Case Investigation on the Influence of In‐Snow Particles' Size and Composition on the Snow Light Absorption and Albedo

The snow physical parameters are closely related to the sizes, shapes, and chemical composition of light‐absorbing particles (LAPs). By utilizing a computer‐controlled scanning electron microscope software called IntelliSEM‐EPASTM, we first report the measured size‐resolved concentration of soot, dust, and fly ash particles in fresh (wet) and aged (dry deposition) snow samples collected at an industrial city in China during and after a snowfall at intervals of 6–8 days. Due to wet scavenging by seasonal snow, soot and dust particles in snow are absorbed by 69.7% and 30.3% at wavelengths of 550 nm, lowering snow albedo by 0.0089 and 0.0039, respectively. Soot particle size increases slightly during dry deposition, whereas size‐resolved mineral dust does not undergo a significant shift in particle size. These results indicate the essentiality to involve the effects of accurate size and composition of in‐snow LAPs for a better assessment of snow light absorption and reflectance.

Consequently, dust particles can darken snow and ice surfaces either through wet or dry deposition, leading to a positive surface radiative effect and accelerating snowmelt (Kok et al., 2023). Recent studies have revealed that ash particles also contribute to the reduction of snow albedo He, 2022;Young et al., 2014). Due to limited volcanic eruptions, the effects of size and mass of ash particles on snow albedo reduction are less evaluated (Constantin et al., 2020;Flanner et al., 2021). As a result, the ash-induced snow albedo reduction is much less quantified than those of soot and dust (Dong et al., 2014(Dong et al., , 2017Ren et al., 2017).
The mass absorption cross-section (MAC) is another crucial factor in reducing snow albedo, based on particle refractive index, shape, size, and state of mixing with other particles (Dang et al., 2015;Liu et al., 2017;Niu et al., 2020;Schwarz et al., 2013). Snowmelt processes are markedly shortened by light absorption by LAP-contaminated snow, contributing to accelerated climate change at regional and global levels. Schwarz et al. (2013) demonstrate that BC can be displaced to larger sizes in snow than are typically seen in the atmosphere, resulting in an overestimation of BC MAC and forcing by 40% and 30%, respectively. The MAC of soot/ice (m 2 g −1 ) composites at 460 nm is typically enhanced by 1.8-2.1 relative to interstitial soot (Flanner et al., 2012). Despite global snow surveys and laboratory observations illustrating the role of LAPs in reducing snow albedo (Bond et al., 2013;Chylek et al., 1983;Dang et al., 2015;Schwarz et al., 2013;Wang et al., 2013;Yasunari et al., 2015), questions remain about their importance. For example, there has been no quantitative analysis of size-resolved LAPs at nano-sized resolution in snow with a detection range from 0.2 to 10 μm. LAP size and number distribution vary considerably with the different processes of dry and wet deposition involved in their accumulation or scavenging by snow. Another essential issue that remains unresolved is the effect of light absorption of size-resolved LAPs on snow-albedo feedback.
Various analytical techniques have been employed to detect snow reflectance triggered by the influence of LAPs in snow, glaciers, and ice cores. For example, the single-particle soot photometer (SP2) is primarily used to determine mass concentrations and size distributions of soot in snow without significant sensitivity to other materials (Kaspari et al., 2011;McConnell et al., 2007). However, Schwarz et al. (2012) pointed out that SP2 requires additional uncertainty assessment associated with the aerosolization of liquid water, which may cause a disproportionate bias (60% due to larger soot particle size), which is much larger than the typical value obtained from atmospheric sampling (30%). Most LAPs in a snowpack appear as external mixing with snow grains when viewed with a SEM (Aoki et al., 2000). Dust-induced reductions in snow albedo decrease by up to 30% by assuming uniform distributions of dust particles inside ice grains with a dust-effective radius of 1-5 μm (Shi et al., 2021). In ice cores from Antarctica, Spaulding et al. (2011) observed large insoluble impurities that are likely the result of dust embedded in firn or ice grains. A study conducted at mid-latitudes of NE China evaluated six sample preparation methods for detecting insoluble soot and dust in seasonal snow using TEM . Despite the advantages of these methods, a number of limitations still exist in determining size and number distributions of insoluble LAPs in snow Schwarz et al., 2012;Tanikawa et al., 2020). In particular, large amounts of insoluble particles must be identified and analyzed to ensure adequate representation, and manual particle-by-particle analysis is time-consuming and costly.
Recent studies related to atmospheric particulate matter (PM) have increasingly used the computer-controlled SEM (CCSEM) to identify the morphology and composition of ambient particles. IntelliSEM-EPAS TM (Environmental Particle Analysis Software), which is an advanced CCSEM software, can eliminate the potential effects of operator bias, fatigue, and subjectivity inherent in manual microscopy (Shen et al., 2016;Sparks & Wagner, 2021;Zhang & Iwasaka, 2004;Zhang et al., 2003;P. Zhao et al., 2023). The purpose of this study is to describe the IntelliSEM-EPAS technology to evaluate size-resolved LAPs in snow resulting from a combination of fresh (wet) and aged (dry deposition) snow in a heavy industrial area in NE China.

Collection of Snow Samples
As shown in Figure S1a of the Supporting Information S1, Snow samples were collected near the center of Changchun city (43°53′N, 124°13′E), the capital of Jilin Province, in an area surrounded by heavy industrial emission sources (Mu et al., 2023). In winter, air masses over Changchun generally come from the north and northwest. A meteorological station was selected as the snow sampling site to minimize anthropogenic influences ( Figure S1b in Supporting Information S1). During a heavy snowfall from 19 November to 17 December 2020, 1 freshly fallen snow (wet deposition) and 15 aged surface snow samples (dry deposition) were collected at 1-day intervals. In this area, fresh snow is relatively uniform, with a depth of >10 cm (Figure S1c in Supporting Information S1). Surface snow samples (Top 5 cm) were collected using stainless steel scoops (pre-cleaned with alcohol and exposed to UV light for 30 min) and stored in plastic "Whirl-pak" bags. Samples were collected at points 1-2 m apart. Owing to the heavy industrial emissions, 0.5-20 mL of each melted sample was filtered through 0.1 μm Nuclepore filters and stored in a refrigerator pending SEM analysis. To track the accumulation and optical effects of LAPs via wet and dry deposition, five snow samples were used at 6-8 day intervals during the snow sampling periods.

Identification of Snow-LAPs by IntelliSEM-EPAS TM
Using CCSEM technology, the IntelliSEM-EPAS software can perform rapid, accurate, and in-depth analyses of particle characteristics, as illustrated by P. Zhao et al. (2023). In this study, the major LAPs in snow were detected by IntelliSEM-EPAS in the joint laboratory for electron microscopy analysis of atmospheric particles in Beijing. In brief, IntelliSEM-EPAS provides an interface for tracking thousands of individual particles per hour with their morphology, size, and 24 chemical species by combining SEM and energy-dispersive X-ray spectroscopy (EDS) (Shen et al., 2016). All related LAPs in snow with equivalent diameters of 0.2-10 μm (>5,000 particles per sample) are continuously measured and these analysis data are stored in the IntelliSEM-EPAS database for evaluation through the "off-line" Workbench TM module. As the hub of IntelliSEM-EPAS operations, the Workbench provides a powerful, user-friendly interface to allow the analyst to review and visualize the data, spectra, and images collected by Prospector. As a result, IntelliSEM-EPAS software is capable of providing particle locations, morphology, and elemental composition of individual particles efficiently through the use of SEM and EDS (Shen et al., 2016). Furthermore, based on a k-means algorithm, insoluble particles with similar physical and chemical characteristics, as identified by IntelliSEM-EPAS with similar EDS clusters, were classified as homogeneous groups (Ault et al., 2012;Coz & Leck, 2011;Karaca et al., 2019). The method of estimating the average mass concentration of size-resolved LAPs in snow could be found in Section 2.2.2 by P. Zhao et al. (2023).
This study is unique in that it focused primarily on the size and number fraction of snow-LAPs at nano-sized resolution, namely soot, dust, and fly ash. All particles dominated by carbonates (C-rich) were defined as soot (Figures S2a-S2d in Supporting Information S1), and quartz, calcium, and iron oxide particles were categorized as dust (Figures S2e-S2h in Supporting Information S1). SEM and EDX analyses indicate that fly ash has high concentrations of Si/Al, and Fe/Ti spheres (Figures 2i and 2j), which can be attributed to fossil fuel combustion. Insoluble particles not fitting these classes were classified as "other" and not included in calculations of snow albedo reduction. A detailed description of analytical procedures and their accuracy for individual particles has been provided by P. Zhao et al. (2023).

Light Absorption and Snow Albedo Reduction
Light absorption by LAPs was calculated using Mie theory based on their complex refraction index (CRI), size distribution, and density. Flanner et al. (2021) provide a spectrally resolved CRI parameterization of soot of density 1.8 g cm −3 , and Balkanski et al. (2007) provide a CRI parameterization for dust of density 2.5 g cm −3 , enabling analysis of the effect of soot or dust size distribution on particle optical properties including MAC. The spectral albedo of soot and dust-contaminated semi-infinite snow layers under diffuse illumination was simulated using the Spectral Albedo Model for Dirty Snow (SAMDS), as described by Wang et al. (2017) and Shi et al. (2022). The specific equation can be expressed as: Where represents the wavelength, while k(λ) denotes the imaginary component of the complex refractive index of ice. The effective radius of the snow grain is represented by ef , and the density of ice is denoted by ice (= 916.7 kg m −3 ). g is the asymmetry parameter (the average cosine of the phase function of the medium), B is a factor that only depends on the snow grain shape. While MAC and represent the MAC and mass concentration (kg kg −1 ) of impurity i particles, respectively. Here we employed the values of B = 1.27 and g = 0.89 to characterize the spherical shape of snow grains . It is noted that the assumption of semi-infinite snow layers would provide an upper limit of the LAPs' absorption in snow and hence LAP-induced snow albedo reduction, compared to shallower snowpack conditions.

Identification of Major LAPs in Seasonal Snow
In this section, we describe the size-resolved soot, dust, fly ash, and other insoluble impurities in seasonal snow identified through the classification of LAPs, and all size-related parameters were statistically analyzed using lognormal distributions. Size-resolved soot fractions (64.8% by number; 57.6% by mass) are higher for diameters of <1 μm, whereas for those with diameters of >1 μm, the dust in a larger fraction comprised 61% by number and 84.5% by mass (Figure 1a and Figure S3a in Supporting Information S1). Fly ash and other insoluble impurities accounted for only 8.3% (6.2%) of the total number (mass) relative abundances, suggesting that fly ash is limited in its influence on snow albedo reduction. Based on a log-normal number fraction, the soot median diameter peaked at ∼0.48 μm (Figure 1b), compared with 0.96 and 0.70 μm for dust and fly ash (Figures 1c and 1d). This may be the first reported attempt to provide size-resolved information regarding fly ash in seasonal snow. Significantly, the fitting of the soot size distribution (Figure 1b) has overlooked the results pertaining to smaller soot particles (<0.5 μm). This is primarily due to the relatively dark appearance of soot particles when observed under scanning electron microscopy (Bai et al., 2018;Shao et al., 2022), which may restrict the statistical data related to smaller soot particles and ultimately leads to a significant bias in the final fitting results. To ensure the reliability of the fitting results, only the coarse-mode soot particles (>0.5 μm) are considered for fitting the soot size distribution in this paper. Therefore, the lognormal curve fitting for soot seems significant off for smaller size (e.g., <0.5 μm, Figure 1b). In addition, soot mass can also be estimated using a size-resolved soot number fraction combined with ambient particle average mass concentrations (P. Zhao et al., 2023). The retrieved mass fraction of soot (<1 μm) was >57.6% and was primarily responsible for the measured light absorption of 97.3% ( Figure S3a in Supporting Information S1). A log-normal distribution of soot and dust particle sizes is shown in Figure 2. As all snow samples were collected at the beginning of winter, the detection of LAPs in surface snow indicates an enrichment mechanism rather than melting amplification. The soot with a diameter greater than 0.4 μm was attributed to a slight increase in diameter at the beginning of dry deposition following fresh snowfall, decreasing 23 d later (Figure 2a). Conversely, the size-resolved dust did not change significantly in seasonal snow, but the larger diameter dust (>1 μm) had the most significant effect on light absorption (Figure 2b). For a better understanding of this phenomenon, however, more snow samples must be collected from the industrialized areas through the wet and dry deposition of more snow cases. Figure 2c shows that fly ash abundances at the diameter of 4-10 μm increased significantly via dry deposition, which may be a result of its aging processes in the atmosphere. An earlier study found no significant change in snow albedo during snowmelt caused by the volcanic ash of diameter >5 μm (Conway et al., 1996). The peak dust number fractions did not shift significantly during dry deposition, except for a slight increase in soot diameter (Figures 2d and 2e). We conclude that not only does dry deposition affect soot mass loading in seasonal snow, but it also results in a shift in diameter due to internal mixing and hydrophobicity (Ding et al., 2019). For example, the hydrophobicity of LAPs can influence their vertical redistribution in melting snow by affecting their scavenging efficiency and retention time . In Arctic regions, Doherty et al. (2010) found that soot concentrations in surface snow before melting began were double those of subsurface concentrations, which may influence the size of individual particles. Based on our results, it appears that size-resolved BC enhances the effects of surface snow scavenging, although dust particles move into the sublayers after the surface snow has melted. This result is highly consistent with a previous study by Doherty et al. (2013), who reported that soot concentration in the few centimeters of surface snowpacks could increase generally by a factor of five or more due to the snow melting. Moreover, Dong et al. (2009) found that snow dust concentration varied largely year to year, rather than clumping as in three glaciers in northwestern China, and pointed out that the process of aerosol dust deposition in snow remained. As for the fly ash, we were unable to establish log-normal number and mass fraction distributions for each snow sample due to its lower number concentrations.

Light Absorption and Snow Albedo Reduction
The results presented in Figure 3 were primarily derived from published literature. Here, we aim to compare the outcomes of different studies to depict the variability of soot and dust size distribution in snow. To provide further details on these results, the sampling locations, sampling times, sample types, and measurement methods are outlined in Table S1 of the Supporting Information S1. The soot size distributions are globally biased (Figure 3a). A log-normal distribution with the largest soot median radius (0.235 μm) found in this study (black line; Figure 3a) is larger than that typically found in snow and ice cores collected in Beijing (D. Zhao et al., 2021), East Antarctica (Kinase et al., 2020), the Arctic (Mori et al., 2019), and the Himalayan southern range . This can be attributed to the measurement method we used, which can only capture the size distribution of coarse-mode soot (>0.5 μm). Therefore, the soot MAC may be slightly underestimated when detection is limited to soot sizes of <0.5 μm (Figure 3a). Light is absorbed less efficiently by larger soot particles per unit mass than by smaller particles; the average MAC of uncoated soot in seasonal snow here was 0.7 m 2 g −1 at 550 nm, lower than those of other studies (Figure 3b). Moreover, there are slight differences in soot MAC between fresh surface snow (0.76 m 2 g −1 ) and aged snow (0.65 m 2 g −1 ). However, the CCSEM method is more accurate and efficient in detecting large-size distributions of LAPs in snow, particularly for large dust particles. We found that a single lognormal size distribution of medium diameter (D g ) for mineral dust is 0.96 μm with a standard deviation ( g ) of 2.19 (Figure 3c). Peak values of log-normal number fractions were considerably lower than those reported previously for the Sahara Desert (Di Mauro et al., 2019), glaciers of the Tibetan Plateau (Dong et al., 2016), Summit (Drab et al., 2002), and central Japan (Osada et al., 2004). These lower values may be due to the presence of smaller coarse particles in urban or local airborne dust compared with those from long-range sources (Dang et al., 2015;Wang et al., 2013). By using TEM, Ren et al. (2017) found that most dust particles detected in snow from the Gobi deserts are non-spherical particles. Based on the Mie theory calculations, the average MAC of dust at 550 nm was 0.03-0.04 m 2 g −1 (Figure 3d). These findings highlight the urgent need for further investigation into the size distribution of soot and dust in snow, as particle size variations can result in differences in their optical absorption. The soot and dust number distributions determined for 19 November 2020, represent the wet-snow deposition process, whereas samples collected at 6-8 day intervals represent dry deposition. It is thus possible to calculate the scavenging rate of soot and dust from the difference between number and mass fractions at each sampling interval for several days. The number fractions of soot, dust, and fly ash were estimated at a mean scavenging rate of 2.47 × 10 4 N mL −1 d −1 , 2.99 × 10 4 N mL −1 d −1 , and 7.15 × 10 2 N mL −1 d −1 in dry deposition, with mass fractions related to soot, dust, and fly ash of 0.14, 1.41, and 0.06 μg ml −1 d −1 , respectively (Figures 4a and 4b).
During winter, the temperature is far below 0°C and air pollution across NE China is heavy. Therefore, during the study period, the major LAPs in snow were essentially produced by accumulation, with minimal ablation and sublimation.
Based on the retrieved soot and dust concentrations, the albedo reduction of snow was examined using the SAMDS model with a snow grain radius of 100 μm (Figure 4c). The overall snow albedo reduction for fresh snow on 19 November was 0.0089 and 0.0039 for 210 and 1,460 ng g −1 soot and dust contamination at the 550 nm wavelength, respectively. Dry-deposited soot concentrations in surface snow on 25 November and on 3, 11, and 17 December 2020, thus resulted in an albedo reduction of 0.015, 0.0098, 0.0059, and 0.027 at 550 nm in each 6-8 day interval, respectively. Despite the dust mass being significantly higher than the soot mass, dry deposition of snow resulted in decreases of only 0.0095, 0.0061, 0.0072, and 0.019 in snow albedo (Figures 4b and 4c). There was also a reduction of broadband snow albedo (Δα) through wet and dry deposition processes caused by soot and dust with three effective snow radii of 100, 500, and 1,000 μm (Figure 4d). On 19 November (wet deposition), soot and dust did not differ significantly for fresh snow in terms of broadband snow albedo reduction for limited LAP masses with different snow grain sizes. However, the amount of snow albedo reduction increased with dry deposition and the greatest reduction occurred 29 days later on 17 December 2020. Compared with freshly fallen snow (wet deposition), the greatest albedo reduction caused by dry deposition with the observed soot masses was 0.043, 0.090, and 0.119 for effective snow radii of 100, 500, and 1,000 μm, while these values were 40.5%, 37.9%, and 36.1% greater than those caused by dust (Figure 4d), respectively. Therefore, for the range of sizes focused on here, dust replicates the effect of soot on snow albedo reduction at a mass fraction of 4.9%. Additionally, we acknowledge that snow albedo was simulated by assuming LAPs external mixed with spherical snow grains and semi-infinite snow layers in this study. Recent research has demonstrated that snow nonsphericity can interact with LAP-snow mixing, leading to 10%-40% lower LAP-induced albedo reductions for nonspherical snow shapes compared to snow spheres (Dang et al., 2016;He, Flanner, et al., 2018). In contrast, LAP-snow internal mixing may enhance snow albedo reduction by 10%-60% relative to external mixing (Dombrovsky & Kokhanovsky, 2020;He et al., 2019). Furthermore, liquid water can increase the effective grain size by enhancing grain growth and infilling voids between snow grains, while larger snow depths may reduce light absorption from the underlying ground, both of which contribute to enhancing the effect of LAPs on snow albedo (Donahue et al., 2022;He, 2022). All these factors have a considerable impact on the assessment of LAP-induced snow albedo reduction.

Limitations of IntelliSEM-EPAS TM
Although the IntelliSEM-EPAS TM technology can identify the shape and numbers of primary LAPs in seasonal snow, the detection of primary LAPs was based on assumed diameters of 0.2-10 μm. The nano-sized particles of <0.2 μm can only be obtained by manually manipulating the SEM to capture a clear image, and can not be accurately detected by IntelliSEM-EPAS TM , particularly for small-particle insoluble impurities such as soot. Therefore, the technology could be improved or combined with other instruments (e.g., SP2, scanning mobility particle sizers [SMPS], or TEM) for the detection of smaller impurities. Results for overlapping diameters (0.2-1.0 μm) can be comparatively analyzed and cross-validated. Furthermore, all carbonate-dominated particles were identified as soot rather than other LAPs (e.g., dust or organic carbon [OC]), which may overestimate their potential light absorption and radiative forcing. Insoluble OC in snow has been found to play an integral role in dominating light absorption in large-scale snow field experiments across NE China (Wang et al., 2013). Owing to the misidentification of OC at wavelengths between 450 and 600 nm, Light absorption of soot may have been overestimated here by 5%-20%. However, rather than focusing on the effects of scavenging and washout on snow albedo reduction, this study focused on the accumulation processes of heavy-pollution LAPs in snow and their feedback on snow albedo reduction. It is thus necessary to undertake more systematic studies taking into account the loss of major size-resolved LAPs during snowmelt.

Conclusions and Discussions
The size ranges of three types of snow-LAPs at nano-sized resolution in a heavy industrial area in NE China were quantified using CCSEM technology, including soot, dust, and fly ash of particles of 0.2-10 μm in size. The concentrations of soot, dust, and fly ash increased with accumulation time at a rate of 0.14, 1.41, and 0.06 μg mL −1 d −1 , respectively. The soot number fraction constitutes the largest fraction (64.8%) and is highest at diameters of <1 μm, whereas the dust number fraction leads to larger fractions (>61%) with diameters of >1 μm. We also find that the greatest albedo reduction caused by dry deposition with the observed soot masses was 0.043, 0.090, and 0.119, with these values being 40.5%, 37.9%, and 36.1% greater than that caused by dust for effective snow radii of 100, 500, and 1,000 μm, respectively. A slight increase in initial soot size was observed during dry deposition, which is possibly caused by internal mixing and hydrophobicity due to heavy soot emissions during the scavenging, sublimation, and freezing processes of snowmelt water, whereas size-resolved dust does not undergo a significant shift in particle size. In this study, fly ash concentrations in snow are much lower than those of soot and dust. It is hard to determine the concentration of size-resolved fly ash particles in the snow for each of the dry deposition processes. Since fly ash particles also reduce the albedo of snow, CCSEM techniques should be applied to examine the shapes and sizes of fly ash particles in future research.
Compared with bulk sample collection and a variety of other techniques, we emphasize the CCSEM technology is capable of providing an innovative method for detecting the exact shapes and log-normal size distribution of LAPs in snow, thereby significantly enhancing the accuracy of snow radiative forcing in model simulations. By using IntelliSEM-EPAS software, the rapid and accurate detection can be used as a benchmark for examining the processes leading to the accumulation and deposition of major LAPs in snow based on size-resolved measurements. Thus, this study describes a new approach to detecting size-resolved LAPs in the cryosphere.