Ice nucleating particles from multiple aerosol sources in the urban environment under mixed-phase cloud conditions

. Ice crystals occurring in mixed-phase clouds play a vital role in global precipitation and energy balance because of 15 the unstable equilibrium between co-existent liquid droplets and ice crystals, which affects cloud lifetime and radiative properties, as well as precipitation formation. Satellite observations proved that immersion freezing, i.e., ice formation on particles immersed within aqueous droplets, is the dominant ice nucleation (IN) pathway in mixed-phase clouds. However, the impact of anthropogenic emission on atmospheric IN in the urban environment remains ambiguous. In this study, we present in situ observations of ambient ice nucleating particle number concentration ( N INP ) measured at mixed-phase cloud conditions 20 (-30 °C, relative humidity with respect to liquid water RH w = 104%) and the physicochemical properties of ambient aerosol, including chemical composition and size distribution, at an urban site in Beijing during the traditional Chinese Spring Festival. The impact of multiple aerosol sources such as firework emissions, local traffic emissions, mineral dust and urban secondary aerosols on N INP is investigated. The results show that N INP during the dust event reaches up to 160 # L -1 , with an activation fraction (AF) of 0.0036% ± 0.0011%. During the rest of the observation, N INP is on the order of 10 -1 to 10 # L -1 , with an average 25 AF between 0.0001 to 0.0002%. No obvious dependence of N INP on the number concentration of particles larger than 500 nm ( N 500 ) or black carbon (BC) mass concentration ( m BC ) is found throughout the field observation. The results indicate that mineral dust dominates N INP , although the observation took place at an urban site with high background aerosol concentration. Meanwhile, the presence of atmospheric BC from firework and traffic emissions, along with urban aerosols formed via secondary transformation during heavily polluted periods do not influence the observed INP concentration. Our study corroborates previous laboratory and field findings that anthropogenic BC emission has a negligible effect on N INP , and that N INP is unaffected by heavy pollution in the urban environment under mixed-phase cloud conditions. of HINC in this study is identical to the setting of Lacher et al. (2017). Therefore, only particles larger than 5 μm detected by the HINC OPC are counted as ice 135 crystals. For more detailed HINC design and operating principle information, please refer to Lacher et al. (2017) and Kanji and Abbatt (2009). In this study, both warm and cold walls of HINC lined with glass-fiber filter paper were wetted with ~150 mL de-ionized water each day before the experiment start, or after running experiment for 4 hours. After draining for ~15 min, the wall temperatures of warm and cold walls would be set to -20 and -40 °C respectively to achieve desired lamina temperature (- 140 30 °C) and RH w (104%). The sampling flow rate of HINC was 0.26 LPM, surrounded by 2.57 LPM particle-free nitrogen sheath gas. During the experiment, sampling air would pass through a particle filter for 5 minutes after every 15 minutes of measurement to quantify HINC background count detected by the OPC. HINC background counts follow a Poisson distribution, based on which the average background count is determined. Average ice crystal concentration (equivalent to N INP ) of the 15-minute measurement is calculated by firstly subtracting the average background particle counts from measurement counts, and 145 secondly converting particle counts to number concentration using HINC sampling flowrate. This study reports positive N INP only, because the negative values indicate that the signal of OPC during the measurement is undistinguishable from background noise. N INP was investigated in relation to N 500 and m BC . The relationships of m ammo with N INP , as well as PM 10-2.5 with N INP , during the dust event are also presented. The results show that N INP , as well as AF of ambient 340 particles during dust event are substantially higher than all other scenarios. N INP could reach 160 # L -1 during the dust event, while it ranges from 10 -1 to 10 1 # L -1 on other days. AF during the dust event (0.0036% ± 0.0011%) is 20 to 30 times higher than clean (0.0001% ± 0.0001%) and heavily polluted days (0.0002% ± 0.0002%). During the dust event, N INP and m ammo exhibited synchronized variation, and N INP exhibited slight dependence on PM 10-2.5 ( R 2 = 0.24). The parameterization proposed by DeMott et al. (2010) predicts more than 60% of measured N INP within a factor of 2.5 during the dust event. Mass 345 concentration measurements suggest that large amounts of aerosols containing chloride and BC appeared after the celebrations on Feb. 11 th and 26 th nights due to firework emission. Meanwhile, the stagnant and humid meteorology condition provides ideal condition for secondary aerosol formation. But there is no significant difference between N INP on heavily polluted and clean days, implying the urban aerosols from multiple sources with complex chemistry might not be effective INP. Besides, the diurnal increase of m BC from petrol passenger vehicle emissions during rush hours and from diesel truck emissions after 350 20:00 (UTC+8) on clean days does not lead to distinguishable higher N INP , implying that local traffic emission also has negligible impact on N INP . Our study reveals that mineral dusts, even though present in relatively low number concentration out of the high background particle number concentration, dominate immersion INP population in the urban environment. Furthermore, our results agree with previous literature from laboratory and field studies that atmospheric BC from both local traffic and firework emissions has negligible effects on mixed-phase cloud formation, and that N INP is unaffected by heavy 355 pollution.


Introduction
Mixed-phase clouds occur where super-cooled liquid water droplets co-exist with ice crystals and are normally sustained between -38 and 0 °C in the atmosphere, with ice melting rapidly at warmer temperature and droplets freezing homogeneously 35 at colder temperature (Boucher et al., 2013;Korolev et al., 2017). The Wegener-Bergeron-Findeisen process in mixed-phase clouds favors ice crystal growth at the cost of liquid droplet evaporation (Wegener, 1911;Bergeron, 1935;Findeisen, 1938), leading to ice water content and ice crystal size change, which further result in changes of mixed-phase cloud lifetime and radiative properties, as well as global precipitation pattern (Cantrell and Heymsfield, 2005;Field and Heymsfield, 2015;Mülmenstä dt et al., 2015;Korolev et al., 2017;Heymsfield et al., 2020). Satellite observations demonstrate that the 40 predominant ice formation pathway in mixed-phase clouds is immersion freezing (e.g., Ansmann et al., 2008;de Boer et al., 2011;Silber et al., 2021). In this mode, ice nucleating particles (INPs) immersed within super-cooled aqueous droplets provide an interface that decreases the liquid-solid phase transition energy barrier and aids droplet freezing by so called heterogeneous ice nucleation (IN, Pruppacher and Klett, 2010;Vali et al., 2015;Kanji et al., 2017).
Most of the particles in highly populated urban areas originate from local emissions, including ground transportation, 45 cooking, coal and biomass burning, leading to significant production of carbonaceous particles, including organic compounds and elemental carbon, as well as inorganic salts. Apart from local emissions, regional transportation also contributes significantly to urban particle population under appropriate meteorology conditions, during which aging can significantly modify particle physicochemical properties, such as chemical composition, morphology, and mixing state (Lin et al., 2016;Sun et al., 2016;Hua et al., 2018;Zhang et al., 2020b;Lei et al., 2021;Li et al., 2021). Previous studies have confirmed that 50 several kinds of atmospheric particles, including mineral dusts, carbonaceous particles, and biogenic species, can act as immersion INP and catalyze ice crystal formation below 0 °C (Murray et al., 2012;Kanji et al., 2017 and references therein).
Previous modelling work confirmed that anthropogenic INP emission could alter the size of ice crystals in clouds and change cloud lifetime and global precipitation pattern . Yet, there is limited published direct evidence on the contribution of anthropogenic particles to ice crystal formation in highly populated areas (Knopf et al., 2010;Corbin et al., 70 2012;Chen et al., 2018;Che et al., 2019;Che et al., 2021). Knopf et al. (2010) used filter samples collected from a highly populated urban area in Mexico City and an optical IN microscopy technique to report that anthropogenic particles dominated by organic components might catalyze ice formation well below water saturation at temperature below -38 °C. Such organicrich anthropogenic particles also demonstrated ice formation potential via immersion pathway above -38 °C in their study. Corbin et al. (2012) suggested that coupling atmospheric dust, elemental carbon, and biomass burning particle concentration 75 together provided the best estimation for atmospheric INP concentration in downtown Toronto at -34 °C just below water saturation, but the share of each particle category remained unclear due to limited data. Chen et al. (2018) quantified off-line immersion INP concentration using filter samples collected every 12 hours during heavily polluted 2016 wintertime in Beijing.
Even though high level PM2.5 with complex chemical composition was sampled during a heavy pollution period in the urban area, these aerosols did not act as superior INPs, and the highest INP concentration measured at -26 °C was below 10 # L -1 , 80 similar to what was observed in remote regions such as the Swiss Alps (Boose et al., 2016a;Lacher et al., 2017). In other words, the INP concentration reported by Chen et al. (2018) was insensitive to particle number concentration and particle chemistry in an atmosphere dominated by anthropogenic emissions. The absence of a correlation of immersion INP concentration with particle number during a pollution period was further supported by Bi et al. (2019) in an online immersion INP concentration field observation at an urban site in Beijing during May to June, 2018, using a continuous flow diffusion 85 chamber (CFDC) operated above water saturation between -20 °C to -30 °C. However, Che et al. (2019) reported a positive correlation between the total atmospheric INP concentration and air pollution degree during springtime in Beijing. INP concentration was measured by a Bigg-mixing cloud chamber for one month in 2017, and the total atmospheric INP concentration could reach 1500 # L -1 at -30 °C (Che et al., 2019(Che et al., , 2021. Currently, a knowledge gap still exists on the magnitude and dominant source of ambient INPs in highly populated urban 90 area, as well as the dependence of INP concentration on anthropogenic particle emission, hampering the estimation of global atmospheric INP concentrations (Boucher et al., 2013;Seinfeld and Pandis, 2016). In this paper, we report the in situ INP concentration measured at mixed-phase clouds condition (-30 °C , relative humidity with respect to liquid water of 104% RHw = 104%) during the traditional Chinese Spring Festival at an urban site in Beijing. The correlations between immersion INP concentration, meteorology condition, and aerosol physiochemical properties are also explored.

100
The sampling site (39°59′20″N, 116°18′26″E) is located on the roof of a six-floor building (~30 m above ground level) at Peking University, which is adjacent to the north-western 4 th ring road of Beijing. The site lies about 250 m west of a busy street with heavy traffic. At the sampling site, meteorological parameters, including wind speed, wind direction, RHw, and temperature, were measured by a weather station (MetOne Inc.). The mass concentration of particulate matter (PM) with aerodynamic diameter (da) smaller than 2.5 and 10 μm (PM2.5 and PM10, respectively) were measured by a tapered element 105 oscillating microbalance (TEOM) monitor. The temporal resolutions of meteorology and PM data were 1 minute. Ambient air was sampled through a stainless tube with an inner diameter of 12.7 mm. The tube inlet was bent facing downwards (see Fig. 1) to prevent water contamination and allowed particles smaller than 20 μm to enter. The sample flow was then split and pumped into different instruments. The relative humidity of the sample stream (RHw, sample) was kept below 2% by passing through two consecutive 47 cm Nafion™ dryers (Perma Pure, LLC.) using 4 LPM nitrogen as sheath gas during 110 the experiment. A schematic of the setup is shown in Fig. 1.

Particle number size distribution
Sub-micron particle number size distribution was measured by a scanning mobility particle sizer (SMPS, model 3082, comprising a 3082 classifier, a 3081 long DMA, and a 3776 CPC; TSI Inc.). The sampling flow rate of SMPS was set to 0.3 115 LPM with a sheath-to-sample ratio of 10:1, resulting in an electrical mobility size range from 14.6 nm to 710.5 nm.
An aerodynamic particle sizer (APS, model 3021; TSI Inc.) was used to provide number size distribution for ambient aerosols with da ranging from 0.542 μm to 19.81 μm. The aerodynamic particle number size distribution obtained from APS could be converted to particle mobility size (dm) distribution by assuming the effective density of ambient particles to be1.5 g cm -3 (Khlystov et al., 2004;Chen et al., 2018). The total inflow rate of APS was 5 LPM, of which 1 LPM was sample flow, 120 and a remainder 4 LPM (passing through a built-in particle filter) as a sheath flow.

Particle chemical composition
Real-time non-refractory PM1 (da smaller than 1.0 μm) mass loading and chemical composition was measured by an Aerosol Chemical Speciation Monitor (ACSM; Aerodyne Inc.) equipped with a quadrupole analyzer. The sampling flow rate of ACSM was 0.1 LPM. A PM2.5 cyclone was installed upstream ACSM inlet to prevent inlet clog by particles with da larger than 2.5 125 μm. The time resolution of an ACSM scan was set to 15 minutes. Meanwhile, BC mass concentration was monitored by a multi-angle absorption photometer (MAAP, model 5012; Thermo, Inc.) with a temporal resolution of 1 minute.

Ice nucleating particle (INP) concentration
In situ immersion INP concentration was measured by a Horizontal Ice Nucleation Chamber (HINC) at fixed lamina condition throughout the observation period, i.e. with a lamina temperature (Tlam) of -30 °C and RHw = 104% (equivalent to RHi = 140%, 130 where the subscript i denotes ice). HINC is a CFDC type instrument made of two flat parallel copper plates. The temperature of each plate is controlled independently to create supersaturation along the chamber centerline lamina. To minimize the impact of convection, the top plate of HINC is warmer than the bottom plate. Ice crystal size and number was measured by a 6-channel optical particle counter (OPC; MetOne Inc.). The injector position, and thus the flow structure of HINC in this study is identical to the setting of Lacher et al. (2017). Therefore, only particles larger than 5 μm detected by the HINC OPC are counted as ice 135 crystals. For more detailed HINC design and operating principle information, please refer to Lacher et al. (2017) and Kanji and Abbatt (2009).
In this study, both warm and cold walls of HINC lined with glass-fiber filter paper were wetted with ~150 mL de-ionized water each day before the experiment start, or after running experiment for 4 hours. After draining for ~15 min, the wall temperatures of warm and cold walls would be set to -20 and -40 °C respectively to achieve desired lamina temperature (-140 30 °C) and RHw (104%). The sampling flow rate of HINC was 0.26 LPM, surrounded by 2.57 LPM particle-free nitrogen sheath gas. During the experiment, sampling air would pass through a particle filter for 5 minutes after every 15 minutes of measurement to quantify HINC background count detected by the OPC. HINC background counts follow a Poisson distribution, based on which the average background count is determined. Average ice crystal concentration (equivalent to NINP) of the 15minute measurement is calculated by firstly subtracting the average background particle counts from measurement counts, and 145 secondly converting particle counts to number concentration using HINC sampling flowrate. This study reports positive NINP only, because the negative values indicate that the signal of OPC during the measurement is undistinguishable from background noise.
https://doi.org/10.5194/acp-2021-922 Preprint. Discussion started: 21 January 2022 c Author(s) 2022. CC BY 4.0 License. Ambient particle number concentration entering HINC was monitored by a CPC (Model 3775; TSI Inc.) connected in parallel with HINC at the aerosol inlet (see Fig. 1). Activation fraction (AF), i.e. the ratio between ice crystal number 150 concentration at HINC outlet (calculated from OPC counts, as stated above) and total particle number concentration at HINC inlet (measured by CPC) were simultaneously monitored and calculated.

Overview
The observation lasted from Feb. 10 th to 28 th , overlapping with the traditional Chinese Spring Festival for the year 2021. Figure  155 2 displays the chemical composition of non-refractory PM1 mass concentration in the upper panel, and particle number size distribution in the lower panel. Figure    indicating secondary pollutant formation (e.g., Wu et al., 2018). Such secondary pollutant transformation processes produce particles in the accumulation mode, as illustrated by the high level of particle concentration peaking between 200-300 nm in 180 On Feb. 18 th , the first workday after the Spring Festival holiday (Feb. 11 th to Feb. 17 th ), people swarmed into downtown Beijing and the mass concentrations of non-refractory PM1 and PM2.5 increased during rush hours, as can be seen in Fig. 2 and  185 3, respectively. mBC also increased because of increasing use of passenger vehicles during rush hours on workdays after Feb.
During the whole observation, there was minor difference between ambient PM2.5 and PM10 mass concentrations except for Feb. 21 st afternoon, when there was a significant increase of PM10 mass concentration as highlighted by the orange shading in Fig. 3, implying that large (dust) particles with da ranging between 2.5 μm and 10 μm were present. Besides, aerodynamic 190 particle number size distribution exhibited a clear shift towards the larger end during the dust event (Fig. A1) with a mode size around 1 μm, which further confirmed the presence of large particle during the dust event. Aerosol optical depth (AOD) derived from MODIS Aqua Deep Blue Collection 6 dataset at 550 nm (Acker and Leptoukh, 2007) also shows elevated aerosol loading on Feb. 21 st afternoon at the sampling site compared to Feb. 20 th (Fig. A2). Based on measured particle mass concentration level and meteorology conditions, the observation days are categorized into different scenarios, i.e. dust event, clean, and 195 heavy pollution days, as summarized in Table 1. Table 1 The date, number concentrations of immersion INP (NINP), ambient particle (NCPC), particles larger than 500 nm (N500) and 1000 nm (N1000), as well as mass concentrations of BC particles (mBC) and ammonium salt (mammo), and activation fraction (AF) for each scenario. The numbers are average values, and numbers in parentheses denote one standard deviation (σ) from the average.

Scenario
Date of Feb.

Contribution of mineral dust to NINP during the dust event
On the afternoon of Feb. 21 st , 2021, a dust event occurred at the sampling site, as indicated clearly by the significant difference between PM10 and PM2.5 mass concentrations in Fig. 3. PM10 mass concentration reached 250 μg m -3 and was 3 to 5 times as much as the PM2.5 mass concentration during the dust event.
The dust event is characterized with a substantially higher AF of 0.0036% ± 0.0011% compared to other days, as listed 210 in Table 1. During the dust event, NINP was 1 to 2 orders of magnitude higher than clean days, ranging from 40 to 160 # L -1 . Meanwhile, the ambient particle number concentration entering HINC (NCPC) during the dust event is only half to two thirds of the clean-day concentration level (Table 1) Previous laboratory studies have shown that larger particles, especially those larger than 500 or 1000 nm, exhibit superior INP activity based on surface active site density theory (e.g., Connolly et al., 2009;Welti et al., 2009;Lüönd et al., 2010;225 Hoose and Möhler, 2012;Ardon-Dryer and Levin, 2014;Chen et al., 2021). Bi et al. (2019) observed significant increase of NINP when ambient particle peak size shifted towards the larger end of size spectra (exceeding 1000 nm) during springtime dust events in rural Beijing. N500 is generally below 50 # cm -3 during springtime dust events in Beijing (Bi et al., 2019), and is comparable to the results presented in Fig. 5a. To quantify the impact of N500 on NINP, linear regression analysis between logarithms of NINP and N500 is performed for dust event using ordinary least square (OLS) method, as shown in Fig.  230 5a. The triangles indicate measured data, and the blue markers are the predicted data using the fitted linear regression parameters. The coefficient of determination (R 2 ) between log10(NINP) and log10(N500) is 0.11 (Fig. 5a), reflecting that N500 is not likely to be correlated with NINP, and might have limited impact on NINP during the dust event. Besides, the significant difference between PM10 and PM2.5 mass concentrations (lower panel of Fig. 3) indicates that large (dust) particle (occupies high mass concentration but low number concentration) with high IN activity appeared during the dust event. It would be 235 worthwhile to explore the connection between the mass concentration difference of PM10 and PM2.5 (PM10-2.5) and NINP in the urban environment. OLS linear regression analysis between NINP and PM10-2.5 data collected during the dust event exhibits a https://doi.org/10.5194/acp-2021-922 Preprint. Discussion started: 21 January 2022 c Author(s) 2022. CC BY 4.0 License. stronger yet not statistically significant correlation (R 2 = 0.24, Fig. C1) compared to the correlation between NINP and N500. An earlier study in east Mediterranean urban region claimed that immersion IN activity of particles collected during dust storms correlated well (R 2 = 0.47) with PM10-2.5 between -10 °C and -30 °C (Ardon-Dryer and Levin, 2014). However, the correlation 240 between NINP and PM10-2.5 at -30 °C during the dust event in this study suggests that PM10-2.5 is not well correlated with NINP in the urban environment. The upper panel of Fig. 6 displays the diurnal profile of measured NINP and mass concentration of atmospheric ammonium (mammo) during the dust event. It can be seen that except for 13:00 (UTC+8), NINP profile seems to follow mammo profile. Wu et al. (2020) reported that ammonium ions could form and accumulate on mineral dust surface in the form of ammonium nitrate in the highly populated urban environment. The synchronized trends of NINP and mammo suggest that the trace amount of 250 atmospheric ammonium (below 1 μg m -3 compared to 5-10 μg m -3 on other days, as listed in Table 1) might be internallymixed with ambient mineral dust particles. The ammonium content on mineral dust surface might promote their IN activity due to strengthened ammonium ion surface adsorption followed by the formation of ice-favorable structure on dust particle surfaces (Boose et al., 2016b;Kumar et al., 2018;Whale et al., 2018;Kumar et al., 2019). To quantify the correlation between atmospheric ammonium content and NINP during the dust event, and to investigate whether the observed enhancement of 255 mineral dust IN activity by ammonium salts in previous studies (Boose et al., 2016b;Kumar et al., 2018;Whale et al., 2018;Kumar et al., 2019) still holds for the urban environment, linear regression analysis between NINP and mammo is performed, as shown in the lower panel of Fig. 6. The blue markers are fitted NINP based on the OLS regression parameters, and the blue shading refers to ± 1σ range (calculated from measured NINP) from fitted NINP. More than 60% of measured NINP fall into the shaded area, suggesting that NINP might be associated with mammo during dust events in the urban environment. However, more 260 field observations in urban areas, as well as systematic laboratory studies using natural mineral dust samples (e.g., Saharan dust and Asian dust, etc.) are required to further investigate the connection between mineral dust surface characteristics and IN activity, and the underlying mechanism.

Traffic emission
Clean days, when compared with heavy polluted or dusty days, provide ideal background to investigate the impact of primary particle emission sources, especially local traffic emission (from clean days), on NINP in urban regions. The major particle formation pathway in gasoline passenger vehicle exhausts is volatile organic compound (VOC) nucleation, producing large 290 number of nanoparticles (diameter smaller than 50 nm) with low mass concentration (Raza et al., 2018 and references therein).
On the other hand, diesel engine particle emission is dominated by BC particles ranging between 80-200 nm (Kittelson, 1998).
The increase of mass concentrations of organics (morg) and BC between 16:00 to 20:00 (UTC+8) in Fig. 8 corresponds to the evening rush hours, during which the emission of gasoline passenger vehicles dominating ambient particle population in the urban region. There is a further increase of mOrg and mBC after 20:00 (UTC+8) followed by a plateau in Fig. 8. According to 295 Beijing municipal administrative regulation, heavy-duty diesel trucks for goods transportation, as well as gasoline passenger vehicles with foreign plates (issued by cities other than Beijing) are only permitted to enter urban Beijing after 20:00 (UTC+8).
Increasing emission from on-road heavy-duty diesel trucks and gasoline passenger vehicles with foreign plates are highly likely to be responsible for the increasing mOrg and mBC after 20:00 (UTC+8, Hua et al., 2018;Zhang et al., 2019).  NINP of clean days ranges between 0.33 to 16 # L -1 which is the same order of magnitude as the immersion NINP results of Schill et al. (2016), who reported NINP for both freshly-emitted and aged BC on the orders of 10 -1 to 10 1 # L -1 measured at similar experiment condition (-30 °C and RHw = 105%) to this study, using BC generated from an off-road diesel engine. To investigate the impact of traffic emission on NINP, linear regression analysis is performed between NINP and mBC, a widely used 305 cursor of traffic emission. As shown in Fig. 9a, the correlation between NINP and mBC is poor (R 2 = 0.01), implying that NINP is independent of mBC on clean days. Recently, Kanji et al. (2020) also reported that BC might not act as effective immersion INP based on laboratory experiments. The absence of a relationship between NINP on mBC on clean days in this study is consistent with previous findings (Schill et al., 2016;Kanji et al., 2020;Schill et al., 2020).

Firework emission
Heavy pollution accompanied by the presence of substantially higher mass concentrations of non-refractory PM1, PM2.5 (128 315 ± 44 μg m -3 ) and BC (up to 6 μg m -3 ) occurred during Feb. 12-13 and 24-27, as shown in Fig. 2-4. N500 was also substantially higher during heavy pollution than other days (Table 1). As stated above, large amounts of particles comprising carbonaceous material and chloride emerged after Spring and Lantern Festival celebrations due to firework emissions which also contain trace amounts of potassium and other metal elements (Jiang et al., 2015;Kong et al., 2015;Cao et al., 2017). With almost 10 times as much N500 as that of the dust event during heavy pollution, it was expected that a substantially 320 higher INP concentration should be observed if these large particles are effective INP. However, the particle population during heavy pollution did not exhibit superior IN activity, with the majority of NINP fell into the range of 0 to 25 # L -1 . As shown in Fig. 5b and Fig. 9b, the OLS linear regression results further suggest that NINP is likely to be independent of N500, as well as mBC during heavily polluted days, with R 2 between log10(NINP) and log10(N500), mBC being 0.01 and 0, respectively. The independence of NINP on mBC is compliant with the results of Adams et al. (2020), in which there was a substantial growth (by 325 more than an order of magnitude) of ambient particle number concentration and mBC from combustion and firework emissions, but no significant NINP change was observed. Chen et al. (2018) conducted off-line NINP measurement using filtered samples collected at the same sampling site as this study during heavy pollution, and found no dependence of NINP on neither the mass concentrations of PM2.5 and BC, nor N500 as well. As aforementioned, the synergetic heavy pollution after festival celebrations were induced by secondary pollutant formation via liquid phase reaction Wu et al., 2018) and firework 330 emission. Under the mixed-phase cloud condition (-30 °C, RHw = 104%) in this study, such particles are very likely to become aqueous droplets or contain liquid films on solid particles, which might require conditions for homogeneous freezing to nucleate ice.

Conclusion
Continuous in situ observation of INP number concentration (NINP) and physiochemical properties, including chemical 335 composition and size distribution, of ambient particles at an urban site in Beijing during the traditional Chinese Spring Festival has been performed at mixed-phase cloud condition (-30 °C , RHw = 104%) for 18 days. The impact of different scenarios, such as the synergetic heavy pollution induced by secondary aerosol formation and firework emissions, a dust event, and local traffic emissions on NINP has been explored. NINP was investigated in relation to N500 and mBC. The relationships of mammo with NINP, as well as PM10-2.5 with NINP, during the dust event are also presented. The results show that NINP, as well as AF of ambient 340 particles during dust event are substantially higher than all other scenarios. NINP could reach 160 # L -1 during the dust event, while it ranges from 10 -1 to 10 1 # L -1 on other days. AF during the dust event (0.0036% ± 0.0011%) is 20 to 30 times higher than clean (0.0001% ± 0.0001%) and heavily polluted days (0.0002% ± 0.0002%). During the dust event, NINP and mammo exhibited synchronized variation, and NINP exhibited slight dependence on PM10-2.5 (R 2 = 0.24). The parameterization proposed by DeMott et al. (2010) predicts more than 60% of measured NINP within a factor of 2.5 during the dust event. Mass 345 concentration measurements suggest that large amounts of aerosols containing chloride and BC appeared after the celebrations on Feb. 11 th and 26 th nights due to firework emission. Meanwhile, the stagnant and humid meteorology condition provides ideal condition for secondary aerosol formation. But there is no significant difference between NINP on heavily polluted and clean days, implying the urban aerosols from multiple sources with complex chemistry might not be effective INP. Besides, the diurnal increase of mBC from petrol passenger vehicle emissions during rush hours and from diesel truck emissions after 350 20:00 (UTC+8) on clean days does not lead to distinguishable higher NINP, implying that local traffic emission also has https://doi.org/10.5194/acp-2021-922 Preprint. Discussion started: 21 January 2022 c Author(s) 2022. CC BY 4.0 License. negligible impact on NINP. Our study reveals that mineral dusts, even though present in relatively low number concentration out of the high background particle number concentration, dominate immersion INP population in the urban environment.
Furthermore, our results agree with previous literature from laboratory and field studies that atmospheric BC from both local traffic and firework emissions has negligible effects on mixed-phase cloud formation, and that NINP is unaffected by heavy 355 pollution.   temperature gradient and supersaturation in the chamber (Lacher et al., 2017). The RHi within HINC lamina changed continuously from 100% to 160% at each lamina temperature. The concentration of the ammonium nitrate solution was 0.0025 mol L -1 . The solution was atomized by a nebulizer (TSI Inc.) using 1.5 LPM nitrogen gas. The flow stream was dried to RHw < 2% by passing through a 47 cm Nafion TM dryer and was then size selected by a DMA (model 3081 long; TSI Inc.). Particle number concentration entering HINC was measured online by a CPC (model 3775; TSI Inc.). 375 Figure B1 shows the AF as a function of RHw at -30 °C for 200 nm ammonium nitrate aqueous droplets. Different OPC channel (> 1 μm, > 2 μm, > 3 μm, > 5 μm) are marked with different colors. It is worth noting that given the flow structure of HINC in this study (Sect. 2.2.3), only particles larger than 5 μm detected by HINC OPC would be recognized as detectable ice crystals (below -38 °C) or water droplets (above -38 °C). As shown in Fig. B1, 200 nm ammonium nitrate aqueous droplets start to grow upon water saturation (black dots), followed by more rapid growth with increasing RHw (grey dots). However, 380 there is no growth in >5 μm channel until RHw exceeds 106%, corresponding to the presence of detectable water droplets larger than 5 μm. Therefore, HINC should be operated below 106% at -30 °C to avoid erroneous count of large (>5 μm) water droplets rather than ice crystals.  the calibration, the average variation of lamina RHw was less than 1.2% (corresponding to a 1.8% variance of lamina RHi) .
Lamina RHw suffers larger variation as RH increases, resulting in RHw = 108% ± 2.1% at -30 °C . The variance of lamina temperature was below 0.2 K throughout the calibration process. As shown in Fig. B1, the IN onset point of 200 nm ammonium 395 nitrate at -40 °C lies on the calculated homogeneous freezing threshold. The IN onset at -45 °C exceeds the homogeneous freezing threshold by 3.5%, yet still below water saturation line. When the lamina temperature is above -38 °C, water drops require RHw substantially higher than 104% (dashed line) to be detected in the >5 μm OPC channel, as such we are confident that signals arising in the >5 μm OPC channel at RHw = 104% are due to ice crystal formation.