 65%-86%. The and obtained from radiosondes launched at ~1115 UTC were used to determine the layers .002 µm to 20 µm obtained by combining aerodynamic particle sizer (APS), scanning mobility particle sizer (SMPS), and nano-scanning mobility particle sizer (Nano-SMPS) measurements. Figure 4 shows the profiles of W, RH, θ, β 532 , the depolarization ratio, and the color ratio, which indicates aerosol particle size in each case. In order to ensure that the two cases had the same range of relative humidity, two atmospheric layers were selected to study the hygroscopic growth: 410-540 m for Case I and 580-1000 m for Case II and the relative humidity range of both cases was 65-86%. The W and θ obtained from radiosondes launched at~1115 UTC were used to determine the layers of well mixing. As shown in Figure 4 and Table 2, both are generally constant within each layer of each case, indicating uniformly atmospheric mixing and vertical homogeneity in both cases. Therefore, the hygroscopic growth that occurred in both cases, resulting in changes in aerosol optical properties, was most likely caused by the uptake of water [54,55].

Lidar-Estimated Hygroscopicity
When the aerosol layer is uniformly mixed, the variation of aerosol optical properties with height can be considered mainly caused by water uptake in the aerosol layer. Both the β (Figure 4a,g) and RH (Figure 4c,i) increased with altitude within each layer. Table 3 shows the specific numerical changes for the two case studies. The volume depolarization ratio, however, decreased with altitude in each layer (Figure 4e,k; Table 3), indicating that the proportion of spherical particles increased with altitude. The volume color ratio (Figure 4f,l; Table 3), which indicates the size of aerosol particles per unit volume, increased in both cases within each layer. Given that the β increased with increasing RH in the two layers, hygroscopic growth likely occurred. All these suggest that the aerosol particles gradually grew from the bottom to the top of each layer, accompanied by aerosol particles to become more spherical [18,42,56].
Remote Sens. 2020, 12, 785 9 of 17 Remote Sens. 2020, 11, x; doi: FOR PEER REVIEW www.mdpi.com/journal/remotesensing of well mixing. As shown in Figure 4 and Table 2, both are generally constant within each layer of each case, indicating uniformly atmospheric mixing and vertical homogeneity in both cases. Therefore, the hygroscopic growth that occurred in both cases, resulting in changes in aerosol optical properties, was most likely caused by the uptake of water [54,55]. When the aerosol layer is uniformly mixed, the variation of aerosol optical properties with height can be considered mainly caused by water uptake in the aerosol layer. Both the β (Figure 4a,g) and RH (Figure 4c,i) increased with altitude within each layer. Table 3 shows the specific numerical changes for the two case studies. The volume depolarization ratio, however, decreased with altitude in each layer (Figure 4e,k; Table 3), indicating that the proportion of spherical particles increased with altitude. The volume color ratio (Figure 4f,l; Table 3), which indicates the size of aerosol particles per unit volume, increased in both cases within each layer. Given that the increased with increasing RH in the two layers, hygroscopic growth likely occurred. All these suggest that the aerosol particles gradually grew from the bottom to the top of each layer, accompanied by aerosol particles to become more spherical [18,42,56].
The aerosol particle hygroscopic enhancement factors of the two cases were calculated using the Ha nel and Kasten parameterizations of Equation (4). The initial RH of both cases was 65%. Table 4 summarizes the fitting results, and Figure 5 shows the ( , ) and the fitting results for each case. The Case II hygroscopic growth factor was much stronger than that of Case I.  [22,23,41,42,54]. Both of the parameters b from the Kasten model and γ from the Ha nel model  The aerosol particle hygroscopic enhancement factors of the two cases were calculated using the Hänel and Kasten parameterizations of Equation (4). The initial RH of both cases was 65%. Table 4 summarizes the fitting results, and Figure 5 shows the f β (RH, λ) and the fitting results for each case. The Case II hygroscopic growth factor was much stronger than that of Case I. β in Case I increased 1.4 times [ f β−caseI (86%)] in the RH range of 65-86%, while β in Case II increased 3.1 times [ f β−caseII (86%)] in the same RH range, similar to the increase reported in previous studies for other regions [22,23,41,42,54]. Both of the parameters b from the Kasten model and γ from the Hänel model indicate the strength of hygroscopicity. These parameters in Case I were less than the parameters for Case II (Case I: b = 0.33, γ = 0.307; Case II: b = 1.34, γ = 1.138) [57]. More intuitively, the color ratio in Case II (Figure 4l) was significantly higher than that in Case I (Figure 4f) within the same range of RH change. Mineral dust particles themselves do not have the characteristics of hygroscopic growth. The hygroscopic growth of dust particles is often due to the change in water absorption of other substances covering the surface of dust particles [58]. Therefore, as stated before, Case I aerosol undergoing hygroscopic growth was aerosol mixed with dust. The hygroscopic growth of dust particles is often due to the change in water absorption of other substances covering the surface of dust particles [58]. Therefore, as stated before, Case I aerosol undergoing hygroscopic growth was aerosol mixed with dust.

Isolating Non-Dust Fine-Mode Aerosol Hygroscopic Properties
To better understand the hygroscopic growth and its dominant influential factors, the one-step POLIPHON method is used to differentiate contributions from dust and non-dust. Figure 6 shows the 532 nm profiles for the mean total, non-dust (mainly fine-mode) and dust (mainly coarse-mode) components. The area of high depolarization ratio in Figure 2d denotes the dust process that was first transmitted to the local area at high altitude and then the local dust process. In Case I, the irregular dust particles with high vertical depolarization ratio mainly concentrated near the ground (below 480

Isolating Non-Dust Fine-Mode Aerosol Hygroscopic Properties
To better understand the hygroscopic growth and its dominant influential factors, the one-step POLIPHON method is used to differentiate contributions from dust and non-dust. Figure 6 shows the 532 nm β profiles for the mean total, non-dust (mainly fine-mode) and dust (mainly coarse-mode) components. The area of high depolarization ratio in Figure 2d denotes the dust process that was first transmitted to the local area at high altitude and then the local dust process. In Case I, the irregular dust particles with high vertical depolarization ratio mainly concentrated near the ground (below 480 m) and contributed more to the overall β in this area. This suggests that dust was generated locally [59,60]. Within the same range of RH, the backscattering coefficient of the non-dust particles increased faster than the total backscattering coefficient. Figure 7 shows the Case I non-dust particle f β (RH, λ 532 ). For a given RH, the f β (RH, λ 532 ) of non-dust fine-mode particles in Case I is much higher than that when particles of all modes are considered. The non-dust-particle f β (RH, λ 532 ) was fitted using the Kasten parameterization. The Case I non-dust-particle b value is 0.98, which is greater than the Case I total-particle b value (0.33) and closer to the Case II total-particle b value (1.34). The hygroscopic growth intensity of Case I is similar to that of Case II, which represents mainly non-dust fine particles. This suggests that fine-mode non-dust particles play a dominant role in hygroscopic growth. Di Girolamo (2012) observed a similar phenomenon in their study on aged dust particles mixed with maritime, urban, and organic aerosols [61].
Remote Sens. 2020, 11, x; doi: FOR PEER REVIEW www.mdpi.com/journal/remotesensing faster than the total backscattering coefficient. Figure 7 shows the Case I non-dust particle ( , ). For a given RH, the ( , ) of nondust fine-mode particles in Case I is much higher than that when particles of all modes are considered. The non-dust-particle ( , ) was fitted using the Kasten parameterization. The Case I nondust-particle b value is 0.98, which is greater than the Case I total-particle b value (0.33) and closer to the Case II total-particle b value (1.34). The hygroscopic growth intensity of Case I is similar to that of Case II, which represents mainly non-dust fine particles. This suggests that fine-mode non-dust particles play a dominant role in hygroscopic growth. Di Girolamo (2012) observed a similar phenomenon in their study on aged dust particles mixed with maritime, urban, and organic aerosols [61].  Remote Sens. 2020, 11, x; doi: FOR PEER REVIEW www.mdpi.com/journal/remotesensing [59,60]. Within the same range of RH, the backscattering coefficient of the non-dust particles increased faster than the total backscattering coefficient. Figure 7 shows the Case I non-dust particle ( , ). For a given RH, the ( , ) of nondust fine-mode particles in Case I is much higher than that when particles of all modes are considered. The non-dust-particle ( , ) was fitted using the Kasten parameterization. The Case I nondust-particle b value is 0.98, which is greater than the Case I total-particle b value (0.33) and closer to the Case II total-particle b value (1.34). The hygroscopic growth intensity of Case I is similar to that of Case II, which represents mainly non-dust fine particles. This suggests that fine-mode non-dust particles play a dominant role in hygroscopic growth. Di Girolamo (2012) observed a similar phenomenon in their study on aged dust particles mixed with maritime, urban, and organic aerosols [61].

The Influence of Chemical Composition Inferred from ACSM Measurements
Removing dust particles from the mean total profile in the dust case increased the hygroscopic strength of the remaining fine-mode particles, which became closer to that of aerosol in the case of a clean atmosphere. In other words, fine-mode aerosol particles may be instrumental in hygroscopic growth. The specific chemical composition of aerosols in each mode is important when determining f β (RH) [62].
Here, the chemical compositions of fine-mode particles in Cases I and II were analyzed. Because the ACSM's inlet impactor has a diameter of 2. µm, only fine-mode aerosols were considered. Figure 8 shows the percentages of aerosol chemical compositions obtained from the ACSM measurements around the times of the two cases. The chemical compositions in the two cases were similar. Organics dominated in both cases (Case I: 43.2%; Case II: 46.1%). The percentages of sulfate in Cases I and II were 25.1% and 19.3%, and those of nitrate were 19.1% and 21.2%, respectively. The proportions of amine (Case I: 11.3%; Case II: 11.8%) and chloride (~1%) were the smallest. Because of its low percentage and relatively low hygroscopicity, chloride was neglected in this study [10,13,26]. Table 5 summarizes the volume fraction and κ values of each case. The κ values were 0.357 and 0.344 for Cases I and II, respectively, suggesting that the fine-mode aerosol hygroscopicity in both cases was relatively strong and approximately the same. This means that the hygroscopicity of aerosols in Beijing depends on fine-mode aerosol particles and that the presence of mineral dust (non-hygroscopic particles) will reduce the overall hygroscopicity to some extent [20,62].

The Influence of Chemical Composition Inferred from ACSM Measurements
Removing dust particles from the mean total profile in the dust case increased the hygroscopic strength of the remaining fine-mode particles, which became closer to that of aerosol in the case of a clean atmosphere. In other words, fine-mode aerosol particles may be instrumental in hygroscopic growth. The specific chemical composition of aerosols in each mode is important when determining ( ) [62]. Here, the chemical compositions of fine-mode particles in Cases I and II were analyzed.
Because the ACSM's inlet impactor has a diameter of 2.5 μm , only fine-mode aerosols were considered. Figure 8 shows the percentages of aerosol chemical compositions obtained from the ACSM measurements around the times of the two cases. The chemical compositions in the two cases were similar. Organics dominated in both cases (Case I: 43.2%; Case II: 46.1%). The percentages of sulfate in Cases I and II were 25.1% and 19.3%, and those of nitrate were 19.1% and 21.2%, respectively. The proportions of amine (Case I: 11.3%; Case II: 11.8%) and chloride (~1%) were the smallest. Because of its low percentage and relatively low hygroscopicity, chloride was neglected in this study [10,13,26]. Table 5 summarizes the volume fraction and κ values of each case. The κ values were 0.357 and 0.344 for Cases I and II, respectively, suggesting that the fine-mode aerosol hygroscopicity in both cases was relatively strong and approximately the same. This means that the hygroscopicity of aerosols in Beijing depends on fine-mode aerosol particles and that the presence of mineral dust (non-hygroscopic particles) will reduce the overall hygroscopicity to some extent [20,62].

Conclusions
An enhanced observation experiment was conducted in the Beijing climate observatory from December 2018 to March 2019. During this time, the hygroscopic growth characteristics of two different atmospheric conditions (dusty, non-dust) were studied using data collected by a ground-based lidar system and radiosondes, and using aerosol particle spectrum and chemical composition information. The increase in the aerosol β with increasing RH was used to determine the hygroscopic growth of the aerosols. W and θ vertical profiles provided additional constraints to the selection of hygroscopic growth cases. During hygroscopic growth, the volume depolarization ratio decreased and the color ratio increased, suggesting that aerosol particles grew larger and became more spherical. However, the hygroscopic enhancement properties of these two different types of aerosols differed. The Kasten and Hänel parameterization schemes were applied to the two selected cases using a reference RH of 65% in both cases. It was applied to fit the f β (RH, λ), parameter b at very low values around 0.33 ± 0.103 for aerosol with mineral dust (Case I), and reaching up to 1.34 ± 0.091 for aerosol of moderate background loading (Case II). The strength of the hygroscopic growth of the fine-mode clean aerosol case (Case II) was stronger than that of the aerosol mixed with coarse-mode mineral dust (Case I) due to the different types of aerosols in each case. From lidar-measured β and depolarization ratios and aerosol particle spectra obtained from particle sizers, the type of aerosol in each case was identified: a mixture of irregularly shaped dust aerosol and non-dust fine-mode aerosol (Case I) and mainly fine-mode clean aerosol (Case II). To study the contribution of each aerosol type in Case I to hygroscopic growth, dust aerosol in Case I was separated from fine-mode aerosol using the POLIPHON method. Within the same range of RH, the non-dust aerosol β and the f β (RH, λ 532 ) increased much faster than those for the dust component and the overall aerosol. The chemical composition of the background aerosol in two cases obtained by the ACSM was also used to verify the hygroscopic strength of non-dust fine-mode aerosol particles in Case I and in Case II. The hygroscopic strength kappa coefficient of the non-dust fine-mode aerosol in Case I and that for the total aerosol of Case II were similar (Case I: 0.357; Case II: 0.344), representing the general hygroscopic growth of background aerosol in the region. All the results show that fine-mode aerosol particles play a leading role in the process of moisture absorption growth, and the presence of non-hygroscopic dust particles can weaken the total hygroscopicity per unit volume to some extent.
Data Availability: Data used in this study are available from the first author upon request (wutong0207@outlook.com).
Author Contributions: Z.L. and T.W. determined the main goal of this study. T.W. carried out the study, analyzed the data, and prepared the paper with contributions from all coauthors. J.C. provided guidance on data-processing methods. Y.W. contributed to the improvement of the content of the article and made suggestions on grammatical issues and graphic formatting issues in the article. H.W. provided particle spectrum data and corresponding technical support. X.J. and S.L. provides aerosol chemical composition data and corresponding technical support. C.L. and W.W. participated in the initial observation experiments and contributed to high-quality data collection. M.C. helped modify the grammar of the article and polished the language. All authors have read and agreed to the published version of the manuscript.