Anthropogenic Dust as a Significant Source of Ice‐Nucleating Particles in the Urban Environment

Anthropogenic dust is an important constituent of airborne particles in the urban environment but its ice nucleation activity remains poorly investigated. Here, we studied the sources and ice nucleating properties of size‐resolved particles in the urban atmosphere under mixed‐phase cloud conditions. The heat‐resistant ice nucleating particles (INPs) unexpectedly contributed ∼70% of the supermicron INPs at temperatures below −15°C. A detailed chemical composition analysis of size‐resolved particles revealed that these INPs were associated with anthropogenic dust, such as traffic‐influenced road dust. A parameterization based on supermicron particles was developed to predict the anthropogenic dust INP concentration, given their correlations on concentration and similarity in chemical compositions. Once integrated into global models, this parameterization holds the potential to assess the contribution of anthropogenic dust to INPs on a global scale. Given the considerable presence of anthropogenic dust in the atmosphere and its significant role as INPs, we suggest it may be an important aerosol source influencing cloud microphysics and warrant further investigations.


Introduction
The heterogeneous ice nucleation aided by ice nucleating particles (INPs) is important for ice crystal formation in clouds and remains one of the less understood processes in aerosol-cloud interaction (Fan et al., 2016).The INPs in the atmosphere regulate the microphysics and radiative properties of clouds by altering the number concentration and size of ice crystals, as a result, has an indirect impact on global climate (IPCC report 2021) (Legg, 2021).A fundamental understanding of the abundance and sources of atmospheric INPs is still not achieved due to their rarity and chemical complexity (Kanji et al., 2017).In the urban atmosphere, aerosols originate from both natural and anthropogenic emissions and undergo a number of aging and transformation processes.Such complexity adds to the challenges of identifying the major sources of INPs and quantifying their respective contributions.Addressing these challenges is crucial for improving the accuracy of cloud and climate models in predicting aerosol-ice cloud interactions within urban regions and their subsequent impacts on the urban climate.
The ice nucleation activities of urban aerosol and its constituents have been widely studied under mixed-phase cloud conditions (Bi et al., 2019;J. Chen et al., 2018a;J. Chen et al., 2021aJ. Chen et al., , 2021b;;K. Chen et al., 2021;Hasenkopf et al., 2016;Pereira et al., 2021;Yadav et al., 2019;Zhang et al., 2022).Biological INPs are prevalent in the urban atmosphere and are considered efficient INPs at temperatures above 15°C (Pereira et al., 2021;Yadav et al., 2019).Long-range transported dust plumes originating from desert regions can substantially enhance the INP concentrations in the urban atmosphere (Bi et al., 2019;J. Chen et al., 2021a).The poor ice nucleation behavior of black carbon (BC) or soot particles from fossil fuel combustion under mixed-phase cloud conditions has been verified by both laboratory studies (Kanji et al., 2020;Vergara-Temprado et al., 2018) and field measurements (J.Chen et al., 2018a;Zhang et al., 2022).Ice-active organic components and mineral phase (Jahl et al., 2021) have been found in biomass-burning aerosols, but their contribution to atmospheric INPs has not yet been quantified.The contribution of heat-resistant secondary organic aerosols (which remain stable at 350°C) to immersion-freezing INPs at 30°C in the urban region has recently been pointed out by Tian et al. (2022).However, this study was unable to exclude or evaluate the impact of local dust emission on the refractory INPs, which constitute a substantial portion of urban aerosol (Han et al., 2005).The INP concentrations did not show an obvious increase or decrease during heavily polluted periods in Beijing, indicating the secondary aerosols formed from the increasing anthropogenic precursors were not a major source of the observed INPs (Adams et al., 2020;Bi et al., 2019;J. Chen et al., 2018a;Zhang et al., 2022).This finding is consistent with many field studies conducted in heavily polluted regions (Hasenkopf et al., 2016;Pereira et al., 2021;Yadav et al., 2019) and in regions occasionally influenced by air pollution (Creamean et al., 2018;Wex et al., 2015), where a negligible effect on INP population from the non-dust air pollution was observed.
These studies imply that the INPs in the urban atmosphere are unlikely contributed by anthropogenic pollutants under mixed-phase cloud conditions as far as dust air pollution is not considered.Instead, the natural biological particles and the long-range transported desert dust are still considered common INP sources and there may exist unidentified INP sources on non-dust days.However, note that the atmospheric dust loading can also be changed anthropogenically through changes in land use caused by human activity, the so-called anthropogenic dust (as defined in Zender et al. (2004)).Anthropogenic dust accounts for ∼30%-70% of total dust concentrations in urban areas in recent decades (J.Chen et al., 2018a;Huang et al., 2015).On a global scale, anthropogenic contributions to atmospheric dust loads are between ∼90 and 2,000 Mt. year 1 (Webb & Pierre, 2018).It is an important constituent of airborne particles in the urban atmosphere and can be emitted from activities, such as construction, traffic-generated turbulence, and agricultural and industrial operations (Haynes et al., 2020;Philip et al., 2017).Representative anthropogenic dust species include traffic-influenced road dust (Xia et al., 2021) and soil particles from disturbed soil (Wang et al., 2018).
Despite the high mass loading of anthropogenic dust in the urban atmosphere (5∼200 Tg.year 1 reported by Xia et al. (2022)) and its direct and indirect effects on the urban climate (Philip et al., 2017;Xia et al., 2021), limited studies have investigated the ice nucleation properties of this dust species compared to other pollutants.Soil dust particles serve as INPs in a wide temperature range of 35°to 6°C (Hill et al., 2016;O'Sullivan et al., 2014;Pereira et al., 2022;Steinke et al., 2016;Tobo et al., 2014).Their ice nucleation activities are influenced by the composition, particularly the presence of the mixture of dust with biological compounds (Conen et al., 2011) and organic matters (OM) (Pereira et al., 2022;Tobo et al., 2014).Compared to ground-based soil dust, limited attention has been given to the ice nucleation activity of airborne soil dust originating from disturbed soils.Studies have pointed out that atmospheric INPs in South America can come from airborne agricultural dust (Gong et al., 2022;Testa et al., 2021).Corbin et al. (2012) showed dust particles were enriched in ice residues activated by INPs (at a temperature of 34°C and a relative humidity of 95% with respect to water) in Toronto.The primary source of the detected dust is likely vehicular resuspension from nearby roads, implying the potential contribution of road dust to INPs (Corbin et al., 2012).
In the present study, ambient particles were collected in the urban atmosphere and their ice nucleation activities were investigated to identify the potential sources of INPs.To minimize the impact of natural dust on our observation, this work was conducted during the summer in Beijing when dust events were absent.In the urban atmosphere, the size of aerosol particles varied in a wide range, primarily attributed to diverse aerosol emission sources and complex secondary formation processes.Recognizing the significant role of particle sizes in influencing both the chemical composition and ice nucleation activity of the particles, particles of different sizes were collected and investigated separately.

Aerosol Sampling and Characterization
Aerosol samples were collected at Peking University Urban Atmospheric Environment Monitoring Station (PKUERS) during the summer period from June 22 to 21 July 2020.PKUERS is located on the campus of Peking University and is positioned 20 m above the ground level.This site serves as a representative urban location affected by multiple anthropogenic emissions, including transport emissions and fossil fuel combustion (J.Chen et al., 2018a;Zhang et al., 2022).Aerosols were collected onto 47 mm diameter polycarbonate filters (Whatman, 111,107) by a Micro-Orifice Uniform Deposit Impactor (MOUDI, MSP Corporation, USA) operating at a flow rate of 30 L min 1 .MOUDI classifies and collects particles based on different aerodynamic diameters (AD).Each sample set, encompassing filter samples collected particles with different cut-off sizes, was collected for 24 hr.The detailed sampling information is provided in Table S1 in Supporting Information S1.Each set includes particles with cut-off sizes of 0.56, 1.0, 1.8, 3.2, and 5.6 μm, an aerosol population that is of interest in deriving INP parameterizations.All sizes reported in the present study represent the 50% cut-off AD (D 50 ), which corresponds to the AD of particles trapped with an efficiency of 50%.In total, eight sets (corresponding to 8 sampling days) of filter samples were collected.
An aerodynamic particle sizer (APS, model 3021, TSI) measured the number concentration of particles with AD ranging from 0.542 to 19.81 μm (N >* denotes the number concentration of particles larger than * μm, N * represents the number concentration of particle with the size of * μm).Note that APS and MOUDI both detect particles based on their aerodynamic diameter.Elemental analysis was performed using Inductively Coupled Plasma-Mass Spectrometry (ICP-MS, Bruker, aurora M90) on the same water suspensions used for ice nucleation measurements (see Section 2.3 for further details).The elements analyzed included Na, Mg, Al, K, Ca, Mn, Fe, Zn, As, Ba, Pb, and others (V, Co, Se, Sr, Mo, Tl, Bi, Th, U, Cd, Ni, Cu, Ti, Cr, and P).The mass concentrations of the organic carbon (OC) and element carbon (EC) in aerosols during the sampling period were measured by the Sunset ECOC analyzer (Sunset Lab 4) with quartz filter samples.These quartz filter samples (Whatman, 1851-865) were collected by a sequential air sampler (Model Partisol 2025i, Thermo Scientific) with a PM 2.5 cut-off size.The sampler was placed at the same location as MOUDI and operated at a flow rate of 16.7 L/min.Before being used for sampling, the quartz filters were pre-baked at 550°C for 9 hr.All collected filter samples were stored at 20°C until analysis.

Meteorological Measurement and Back Trajectories Analysis
Meteorological parameters, including the temperature (T), wind speed (WS), wind direction (WD), and relative humidity (RH) were monitored during the sampling period by an automated weather station (Model 580w, Field Environmental Instruments Inc.).The temporal evolution of these parameters is illustrated in Figure S1 in Supporting Information S1.The mean recorded temperature during this campaign is 26.8 ± 6.6°C.RH exhibited a noticeable diurnal pattern, with a mean value of 52.3% ± 19.1%.The average WS throughout the campaign was 2.5 ± 4.2 m/s, showing an overall stagnated atmospheric condition.In addition, concentrations of gas pollutants including NO and NO 2 were continuously detected by a commercial instrument (Model 42i-TLE, Thermo Scientific) with a 1-min time resolution.
The backward trajectories of air masses during the sampling period were computed using the NOAA HYSPLIT (HYbrid Single-Particle Lagrangian Integrated Trajectory) model (Rolph et al., 2017;Stein et al., 2016).Each trajectory was calculated at 6-hr intervals, resulting in four trajectories for each sample calculated at 1:00, 7:00, 13:00, and 19:00 UTC.Every trajectory ended at an altitude of 500 m above ground level, with a time resolution of 1 hr.

Ice Nucleation Measurement
The ice nucleation measurements were performed using the Peking University Ice Nucleation Array (PKU-INA), a cold-stage-based device to measure the freezing ability of droplets under mixed-phase cloud conditions (J.Chen et al., 2018b).Each filter sample was first immersed in 7 mL of distilled water and shaken by the vortex for 40 min to wash particles off.The resulting suspension was then pipetted onto the cold stage to form 90 droplets, each with a volume of 1 μL.Droplets were separated by a spacer and then sealed by a cover glass to avoid the Wegener-Bergeron-Findeisen process.Droplets were cooled to 32°C at a rate of 1°C/min.Simultaneously, the condition of all the droplets was recorded every 6 s by a high-speed camera (Q-imaging MicroPublisher 5.0 RTV, QImaging, Surrey, BC, Canada) mounted on top of the cold stage.The obtained images were processed using a customized MATLAB program to identify the freezing temperatures of the droplets based on the brightness change of each droplet upon its phase transition.
The frozen fraction (ƒ ice ) of the droplets at each temperature can be obtained using Equation 1, assuming a timeindependent ice nucleation process of droplets: where N ƒ is the number of frozen droplets at a given temperature and N t is the total number of droplets (i.e., 90 in this study).f ice of droplets generated from both sample filters and blank filters is shown in Figure S2 in Supporting Information S1.The droplets from blank filters were generated following the same procedure as that applied for sample filters and their f ice BF was used as a negative control.The higher freezing temperatures of droplets from sample filters ( 28.5°C ∼ 17°C) compared to those from blank filters ( 25.6°C ∼ 5.4°C) indicate the presence of INPs.
The number concentration of INP (N INP ) per unit volume of sampled air collected on each sample filter (i.e., at each cut-off size) is calculated based on the f ice and the total volume of the sampled air collected in each droplet (V air ) using Equation 2 (J.Chen et al., 2018b): The N INP background obtained from the blank filter was determined using the f ice BF and assumed the same V air as each sample filter (Equation 2).This background was then subtracted from the N INP of sample filters, resulting in the actual N INP collected on each filter.The total number concentration of INPs (total N INP ) for each day was calculated by adding up the N INP with different cut-off sizes at the same temperature.
The cumulative number concentration of ice active sites per unit surface area of INPs (n s ), as commonly used in various studies to describe the ice nucleation ability of particles (Connolly et al., 2009;Hiranuma et al., 2015;Niemand et al., 2012), is calculated according to Equation 3: For particles collected at each cut-off size, ) is the total surface area of the particles per unit volume of air sampled at each stage.Α is calculated based on the number concentration of particles per unit volume of air with different aerodynamic sizes (D p ) measured by APS and assuming spherical particle shape.

Overview of the INP Concentration
Figure 1a shows the total N INP (the sum of size-resolved N INP ) detected on different days as a function of temperatures.The temperature dependencies of total N INP are similar and the variations of N INP are less than one magnitude from 21°C to 5°C.This implies the INPs on different days may originate from similar aerosol sources and there are no specific strong sources of INPs presented on one particular day during the sampling period.The similar aerosol sources of different days are supported by the stagnant meteorological conditions during sampling time (Section 2.2), which indicates the aerosol was primarily influenced by stable sources from local-and regional-emissions rather than particles from long-range transportation.
The size-dependent N INP with D 50 = 0.56 μm, 1 μm,1.8μm, 3.2 μm, and D 50 = 5.6 μm is shown in Figure 1b.Each curve in Figure 1b corresponds to N INP with a specific cut-off size collected on a given day.On average, N INP in size larger than 1 μm explained 95.2% ± 4% of the total N INP over the temperature range from 20°C to 10°C, meaning that supermicron INPs dominate the INP concentration in the urban atmosphere of Beijing.The prevalence and importance of supermicron INPs are also pointed out by other studies conducted on a global scale (Gong et al., 2020;Ladino et al., 2019;Mason et al., 2016;Mitts et al., 2021;Hiranuma et al., 2021).In the urban atmosphere, supermicron particles can originate from dust particles, tire debris, biological particles and sea salt (see Wu and Boor (2021) and the references therein).The concentration of supermicron particles has a strong correlation with the INP concentrations detected in the urban environment, suggesting their significance as a source of INPs (Che et al., 2021;J. Chen et al., 2021a;K. Chen et al., 2021;Jiang et al., 2023).Further discussion on the sources of detected supermicron particles and INPs will be provided in subsequent sections.

The Contribution of Anthropogenic Dust to INPs
The extracted sample suspensions were heated to 95°C for 20 min.The total N INP detected each day decreased after heat treatment (Figure 1 The heat-resistant INPs can originate from dust particles (refer to (Hill et al., 2016;O'Sullivan et al., 2014;Perkins et al., 2019)), biogenic OM such as cellulose and pollen (Daily et al., 2022) and some unidentified OM (McCluskey et al., 2018;Tian et al., 2022).The contribution of soot particles and other inorganic components (inorganic salts) on the heat-resistant INPs are excluded due to their poor ice nucleation activities under mixedphase cloud conditions (J.Chen et al., 2018a;Kanji et al., 2020;Zhang et al., 2022).
The source of aerosols and heat-resistant INPs was explored based on the chemical composition analysis of collected particles.The contribution of OM to these INPs is considered to be minor under the determined conditions, due to the insignificant mass fraction of OM in supermicron particles in the urban atmosphere (Li et al., 2023;Zheng et al., 2023) and a poor correlation between detected heat-resistant INPs and organic carbon was observed (R 2 = 0.19, Figure S3 in Supporting Information S1).
The chemical elements of supermicron particles (D 50 = 3.2 μm) (Figure 3 and Table S2 in Supporting Information S1) show that crustal elements, including Ca (50.46% ± 2.27%), Mg (10.98% ± 0.82%), Fe (14.89% ± 1.48%) and Al (10.42% ± 1.45%), constitute a major mass fraction (88.64% ± 2.92%) of the total element mass of particles, demonstrating the strong contribution from dust particles.In contrast, fine particles (D 50 = 0.56 μm) contain less crustal elements but an increasing mass percentage of non-dust related elements (Zn, Pb, and Na) (22.76% ± 4.48%) (Figure 3 and Table S2 in Supporting Information S1), suggesting influence from non-dust anthropogenic sources.In particular, a distinguished mass fraction of K has been found in supermicron particles (4.2%) and submicron particles (25.7%).K can be generated from natural sources, such as dust particles like Kfeldspar (Atkinson et al., 2013) as well as anthropogenic sources like biomass burning and coal combustion (Yu et al., 2018).A strong correlation  Earth's Future 10.1029/2023EF003738 has been observed between the mass concentration of Ca and K in supermicron particles (R 2 = 0.66, Figure S4a in Supporting Information S1) whereas this correlation weakens in submicron particles (R 2 = 0.23, Figure S4b in Supporting Information S1), suggesting that K in supermicron particles is mainly from dust particles other than combustion processes.We therefore suggest a limited impact of biomass burning and coal combustion emissions on the observed supermicron INPs.However, we also note the contribution of the mineral phase from biomass burning aerosols to the observed INPs cannot be entirely ruled out.The distinct aerosol sources of supermicron and submicron INPs are also evident in their distinguished n s (calculated based on Equation 3) shown in Figure S5 in Supporting Information S1.As observed by many studies, if INPs are from the same aerosol source, they are likely to have similar n s values which are normalized by particle surfaces (Welti et al., 2009;Wex et al., 2015).
Dust particles in the urban atmosphere can either be long-range transported desert dust or anthropogenic dust (Han et al., 2005;Zender et al., 2004).Here, we confirmed natural desert dust has a negligible impact on heatresistant INPs, whereas anthropogenic dust is a major source, as supported by the following evidence.No obvious enhancement in the number concentration of supermicron particles (Figure S6a in Supporting Information S1) and in n s has been observed during the sampling period, which typically occurred during Asian dust storm events (Figure S6b in Supporting Information S1) [J.Chen et al., 2021a].The n s obtained in this study (3 × 10 3 to 2 × 10 5 m 2 at 15°C) were two orders of magnitude lower than those obtained during spring Asian dust events (10 5 to 10 7 m 2 at 15°C) at the same location [J.Chen et al., 2021a].The analysis of 72-hr backward trajectories (calculated at 6-hr intervals at 1:00, 7:00, 13:00, and 19:00 UTC at the altitude of 500 m) (Figure S7 in Supporting Information S1) reveals that air masses did not pass over the desert regions.Asian dust events occur especially during spring months and are not characteristic of Beijing's summer season (Shao & Dong, 2006), whereas anthropogenic dust from local emissions can contribute ∼80% of the total dust (Han et al., 2005).
The impact of a representative anthropogenic dust species, the traffic-influenced road dust, on the collected aerosols and INPs is proved by the strong correlation between the hourly mean concentration of supermicron particles and nitric oxide (NO) during the sampling time (R 2 = 0.53, Figure S8 in Supporting Information S1).Supermicron particles and NO are indicators of the dust particles and vehicle primary emissions, respectively (İbrahim Aslan, 2018;Seinfeld & Pandis, 2012).Good correlations were observed between heat-resistant N INP at 16°C and the mass percentage of Ba and Zn in supermicron particles (R 2 = 0.39 and R 2 = 0.60, Figure S9 in Supporting Information S1).Ba and Zn are two tracers for the road dust particles (Gietl et al., 2010;Harrison & Alghamdi, 2023;Peltier et al., 2011).This further supports the role of road dust in contributing to heat-resistant INPs.Other anthropogenic dust species, such as those generated from agricultural and construction activities may also contribute to the collected supermicron particles, but their influence on INPs cannot be validated in this study due to the absence of reliable tracers.Given that construction-related dust is commonly found in urban regions (Azarov et al., 2019;Yan et al., 2020), we anticipate its contribution to the observed anthropogenic dust INPs.Assessing the impact of soil dust from agricultural activities is challenging.
The n s values of the heat-resistant INPs obtained here are three orders of magnitude lower than those reported for soil dust (both original and H 2 O 2treated soil dust) measured by Tobo et al. (2014) (Figure 4).They are also substantially lower than the n s of inorganic INPs measured by Testa et al. (2021), which was presumably from land surface emission.This comparison implies that the ice nucleation activity of the airborne anthropogenic dust cannot be explained alone by the ground-based and nearsource soil dust.

Parameterization of Anthropogenic Dust INPs
Based on the aforementioned results, the supermicron heat-resistant N INP can be referred to as anthropogenic dust N INP in the urban atmosphere.The heatresistant N INP shows a strong correlation with the number concentration of supermicron particles (N >1 μm , R 2 = 0.67, Table S3 in Supporting Information S1), which can be foreseen from the strong impact of anthropogenic dust on the supermicron particles (discussed in Section 3.2).Thus, N >1 μm is used to predict the concentration of anthropogenic dust INPs.For comparison, the correlations between the anthropogenic N INP and the number concentration of particles of other size ranges (N >500 nm and N >1.8 μm ) are detailed in Table S3 in Supporting Information S1.
The parametrization for predicting anthropogenic dust INPs within a temperature range of 21°C to 7°C was developed based on N >1 μm , following the form of the DeMott et al. (2015) parameterization (referred to as D15) (listed in Table 1).The resulting parameterization exhibits a coefficient of determination (R 2 ) of 0.93 with a 95% confidence level.A comparison between the observational results and predictions is shown in Figure S10 in Supporting Information S1.The ratios of observed values to predicted values fall within a factor of 3, indicating a good prediction from the parameterization.For comparison, the estimation of heat-resistant N INP was also conducted using D15, which was developed based on N >500 nm (Figure S11 in Supporting Information S1).An overestimation of heat-resistant/total INPs by approximately 1-2 magnitudes was observed, indicating poor predictive performance.This overestimation arises due to the fact that particles larger than 500 nm cannot represent the exact size range of the measured INP species, the anthropogenic dust INPs examined in this study, as evident by the divergent aerosol sources and ice nucleation activities of supermicron and submicron particles.The overestimation of N INP in the urban environment by D15 was confirmed by J. Chen et al. (2018a) and Bi et al. (2019) as well.
Overall, the parameterization developed using N >1 μm has demonstrated its effectiveness in predicting anthropogenic dust INPs.This parameterization holds the potential to be used in regional or climate models, enabling the prediction of INPs contributed by anthropogenic dust and the assessment of their further impact on ice formation and cloud properties.Note that using heat-resistant N INP as the proxy for anthropogenic dust N INP in the urban atmosphere can cause a bias.This assumption cannot exclude that some of the measured heat-resistant INPs are contributed by OM (Daily et al., 2022;Hill et al., 2016;O'Sullivan et al., 2014;Tobo et al., 2014), although a poor correlation between OM and heat-resistant INPs and a significant dust influence on aerosol particles and INPs has been found here.On the other hand, some dust species can significantly lose their ice nucleation activity after heat treatment (Daily et al., 2020(Daily et al., , 2022)), suggesting the possibility of underestimating anthropogenic dust N INP based on the heat-lability of particles.These uncertainties underscore the need for further investigations on measuring the abundance of anthropogenic dust and its contribution to atmospheric INPs in the urban environment.

Conclusions and Atmospheric Implications
In this work, we have gained new insight into the size characteristics and source of INPs in the urban environment, which are critical but not well-characterized parameters for global models to accurately predict INPs.Our results indicate that a substantial portion (95.2% ± 4%) of the detected INPs are supermicron particles, underscoring the significant contribution of supermicron particles to the INP population in the urban environment.This finding aligns with many of the previous studies, which recommend supermicron particles as a critical size of INPs in various environments (Burrows et al., 2022;Gong et al., 2020;Hiranuma et al., 2021;Mason et al., 2016;Mitts et al., 2021).The chemical composition of size-resolved particles reveals that anthropogenic dust particles, such as traffic-influenced road dust and construction-related dust, serve as the major source of the observed INPs, which explain ∼70% of the total INPs (heat-resistant INPs) at temperatures below 15 o C.This finding highlights the substantial role of anthropogenic dust particles as INP sources in urban environments, which has not been extensively studied before.The assessment of historical atmospheric dust loads indicates a significant contribution from anthropogenic emissions in urban environments over recent decades (S.Chen et al., 2018;Huang et al., 2015).We therefore anticipate that anthropogenic dust will become an important aerosol source for facilitating ice formation in the atmosphere by serving as INPs.To determine whether these particles can ascend to cloud altitudes and have further influence on cloud microphysics and radiative properties, additional investigations are required to examine their generation mechanism, regional and global abundance as well as the microphysical and dynamical changes of clouds after being perturbed by anthropogenic dust.Nevertheless, we have provided a new parameterization for estimating the concentration of anthropogenic dust INPs based on the concentration of supermicron particles.This parameterization establishes the relationship between observational aerosol particles and anthropogenic dust INPs, providing insights for the modeling community in simulating ice cloud microphysics in environments perturbed by anthropogenic dust.
, indicated by gray circles), indicating that proteinaceous biological INPs contributed to the observed INPs (as referenced in Christner et al. (2008)).Figure 2 furtherly shows the percentage of heat-resistant N INP , which is defined as the ratio of N INP measured after and before heat treatment at different temperatures.The contribution of heat-sensitive proteinaceous biological particles to INPs only becomes important at temperatures above 14.3°C (∼50%).The percentage of heat-resistant N INP progressively increased with decreasing temperatures.At temperatures below 15°C, the heat-resistant INPs account for ∼70% of the total INPs.

Figure 1 .
Figure 1.The total (a) and size-dependent (b) number concentration of INPs (N INP ) as a function of temperatures; The gray circles in (a) represent the number concentrations of heat-resistant INPs after being heated at 95°C for 20 min.The cut-off size in (b) corresponds to the aerodynamic diameter of particles trapped with an efficiency of 50% at a given stage.

Figure 2 .
Figure 2. The percentages of heat-resistant N INP versus temperatures; only values at temperatures above 16.5°C were calculated where droplets in all samples did not freeze completely.

Figure 3 .
Figure 3.The elemental compositions of particles with different sizes.

Figure 4 .
Figure 4.The n s of heat-resistant INPs from the present study (purple circles) and those from the soil dust measured by Tobo et al. (2014) and inorganic INPs influenced by Argentinian land surface emission in Testa et al. (2021).Note that the unit of n s inTobo et al. (2014) was cm 2 while here we converted the data to m 2 .