Seasonal Variation of Wet Deposition of Black Carbon at Ny‐Ålesund, Svalbard

Black carbon (BC) aerosol deposited in and onto Arctic snow increases the snow's absorption of solar radiation and accelerates snowmelt. Concentrations of BC in the Arctic atmosphere and snow are controlled by wet deposition; however, details of this process are poorly understood owing to the scarcity of time‐resolved measurements of BC in hydrometeors. We measured mass concentrations of BC in hydrometeors (CMBC) and in air (MBC) with 16% and 15% accuracies, respectively, at Ny‐Ålesund, Svalbard during 2012–2019. Median monthly MBC and CMBC values showed similar seasonal variations, being high in winter‐spring and low in summer. Median monthly BC wet deposition mass flux (FMBC) was highest in winter and lowest in summer, associated with seasonal patterns of CMBC and precipitation. Seasonally averaged BC size distributions in hydrometeors were similar except for summer. Measurements of MBC and CMBC in spring 2017 showed a size‐independent removal efficiency, indicating that BC‐containing particles were efficiently activated into cloud droplets. These observations at Ny‐Ålesund were compared with observations at Barrow, Alaska, during 2013–2017. The near‐surface MBC at Ny‐Ålesund and Barrow had similar seasonal patterns; however, the two sites differed in CMBC and FMBC. In summer, CMBC was low at Ny‐Ålesund but moderate at Barrow, likely reflecting differences in MBC in the lower troposphere. Seasonally averaged BC size distributions in hydrometeors were similar at both sites, suggesting that average BC size distributions are similar in the Arctic lower troposphere. The efficiency of BC removal tends to be size‐independent during transport, leading to the observed similarity.


Introduction
Black carbon (BC) particles efficiently absorb solar radiation and lead to heating of the atmosphere. Moreover, BC deposited on surface snow reduces the snow albedo and accelerates snowmelt (AMAP, 2015;Bond et al., 2013;Hadley & Kirchstetter, 2012;Wang et al., 2020). Recent global climate models in the framework of the sixth phase of the Coupled Model Intercomparison Project (CMIP6; Eyring et al., 2016) have estimated that BC is the second-largest positive effective radiative forcing in the Arctic, after carbon dioxide . Because radiative forcing is strongly controlled by the distribution of BC concentrations, it is critically important to estimate BC concentrations accurately in the Arctic. In general, BC particles emitted from midlatitudes and transported to the Arctic as well as those emitted in the Arctic are removed from the atmosphere mainly by wet deposition (Koike et al., 2021;Schmale et al., 2018;Willis et al., 2018). Wet deposition is a key process controlling the spatial distribution of BC in the Arctic atmosphere and snow Mahmood et al., 2016;Matsui & Moteki, 2020).
The majority of global climate models have indicated that the amount of wet deposition of BC in the Arctic region (>66°N) is high in summer and low in winter, depending strongly on the spatial and temporal distribution of precipitation amount (Mahmood et al., 2016;Sharma et al., 2013). However, the magnitude and the seasonality of calculated wet deposition amounts have not been fully tested against observational data, partly because few suitable datasets (e.g., BC concentrations in snow and rain) exist to quantify wet deposition of BC in the Arctic. Therefore, it is critically important to measure BC concentrations in hydrometeors (falling snow and rain) for quantitative investigations of BC wet deposition in the Arctic.
The size distributions of BC in hydrometeors, mass concentration of BC in hydrometeors (C MBC ), and number concentration of BC in hydrometeors (C NBC ) can be measured by using a system combining a single-particle soot photometer (SP2), based on a laser-induced incandescence technique, with a concentric pneumatic nebulizer (Katich et al., 2017;Mori et al., 2016). This measurement system has been applied to hydrometeor and snowpack samples in several Arctic regions (Jacobi et al., 2019;Konya et al., 2021;Mori et al., 2019Mori et al., , 2020Sinha et al., 2018). Sinha et al. (2018) used an SP2 to measure C MBC in daily snowfall samples during the snowfall season (September-April) at Ny-Ålesund, Spitsbergen, in the Svalbard archipelago, to document the temporal variations of C MBC and BC wet deposition flux during that period. Jacobi et al. (2019) measured C MBC in snowpack at Ny-Ålesund during November-April, quantifying the amount of BC deposited in the snow column. However, the BC wet deposition flux at this site at other times of year has not been investigated in detail. Mori et al. (2020) used an SP2 to measure C MBC in daily snowfall and rainfall samples at Barrow Atmospheric Baseline Observatory near Utqiaġvik in Alaska (Barrow hereafter) during 2013-2017. They showed that BC wet deposition flux by mass (F MBC ), derived from C MBC and precipitation amount, is highest in summer because C MBC and precipitation amount are both highest at that time. Calculations of C MBC and F MBC by two global models have suggested that the enhancement of C MBC in summer arises from the increase of BC mass concentration in air (M BC ) in the lower free troposphere originating from biomass burning . However, it is not clear if these features of BC wet deposition are common in the Arctic. Further measurements of C MBC and F MBC at other locations with different environmental settings are required to elucidate the variations of BC wet deposition in wider regions of the Arctic. measured at Jungfraujoch in Switzerland (Cozic et al., 2007). The calculations suggested that by incorporating the WBF process, M BC in the free troposphere (∼4 km altitude) increases by 45%-95%, whereas C MBC and F MBC decrease by ∼15% and 22%-29%, respectively, in the Arctic (60°N-90°N) (Qi et al., 2017). These model analyses suggest that BC wet deposition in the Arctic is affected by the WBF process in mixed-phase clouds. However, this simulated effect of the WBF process has not been fully tested with observations, mainly due to the paucity of C MBC and F MBC measurements.
To quantitatively understand the characteristics of BC wet deposition in the Arctic, we collected daily samples of snow and rain at Sverdrup Research Station at Ny-Ålesund (Ny-Ålesund, hereafter) from August 2012 to October 2019, and we measured C MBC and C NBC along with M BC at Zeppelin Observatory on top of Mt. Zeppelin near Ny-Ålesund as well as meteorological parameters and cloud-base heights. In addition, BC wet deposition flux was derived from precipitation amounts and the BC concentrations in hydrometeors. We first present the temporal variations of the observed parameters (M BC , C MBC , F MBC , precipitation, and size distributions of BC in hydrometeors) at Ny-Ålesund. We then compare these quantities with those measured at Barrow from July 2013 to August 2017 .
The locations of the observation sites are shown in Figure 1. Local BC emissions from anthropogenic and biomass burning sources near Ny-Ålesund and Barrow are negligible (Lamarque et al., 2010). The relevant measurements are listed in Table 1. The parameters used in this study are summarized in Table A1 of Mori et al. (2020).

BC in Air
Our M BC measurements were made with a continuous soot monitoring system (COSMOS; Kanomax, Inc.) at Zeppelin Observatory just south of Ny-Ålesund (78. 91°N, 11.88°E, 474 m elevation), using a PM 1 impactor inlet. This system measures the light absorption coefficient of BC particles (<1.0 μm) on a quartz-fiber filter (Pallflex E70-2075 W) with a time resolution of 1 min. The interference from light-scattering particles on the filter is negligible because these particles are removed during passage through a heated inlet (300°C).
MORI ET AL. This enables M BC measurements with an accuracy of about 15%, estimated by comparisons with M BC simultaneously measured by the SP2 with 10% accuracy . Detailed descriptions of the BC measuring system are given elsewhere (Kondo et al., 2009;Ohata et al., 2019;Sinha et al., 2017).

Collection of Snow and Rain
Daily snowfall and rainfall samples were collected at Sverdrup Research Station at Ny-Ålesund (78. 92°N, 11.93°E, 5 m elevation), operated by the Norwegian Polar Institute, from August 2012 to October 2019. In total, we collected 283 snowfall samples and 273 rainfall samples ( Table 2). Details of the collection techniques are described elsewhere (Mori et al., 2014Sinha et al., 2018).
In brief, rainfall samples were collected by using a rain collector, consisting of a glass funnel and bottle, to minimize possible loss of BC particles attached to the inner wall (Mori et al., 2014. To minimize contamination during samplings, the rain collector was precleaned with pure water, routinely. Snowfall samples were collected by using two different apparatuses (Sinha et al., 2018): a windsock was used from September 2012 to June 2016 and a box collector was used from December 2016 to October 2019. The windsock, made of plankton netting, had an opening diameter of 17 cm and a length of 40 cm. The box collector was an acrylic box (0.5 m long, 0.4 m wide, and 0.4 m high), inside of which a powder-free polyethylene bag was installed and fixed with a rubber band. Both snowfall collectors were mounted 8.5 m above the ground to avoid collecting wind-blown snow from the surface. When snow and rain were mixed, samples from the snow collector were used. Immediately after collection, the snowfall samples were melted at room temperature and then poured into precleaned vials (Sinha et al., 2018). These were kept in a refrigerator at about 5°C until transport by air freight to the University of Tokyo or the National Institute of Polar Research in Japan, where they were kept at about 5°C until analysis Sinha et al., 2018

BC in Snow and Rain
Mass and number size distributions of BC in hydrometeors were measured by a system consisting of a Marin-5 pneumatic nebulizer (Cetac Technologies Inc.) and the SP2 Sinha et al., 2018).
The melted snow and rainwater samples were first agitated in an ultrasonic bath for 15 min to minimize the loss of BC particles attached to the glass wall (Mori et al., 2014. The water samples were injected into the nebulizer at a constant flow rate of 3.0 × 10 −6 L s −1 to be aerosolized. The extraction efficiency was 50.0 ± 4.4% on average in the 70-2,000 nm size range, and it decreased linearly at larger sizes (2,000-5,000 nm). The BC and non-BC particles generated by the nebulizer were introduced into the SP2 at a flow rate of 16 cm 3 s −1 at standard temperature and pressure. The SP2 then measured the size-resolved BC mass and number concentrations in the carrier gas. A standard SP2 measures the mass of each BC particle with mass equivalent diameter (D BC ) within the 70-850 nm range by assuming a BC particle density of 1.8 g cm −3 (Moteki & Kondo, 2010), whereas our SP2 used in this study measures masses of BC particles within the 70-4,170 nm D BC range by expanding the upper limit of the detected incandescence signal in the standard SP2 . Size distributions of BC in water were derived from the pump and nebulizer flow, the nebulizer efficiency, and the size distribution of BC in the carrier gas within the 70-4,170 nm D BC range. We calculated C MBC and C NBC by integrating the mass and number size distributions, respectively, in water within the 70-4,170 nm D BC range . This technique yields C MBC and C NBC measurements with an accuracy of 16% (Table 3), derived from the respective measurement uncertainties of the pump flow rate (±5%), nebulizer flow rate (±5%), nebulizer efficiency (±10%), and BC concentration in the carrier gas measured by the SP2 (±10%).
The detection of smaller BC particles strongly depends on the laser intensity in the SP2 (Schwarz et al., 2010). Laser beams with weak intensity do not completely evaporate smaller BC particles, which causes BC mass and number concentrations to be underestimated. The laser intensity of our SP2 was sufficiently strong and stable to detect BC particles with D BC = 70 nm during the analysis period. In addition, our SP2 was regularly calibrated by using fullerene soot (Alpha Aeser Inc.) and maintained a similar relationship between MORI ET AL. Uncertainty of F NBC ±34%-44% Propagation of uncertainties of C NBC (16%-19%) and precipitation amount (30%-40%).
Section 2.6 Table 3 Accuracies and Uncertainties of Parameters in This Study the masses of individual BC particles and the incandescence signals they emitted (Moteki & Kondo, 2010) during 2012-2019.
To accurately measure the size distributions of BC in hydrometeors, it is important to evaluate the stability of BC sizes during long storage periods in glass bottles. The shapes of the size distributions of BC in melted snowpack samples, collected at Ny-Ålesund, were stable and the C MBC and C NBC values agreed to within 20% during storage periods as long as a few years . Our melted snow and rain samples were analyzed within a year after their collection, thus any changes in their BC sizes were considered negligible, and the concentrations in hydrometeors were also stable during the period of storage before analysis. The size of BC particles can change by coagulation during the extraction of BC particles with a nebulizer. Mori et al. (2016) used the SP2 to measure the number size distributions of four different fullerene soot suspensions with mass concentrations ranging from 6.9 to 64 μg L −1 . They showed that the four distributions had similar shapes, demonstrating that the effect of coagulation during the extraction process is negligible for samples with C MBC less than 64 μg L −1 . The annual average (±1σ) value of C MBC at Ny-Ålesund during 2012-2019 was 2.8 ± 2.6 μg L −1 , well below this upper limit. Table 3 summarizes the uncertainties in C MBC for precipitation samples acquired by rain, box, and windsock collectors. Mori et al. (2019) showed that in our rain collector, the BC losses due to adhesion on the wall of the glass bottle are small. In addition, the glass funnel is routinely precleaned with pure water to minimize contamination during samplings. Therefore, we estimated that the additional uncertainty associated with the rain collector was negligible and the overall uncertainty in C MBC was 16%.

Uncertainty in C MBC Using Rain, Box, and Windsock Collectors
In the case of box collectors, Sinha et al. (2018) found that contamination of BC particles attached to polyethylene surfaces inside the box collector was negligible, and by comparing C MBC in snowpack and falling snow at Ny-Ålesund, they showed that dry deposition makes only a small contribution to total deposition. Considering these results, we estimated that the uncertainty introduced by snow sampling with the box collector was negligible, and thus the overall uncertainty in C MBC was 16%.
The windsock collector was used to collect about 160 snowfall samples from September 2012 to June 2016, about 30% of the total number of hydrometeor samples in this study. Sinha et al. (2018) compared C NBC and C MBC values for 11 daily snowfall samples collected simultaneously by the windsock and box collectors at Ny-Ålesund from December 2016 to March 2017. The windsock collector samples yielded C NBC and C MBC values for the 70-1,000 nm D BC range that on average were higher by factors of about 1.2 and 1.8, respectively, than those of the samples from the box collector. The increase is possibly due to the continued collection of ambient BC particles by the snow samples during the time they resided in the windsock; this would occur by interception and inertial impaction (Wang et al., 2010). To address this uncertainty, we first corrected the C NBC values by using the cumulative C NBC (integrated from 70 to 200 nm) measured in windsock samples and a correction factor of 1.1. This is a correction factor that Sinha et al. (2018) derived from the cumulative C NBC value (integrated from 70 to 200 nm) of the windsock and box samples and the total C NBC value (integrated from 70 nm to 4,170 nm) of the box samples for the period of December 2016-May 2017. The C MBC values of these windsock samples were corrected by using the corrected C NBC values and a mass per particle value of 4.67 fg, as determined by Sinha et al. (2018) from the correlation between total C NBC and C MBC for their snowfall samples from the box collector. The uncertainties in the corrected C NBC and C MBC values were 19% and 21%, respectively (Sinha et al. 2018). In this study, our analysis of BC size distributions in the snow samples relied only on data from the period when the box collector was used.

Precipitation Amount
Daily precipitation amount (P) was measured by using a Geonor weighing gauge with an Alter windshield at an automatic weather station (AWS) (78.92°N, 11.93°E, 8 m elevation), together with snow depth (Førland & Hanssen-Bauer, 2000;Hanssen-Bauer et al., 1996). The sensor accuracy of P and snow depth was about 0.1 mm and 1 cm, respectively. We classified all precipitation days as snow days or rainy days, defined as days with and without a daily increase in snow depth, respectively.
Previous studies have shown that measurement uncertainties for solid precipitation in the Arctic are caused by wind-induced undercatch, wetting, evaporation, and blowing snow (Førland & Hanssen-Bauer, 2000, 2003Rasmussen et al., 2012). They reported that the uncertainty of P was 20%-50% when compared with total precipitation values measured by a precipitation gauge with a double fence. Considering these results, we estimated the uncertainty of monthly P values obtained from the AWS by comparing them with Global Precipitation Climatology Project (GPCP) One-Degree Daily data (https://psl.noaa.gov/) for the grid box including Ny-Ålesund (78°N-79°N, 11°E-12°E) based on satellite and rain gauge measurements during 2012-2015. The monthly AWS precipitation data were strongly correlated with the GPCP data during the cold season (October-May) (r 2 = 0.84 and AWS/GPCP = 0.62) and the warm season (June-September) (r 2 = 0.90 and AWS/GPCP = 0.71). Thus, the measurement uncertainties of P were about 40% during the cold season and 30% during the warm season, when compared with GPCP data, a result consistent with previous studies Sinha et al., 2018). The possible bias in the AWS data was not corrected in this study.

BC Wet Deposition Flux
Detailed methods for deriving the monthly BC wet deposition flux have been described elsewhere (Mori et al., 2014. In brief, daily wet deposition flux of BC by mass (F MBC ) (µg m −2 d −1 ) was calculated as where C MBC (µg L −1 ) is daily BC mass concentration in hydrometeors and P (mm d −1 ) is daily precipitation amount. Then, the monthly P collection rate in the jth month (R j ) was calculated, defined as the ratio of the sum of daily P for sampling events in the jth month to the total P for all events during the jth month. The monthly F MBC (µg m −2 month −1 ) in the jth month was derived by dividing the sum of daily F MBC in the jth month by R j . Monthly wet deposition flux of BC by number (F NBC ) in the jth month was similarly derived.
The monthly mean C MBC and C NBC in the jth month were derived as the ratios (F MBC ) j /P j and (F NBC ) j /P j , which represent the P-weighted C MBC and C NBC averages, respectively.
The seasonal median values of R j ranged from 0.69 to 0.81, as summarized in Table 2. These were much higher than R j values at Barrow (0.17-0.57; shown in Mori et al., 2020), indicating that our datasets of monthly F MBC and F NBC are statistically more reliable throughout the year because we sampled more of the month's precipitation.
Uncertainty in the estimate of the corrected C MBC for windsock samples was 21% for snowfall events during 2012-2016 (Section 2.4). Except for this period, the measurement uncertainty of C MBC for samples from the rain and box collectors was 16% (Section 2.3). The estimated uncertainty of monthly P values ranged from 30% to 40%, based on the comparison between AWS and GPCP data (Section 2.5). Considering these results, the uncertainty in our estimates of monthly F MBC ranged from 34% to 45%. These uncertainties are summarized in Table 3.

Meteorology
Vertical profiles of meteorological parameters (e.g., pressure, temperature, and relative humidity) have been routinely measured near the hydrometeor sampling site by using radiosondes launched from the balloon launching site (78.92°N, 11.92°E, 15.5 m elevation) at 0000 and 1200 UTC. We used these profiles to calculate the vertical profile of water vapor mixing ratio on the basis of the Clausius-Clapeyron equation (Bolton, 1980;Murphy & Koop, 2005) and to calculate total precipitable water (TPW), representing the total amount of water vapor in a column of air.
Cloud-base heights have been continuously measured at the French-German Arctic Research (AWIPEV) Base (78.92°N, 11.93°E, 20 m elevation) by using a micropulse lidar with a Nd/YLF laser at a wavelength of 523 nm (Campbell & Shiobara, 2008;Shiobara et al., 2003), determined by using a lidar cloud detection algorithm (e.g., Campbell et al., 1998Campbell et al., , 2002. The time resolution of the cloud-base height data is 3 min.

Seasonal Variations of Temperature
Monthly mean temperatures (±1σ) at the altitude of Zeppelin Observatory (∼0.5 km) during 2012-2019 had clear seasonal variations, being highest in July (3.7 ± 2.9°C) and lowest in March (−11 ± 5.8°C) (not shown). Monthly mean temperatures at higher altitudes (about 2.0 km, 3.0 km, and 4.0 km) had similar seasonal patterns. Except for the four months from June to September, monthly mean temperatures above 0.5 km were always below 0°C.

Cloud-Base Height
The micropulse lidar monitors the backscattering signal returned from optically thin and thick clouds, and the cloud-base height is determined as the altitude of the first level at which the backscattering signal sharply increases (Shiobara et al., 2003). We used the data of the cloud-base height for all optically thin and thick clouds from August 2012 to October 2019.
Median values of daily average cloud-base height displayed a clear seasonal variation (Figure 2a), being lowest in summer (1.2 km) and highest in winter (2.0 km). This pattern applied to all days in our analysis period, to the snow days and rainy days (Figure 2a), and to our sampling days (not shown).
The radiosonde measurements at Ny-Ålesund during 2012-2019 showed that average relative humidity over ice for snow and rainy days was about 90%-100% within the 0.6-1.4 km altitude range, although relative MORI ET AL.
10.1029/2020JD034110 8 of 32 humidity over liquid was somewhat lower. The amount of water vapor below 2.0 and 4.0 km accounted for about 70% and 90% of the TPW, respectively. In addition, the water vapor concentration below 2 km was higher on precipitation days than on nonprecipitation days. These results are consistent with the cloud-base height of 1-2 km on precipitation days.
Previous studies have investigated the features of cloud-base height at Ny-Ålesund by using micropulse lidar and ceilometer measurements. One of the studies found that the cloud-base height was mostly below 2 km in summer and early fall in 1998-1999 (Shiobara et al., 2003), and another reported that about 50%-75% of clouds appeared below 2 km in 2004-2013 (Yeo et al., 2018). A 25-year (1992-2017) data set of ceilometer measurements showed that seasonally averaged cloud-base height was lowest during summer (Maturilli & Ebell, 2018). Our results are in agreement with these previous results.
To investigate the occurrence of mixed-phase clouds at Ny-Ålesund during our observation period, we compared the cloud-base heights and cloud-base temperatures with those previously measured at Barrow , where mixed-phase clouds are common (Shupe, 2011). By estimating cloud-base temperature from radiosonde and micropulse lidar measurements, we found that the median (25th percentile, 75th percentile) values of cloud-base temperatures at Ny-Ålesund were −12°C (−18°C, −7.4°C) on snow days and 0.4°C (−5.1°C, 2.9°C) on rainy days, and these two categories had different temperature distributions (Figures 2c and 2e). Shupe (2011) combined micropulse lidar, millimeter cloud radar, microwave radiometer, and radiosonde observations at Barrow during 2004Barrow during -2006 to show that mixed-phase clouds formed mainly at 0-3 km altitude except in summer, when they formed at 2-6 km, and that temperatures in the mixedphase clouds ranged from −14°C to 2°C. We estimated that median (25th percentile, 75th percentile) values of cloud-base temperature at Barrow during 2013-2017 were −12°C (−17°C, −6.7°C) on snow days and −0.3°C (−6.0°C, 2.8°C) on rainy days (Figures 2d and 2f). Although the cloud-base height was somewhat lower at Ny-Ålesund than at Barrow (Figures 2a and 2b), the median values and dispersions of temperature distributions were very similar at the two sites for each snow and rainy day (Figures 2c-2f). These meteorological similarities suggest that mixed-phase clouds were also common at Ny-Ålesund.
Global model calculations showed that at Barrow, about 50% of annual BC wet deposition occurred by scavenging of BC at altitudes below 1 km and about 90% occurred by scavenging of BC at altitudes below 4 km , which suggests that a portion of hydrometeors including BC particles originated in or passed through mixed-phase clouds at Barrow. Considering the meteorological similarities of Ny-Ålesund and Barrow, it is likely that some portion of hydrometeors at Ny-Ålesund interacted with ice particles.
Previous studies have shown that mixed-phase clouds occur mainly below 2 km at Ny-Ålesund throughout the year by combining ceilometer, cloud radar, microwave radiometer, and radiosonde measurements during 2016-2018 (Gierens et al., 2020;Nomokonova et al., 2019). These results confirm that mixed-phase clouds commonly occurred below the lower altitude (∼2 km) during our observation periods.

BC Wet Removal
Time series of daily average M BC and C MBC at Ny-Ålesund are shown in Figure 3, and the respective monthly averages are shown in Figure 4. The monthly average M BC values for precipitation days ( Figure 4) were about 20% lower than those for all days (r 2 = 0.89) on average, which may partly be due to BC loss by precipitation. In this study, we used daily and monthly M BC values for precipitation days for comparison with C MBC .
In this section, we present the features of five parameters measured at Ny-Ålesund: M BC , C MBC , P, F MBC , and size distribution in hydrometeors. Monthly averages of M BC and P-weighted C MBC and monthly sums of total P and F MBC values were obtained for the same month during about 7 years of 2012-2019 (Table S1). For the statistical analyses of each parameter described below, we represented each month of the year by the median of about these seven monthly values, which we denote here as the median monthly value. The seasonal patterns of M BC , C MBC , P, and F MBC at Ny-Ålesund are derived from these median monthly values, consisting of 12 data points for each parameter, to minimize the influence of extreme values. In fact, there were large variabilities (1σ) in seasonal and annual average values of these parameters at Ny-Ålesund (Table 4).

BC Mass Concentration in Air
Median monthly M BC for precipitation days at Ny-Ålesund displayed a marked seasonal variation, being highest in spring and lowest in summer ( Figure 5a and Table 4). In general, the Arctic front during winter and spring expands southward over northern Eurasia, such that air masses significantly influenced by anthropogenic BC emissions, mainly at midlatitudes, are frequently transported to the Arctic. At the same time, the strong temperature inversion causes high aerosol concentrations to accumulate in the planetary boundary layer (PBL). As a result, M BC increases in winter and spring (AMAP, 2015;Browse et al., 2012;Hirdman et al., 2010;Quinn et al., 2007;Sharma et al., 2013). During summer, the Arctic front retreats far to the north, such that air masses are transported mainly from the Arctic Ocean (Eleftheriadis et al., 2009;Hirdman et al., 2010). In addition, because BC wet deposition generally increases during its long-range MORI ET AL.

10.1029/2020JD034110
10 of 32  A study of 20-day back trajectories from the altitude of Zeppelin Observatory (474 m) during 2002-2007, using the Lagrangian particle dispersion model FLEXPART, suggested frequent transport of polluted air masses from northern Eurasia during winter and cleaner air masses from the northern Arctic during summer (Hirdman et al., 2010). In addition, the radiosonde measurements showed that the median monthly upper boundaries of the PBL were higher in summer (∼1,600 m) and lower in winter (∼640 m), suggesting that Zeppelin Observatory was within the PBL throughout the year, on average.

BC Mass and Number Concentration in Hydrometeors
Median monthly P-weighted C MBC and C NBC at Ny-Ålesund displayed a seasonal variation like that of M BC , being high in spring and low in summer (Figures 5b and S1a, Table 4). The median monthly values of P-weighted C MBC and M BC for all precipitation days during the observation period (Figures 5a and 5b, respectively) were well correlated (r 2 = 0.78; Figure 6a). In addition, the median daily values of C MBC and M BC were also well correlated (r 2 = 0.44; Figure 6b) for sampling days in which P exceeded 1.1 mm, or about 80% of all sampling days. However, monthly P-weighted average C MBC was poorly correlated with monthly average M BC for all seasons (r 2 = 0.06; not shown) and for each of the three months of winter (r 2 = 0.01), summer (r 2 = 0.01), and fall (r 2 = 0.05), and moderately correlated for the three months of spring (r 2 = 0.16) ( Figure S2). The correlation between daily C MBC and daily M BC on sampling days was also poor (not shown).
These results indicate that the C MBC -M BC correlation is influenced by the variations of M BC and C MBC through multiple effects and processes that tend to weaken the link between the two quantities on a daily basis. First, precipitation is generally associated with physical processes, including cloud and precipitating particle formation, in volumes of air much larger than the vertical column above the measurement site. Therefore, C MBC during precipitation events may not be directly linked with M BC in the local air column, unless BC is uniformly distributed in the atmosphere. Second, increasing the number fraction of cloud droplets nucleated on BC-free particles leads to a decrease in C MBC , considering that coalescence is a key process to grow cloud droplets to rain droplet size.
Change of M BC with altitude is another factor related to the C MBC -M BC correlation.
F NBC (10 9 m −2 month −1 )  conducted over the northern polar and North Pacific regions in spring 2018, also showed that M BC was vertically rather uniform from altitudes of 0.1-2.5 km . These results suggest that the rather uniform M BC over Ny-Ålesund may account for the moderate correlation between monthly C MBC and M BC during spring. Because no observations of M BC distribution near Ny-Ålesund have been made in other seasons, it is difficult to interpret the seasonal variations in the correlation between C MBC and M BC . Ground-based measurements near Ny-Ålesund during 2014-2017 showed that the backscattering coefficient of aerosols was vertically uniform in the PBL and lower troposphere in all seasons (Shibata et al., 2018). However, this feature may not apply to BC, and the effect of variability of vertical profiles of BC is inconclusive.
The micropulse lidar and radiosonde observations showed that a portion of the hydrometeors including BC particles may have originated in or passed through mixed-phase clouds at Ny-Ålesund (Section 3.2). Global model calculations predict that BC particles are released efficiently from cloud droplets in Arctic mixed-phase clouds due to the WBF process, likely leading to reduced BC removal (Qi et al., 2017;Shen et al., 2017). Considering this, measured C MBC may be affected by the BC scavenging process in mixed-phase clouds mainly during winter and spring, when the WBF process can be important. This process may also contribute to weakening the C MBC -M BC correlation.
On a monthly basis, the variations of M BC and C MBC caused by these multiple processes may average out to some extent. However, these effects likely still degrade the C MBC -M BC correlation for the average (±1σ) number of monthly sampling days of 6.4 ± 4.5 (Figures 4 and S2). The number of sampling days during 2012-2019 ranged within N = 34-80 per calendar month (Figure 5e and Table S1). When the median values are determined by data from months with N ≈ 40, C MBC and M BC are well correlated. On this statistical basis, the BC mass deposited per unit of precipitated water increases with increasing M BC .

Precipitation Amount
Figure 3 shows daily P for each sampling day, and Figure 4 shows monthly P for all precipitation days at Ny-Ålesund during 2012-2019. Median monthly P for 2012-2019 showed a clear seasonal variation, decreasing from winter-spring to early summer and increasing from summer to winter ( Figure 5c and Table 4).
We used TPW to investigate the effect of water vapor amount on the local P. The median monthly TPW showed a clear seasonal pattern, highest in summer and lowest in winter, but its pattern differed from that of monthly P (Figure 7a), and the correlation between monthly TPW and P was only moderate (r 2 = 0.14). The error bars in Figure 7a indicate measurement uncertainties of 40% for median monthly P, which are not large enough to strongly influence the seasonal pattern of monthly P.
The value of P may be more strongly influenced than TPW by variabilities in synoptic disturbance through complicated dynamic and thermodynamic processes at Ny-Ålesund, weakening the correlation between P and TPW. On the other hand, the seasonal patterns of monthly TPW and P were similar at Barrow (Figure 7b), indicating that monthly P was strongly influenced by the local TPW . Previous studies used global models to investigate the distribution of P around Svalbard in winter (Maturilli & Kayser, 2017;Rinke et al., 2017) and in summer (Serreze & Barrett, 2008;X. D. Zhang et al., 2004). During winter, intense cyclones in the North Atlantic storm tracks lead to strong inflow of warm and moist air from MORI ET AL.
10.1029/2020JD034110 12 of 32 lower latitudes to the Arctic (Maturilli & Kayser, 2017;Rinke et al., 2017). Consistently, the spatial distribution of monthly P from GPCP data during winter shows a marked latitudinal gradient of P southwest of Svalbard (Figure 8a). During summer, lows over the Eurasian continent and cyclogenesis over the Arctic Ocean cause an increase in cyclonic activity over the Arctic Ocean (Serreze & Barrett, 2008;X. D. Zhang et al., 2004). In fact, values of monthly P over the Arctic Ocean are uniformly higher in summer than in winter although the precipitation amount around Ny-Ålesund is lower (Figure 8b).  Table 4). The high F MBC in winter was due to high C MBC and P values, and the low F MBC in summer was due to low C MBC and P values. Figure S3 shows the correlation between monthly F MBC and P for each season. The monthly F MBC values were strongly variable at the higher monthly P values in all seasons due to the large variability in the monthly P-weighted C MBC . We characterized the variance of the monthly F MBC by using the average and 1σ of monthly P-weighted C MBC and P, described in detail in the Appendix. The variance of monthly F MBC was associated with the variances of monthly P-weighted C MBC and monthly P to comparable degrees.

BC Wet Deposition Flux
MORI ET AL.

10.1029/2020JD034110
13 of 32  Mori et al. (2020), our estimate was much higher than their estimate for the latitude band 65 ± 2.5°N (0.07 mg m −2 month −1 ) and somewhat lower than their estimate for the latitude band 60°N ± 2.5°N (0.29 mg m −2 month −1 ). These results underscore the importance of wet removal at Ny-Ålesund during winter, which was as high as the estimate of BC emission flux around 60°N.  Their median values of daily C MBC and F MBC were 2.3 μg L −1 and 0.42 μg m −2 d −1 , respectively. We collected 15 snowfall samples at Ny-Ålesund during the same period. The median value of daily F MBC at Ny-Ålesund (20 μg m −2 d −1 ) was much higher than at Alert, a result of higher median values of daily C MBC (6.3 μg L −1 ) and P (4.6 mm d −1 ). The average (±1σ) ratio of C MBC (D BC = 70-380 nm) to total C MBC (D BC = 70-4,170 nm) was 0.62 ± 0.19 at Ny-Ålesund, indicating that differences in the BC size range measured were not the major cause of the large difference in F MBC between Ny-Ålesund and Alert.

Snow and Rain
BC in hydrometeors had similar mass size distributions in all seasons except summer, when the size distribution shifted toward smaller sizes (Figure 9c), whereas their number size distributions were similar throughout the year (Figure 9d). In addition, hydrometeors on snow and rainy days had similar seasonally averaged BC mass size distributions in all seasons except winter, when only a few samples were collected on rainy days (Figures 9e-9h). The number size distributions on snow and rainy days were also similar (not shown). Note that the annual mass and number size distributions of BC in hydrometeors (Figures 9a  and 9b, respectively) had large variabilities, represented by error bars (25th and 75th percentiles) for each size category.
We quantified the features of the BC size distributions with the following parameters: mass median diameter (MMD), count median diameter (CMD), geometric standard deviations (σ gm and σ gc ) for the mass and number size distributions in precipitation, the ratio of C MBC for larger particles (D BC = 600-4,170 nm) to total C MBC (D BC = 70-4,170 nm), or f 600 , and BC mass per particle m BC (≡C MBC /C NBC ), as summarized in Table 5. The f 600 and m BC parameters are useful for investigating changes in the size distribution of BC particles, and they were well correlated with each other (r 2 = 0.49). The choice of 600 nm is somewhat arbitrary, and other diameters can be used as thresholds, for example, 800 nm (f 800 ) and 1,000 nm (f 1000 ). The ratios f 800 and f 1000 were strongly correlated with f 600 (r 2 = 0.93 and 0.88, respectively), consistent with the result of Mori et al. (2019). The median values of f 600 and m BC on the sampling days were somewhat larger in winter than in summer, corresponding to the differences in the shapes of the size distributions in the two seasons (Table 5 and Figure 9).
The ratio of C NBC for smaller particles (D BC = 10-70 nm) to total C NBC (D BC = 10-4,170 nm) was estimated to be about 0.30 ( Figure S4). The mass contribution of the smaller particles to the total C MBC was negligible (Figure 9a).

Snow and Air
We investigated the size dependence of BC wet deposition by comparing the number size distributions of BC in hydrometeors and BC in ambient air. We measured the BC number and mass size distributions in air and the mixing states of BC-containing particles at Zeppelin Observatory from February 28 to 4 March 4, 2017 and from March 21 to 29, 2017, using our standard SP2 . These campaign data consisted of 1-h averages within the 70-850 nm D BC range and 1-h average mixing states of BC-containing particles with D BC of ∼186 nm. As an indicator of the mixing state, we used the mass-equivalent shell-tocore ratio, defined as the ratio of the diameter of a BC-containing particle with D BC = 186 nm to that of the BC-core particle. A detailed description of the method to derive the shell-to-core ratio is given in Moteki et al. (2019). Table S2 shows average values related to BC in air during the whole campaign period and those in snow samples collected at Sverdrup Research Station on three different days during the campaign period (March 4, 23, and 24, 2017). The average shape of the BC size distributions from the three precipitation days was very similar to the size distribution for the whole campaign period ( Figure S5). The median values of M BC , MMD, and shell-to-core ratio were also comparable to those measured by Zanatta et al. (2018)   although the C MBC and C NBC were higher by a factor of 2 and 3, respectively, and the MMD was somewhat lower. Figure 10a shows the daily average BC number size distributions in ambient air and falling snow on the three precipitation days, along with their combined averages; the corresponding BC mass size distributions are shown in Figure 10b. First, the ratio of the normalized number size distribution for BC in snow (dC NBC (D BC )/dlogD BC ) to that in air (dN BC (D BC )/dlogD BC ) was derived as This size-resolved removal efficiency, normalized to an average value of unity, is referred to as RE (D BC ) and is shown in Figure 10a.
MORI ET AL.  The RE for BC between D BC = 70 and 850 nm depended little on particle size, indicating that the shapes of the BC size distributions in ambient air at the altitude of Zeppelin Observatory and in falling snow were very similar. BC sizes in hydrometeors should be affected by those in ambient air not only at the surface, but also in the lower free troposphere . The monthly P-weighted C MBC was not directly linked with the near-surface M BC at Ny-Ålesund, as discussed in Section 4.2. Despite possible spatial variability in M BC , this result suggests that the BC size distributions have similar shapes, on average, from the PBL to the lower troposphere over Ny-Ålesund.

Removal Efficiency of BC
In general, aerosol particles are more efficiently activated as cloud condensation nuclei (CCN) as their size and hygroscopicity increase (Petters & Kreidenweis, 2007). Thus, the RE for BC is strongly influenced by the microphysical properties of BC-containing particles (e.g., their mixing state and size) because wet removal of aerosols is mainly controlled by the activation of aerosols to cloud droplets . Aerosol particles are also scavenged by collisions with water and ice particles within clouds and falling hydrometeors, as discussed later in this section.
The median shell-to-core ratio was found to be about 1.2 in the urban atmosphere of Tokyo in Japan, where aerosols were dominantly organic matter, and RE (D BC ) decreased with decreasing D BC Ohata, Schwarz, et al., 2016). This finding indicates that smaller BC-containing particles are not efficiently activated as CCN. On the other hand, the median shell-to-core ratio at a site strongly influenced by outflow from the East Asian continent, where the chemical composition of PM 2.5 is dominantly sulfate, was found to be about 1.4, and RE (D BC ) depended little on particle size within the 70-850 nm D BC range Yoshida et al., 2020), as discussed in more detail in Section 5.5. This suggests that aged BC-containing particles with D BC = 70 nm can be activated.
At Zeppelin Observatory, the median shell-to-core ratio of airborne BC particles was about 1.3 during the three March 2017 precipitation events. Ambient aerosol particles were routinely collected each day at Zeppelin Observatory by using a filter-pack sampler with no size cutoff and the mass concentrations of inorganic ions (e.g., SO 4 2− , NO 3 − , Na + , Mg 2+ , and Cl − ) were measured by ion chromatography. In addition, mass concentrations of organic carbon (OC), collected with a high-volume sampler with a PM 10 impactor on a biweekly basis, were measured by a thermal-optical transmittance analyzer (Sunset Laboratory, Tigard, OR) . The mass concentration of non-sea-salt sulfate is derived from concentrations of Na + , Mg 2+ , and Cl − (Hawley et al., 1988;Keene et al., 1986). These aerosol datasets are available at the Norwegian Institute for Air Research (NILU) data archive (http://ebas.nilu.no/). From February to April 2017, nonsea-salt sulfate accounted for about 66% of the total aerosol mass concentration (i.e., non-sea-salt sulfate, MORI ET AL.

10.1029/2020JD034110
18 of 32 nitrate, and OC). These data were used to estimate the effect of hydrophilic compounds on activation of BC-containing particles into cloud droplets, as discussed below.
The activation of aerosols also depends on the maximum water supersaturation in ambient air. Koike et al. (2019) reported that supercooled water droplets had number concentrations close to those of aerosol particles with dry diameters larger than about 70 nm at Zeppelin Observatory in the summer and winter-spring seasons of 2013-2015, indicating that such aerosols were activated during those seasons. In addition, for the summertime data, they estimated the maximum supersaturation in ambient air by using an air parcel model that thermodynamically calculates cloud particle formation from aerosols within an ascending air parcel, based on the number size distributions of aerosols and cloud droplets and an assumed aerosol composition of ammonium sulfate and updraft velocities of 6-20 cm s −1 . The observed monthly median aerosol number concentrations (70-809 nm) during the summers of 2013-2015 were very low (<120 cm −3 ) at Zeppelin Observatory , indicating relatively high maximum supersaturations (>0.3%) in ambient air depending on the updraft velocity. Jung et al. (2018) also reported, on the basis of simultaneous measurements of number concentrations of aerosol and CCN at Zeppelin Observatory during 2007-2013, that all accumulation mode aerosol particles (60-300 nm) were active as CCN at maximum supersaturations of about 0.2% or higher throughout the year.
We estimated critical supersaturations of BC-containing particles as was done by Ohata, Schwarz, et al. (2016). For this estimate, first, the volume fraction of BC core and BC-containing material was derived from the measured shell-to-core ratio of 1.3. Then we calculated the hygroscopicity parameter κ (Petters & Kreidenweis, 2007) of BC coating material. We made four assumptions for this calculation: (a) the chemical composition of BC coating is the same as that of non-BC aerosols composed of non-sea-salt sulfate, nitrate, and organic matter (OM); (b) mass ratios of OM to OC range from 1.2 to 2.4 (e.g., Aiken et al., 2008;Russell, 2003); (c) κ values of OM range from 0.0 to 0.12 (Lathem et al., 2013;Petters & Kreidenweis, 2007); and (d) the density of OM is about 1.4 g cm −3 (e.g., Kuwata et al., 2012).
The κ values of BC-containing particles were 0.35-0.40 and 0.24-0.32 for OM/OC ratios of 1.2 and 2.4, respectively, within the 0.0-0.12 range of κ values for OM. By using κ-Köhler theory (Petters & Kreidenweis, 2007), we calculated the critical supersaturation of BC-containing particles with D BC = 186 nm to be 0.07%-0.09% for OM/OC ratios of 1.2-2.4. We also calculated the critical supersaturation of BC-containing particles with D BC = 70 nm by estimating their median shell-to-core ratio. The shell-to-core ratio at D BC = 70 nm was obtained by extrapolation from D BC = 186 nm, assuming condensation of sulfuric acid vapor on the surface of BC particles, as described in detail elsewhere (Mori, 2017;Seinfeld & Pandis, 2016). The calculated median shell-to-core ratio for D BC = 70 nm was 1.42, yielding a critical supersaturation of 0.26%-0.33% for OM/OC ratios of 1.2-2.4. These estimates were comparable to or lower than the observed maximum water supersaturations. These results suggest that aged BC-containing particles as small as D BC = 70 nm are removed mainly by activation into cloud droplets. Impaction processes also remove aerosol particles, by collisions between aerosol and cloud particles (water and ice) within clouds and by collisions between aerosol particles and falling hydrometeors below clouds, mainly by Brownian diffusion, interception, and inertial impaction (Wang et al., 2010;Zhang et al., 2013). According to theoretical and experimental studies, the collection efficiency of aerosol particles scavenged in these two settings is low within the 100-3,000 nm aerodynamic diameter range (Wang et al., 2010;Zhang et al., 2013), depending on the sizes and shapes of falling hydrometeors. Moteki et al. (2019) measured the size-resolved RE of BC for 37 rainfall events in East Asia and estimated the relative contributions of nucleation scavenging, in-cloud impaction scavenging, and below-cloud impaction scavenging by using a cloud parcel model including aerosol-cloud impaction processes. The fractional contribution of nucleation scavenging was greater than 0.8 within a 200-2,000 nm D BC range. In midlatitudes, the contributions of in-cloud and below-cloud impaction scavenging increase at D BC < 200 nm to some extent, due to the lower maximum supersaturation in ambient air (∼0.08%) . However, at Ny-Ålesund where the maximum supersaturation is higher (>0.3%), as discussed above, BC-containing particles would be removed mainly by nucleation scavenging.
It should be noted that large fractions of BC are transported from lower latitudes, as shown by trajectory analyses (Eleftheriadis et al., 2009;Liu et al., 2015) and model calculations (AMAP, 2015;Hirdman et al., 2010). We show in Section 5.5 that freshly emitted BC particles become thickly coated during transport from their source regions. Therefore, it is likely that the CCN activities of BC-containing particles in the Arctic are largely set before they reach the Arctic, although the particles are further processed in the Arctic. Therefore, it seems natural that the CCN activities of BC-containing particles are high in general.

Comparison of BC Wet Deposition at Ny-Ålesund and Barrow
In this section, we describe BC wet deposition at Ny-Ålesund (2012Ny-Ålesund ( -2019 and Barrow (2013Barrow ( -2017 on the basis of M BC , C MBC , P, F MBC , and size distribution. These parameters at Barrow are summarized in Mori et al. (2020). For comparison, we used the median of 87 monthly values at Ny-Ålesund during 2012-2019 and that of 35 monthly values at Barrow during 2013-2017 for each parameter, referred to here as the annual median. The annual median values for these and other parameters at Ny-Ålesund and Barrow are summarized in Table 6.

BC Mass Concentration in the PBL
Median monthly M BC at Ny-Ålesund and Barrow had similar seasonal variations, high in winter and spring and low in summer (Figure 5a). Analysis of backward trajectories starting from Zeppelin Observatory during 2002-2007 has shown that polluted air masses were commonly transported from northern Eurasia during winter and relatively clean air masses were transported mainly from the northern Arctic during summer (Hirdman et al., 2010). At Barrow, 10-day backward trajectories from 925 hPa during 1989-2003 have shown that air masses were commonly transported from western and eastern Russia during winter and from the Pacific and western China-Pacific regions during summer (Sharma et al., 2006). In addition, radiosonde measurements in winter showed that average PBL heights at Ny-Ålesund (∼640 m) and Barrow (∼230 m) were higher than the altitudes of the observation sites. These results indicate that high M BC during winter and spring at both sites were caused by the transport of polluted air masses from mid and high latitudes, and the accumulation of BC in the PBL. In addition, the low M BC in summer results from enhanced BC wet deposition during long-range transport, as described in Section 4.1 (AMAP, 2015;Browse et al., 2012;Hirdman et al., 2010;Quinn et al., 2007;Sharma et al., 2006Sharma et al., , 2013. The annual median M BC value at Ny-Ålesund was lower than at Barrow by a factor of 1.8 (Figure 5a and Table 6). The difference between these localities may arise from the difference in the distances from BC emission sources, the different observation altitudes, and the degrees of dilution through mixing and BC wet removal during transport of air from lower latitudes.

BC Mass Concentration in Hydrometeors
Median monthly P-weighted C MBC at Ny-Ålesund are compared with that at Barrow in Figure 5b (C NBC values are compared in Figure S1a). The annual median P-weighted C MBC was lower at Ny-Ålesund than at Barrow by a factor of 2.8 (Figure 5b and Table 6). From January to May, the variations of C MBC at the two sites were similar, although C MBC was lower by factors of 2-3 at Ny-Ålesund (Figure 5b).
The variability of the monthly P-weighted C MBC was generally larger at Barrow than at Ny-Ålesund, partly because Barrow had fewer data. The annual mean (±1σ) value of the number of sampling days at Barrow was 8.4 ± 4.8 per month during 2013-2017 (Figure 5e and Table S1), which is about one-fifth that at Ny-Ålesund (46 ± 12). However, the relatively smooth seasonal variation of median monthly P-weighted C MBC MORI ET AL.  Table 6 Annual Median Values at Ny-Ålesund From August 2012to October 2019and Barrow From July 2013to August 2017 and the similar seasonal patterns of monthly and daily data (not shown) suggest that the average seasonality of monthly P-weighted C MBC is not strongly influenced by small data numbers, except for May, June, and November. The high variability in July and August is discussed below.
Median monthly values of P-weighted C MBC and M BC at Barrow were moderately correlated (r 2 = 0.17) except for July and August (Figure 6c), although the median monthly value was based on about three average values (Table S1). In addition, the median daily values of C MBC and M BC were also moderately correlated (r 2 = 0.15; Figure 6d) for days with P > 0.5 mm, which account for about 80% of the data. The r 2 values were lower at Barrow than at Ny-Ålesund (section 4.2).
In July and August, C MBC at Ny-Ålesund decreased, as did M BC ; however, C MBC at Barrow remained high, while M BC greatly declined which suggests that C MBC did not directly reflect M BC near the surface (Figure 5b). Mori et al. (2020) sought to explain this discrepancy at Barrow by estimating the contributions of C MBC separately for anthropogenic and biomass burning emissions using two different global models. The model calculations showed that M BC increases in the free troposphere in summer due to BC emissions from biomass burning and that BC particles are removed by precipitation below ∼2 km. This result suggests that the high C MBC at Barrow is strongly influenced by increased M BC in the lower troposphere from biomass burning emissions.
To investigate the effects of biomass burning on C MBC at Ny-Ålesund during summer, we used the Global Fire Assimilation System data set (Kaiser et al., 2012) to calculate zonally averaged biomass burning emission fluxes for three latitude bands (55°N ± 2.5°N, 60°N ± 2.5°N, and 65°N ± 2.5°N) at a horizontal resolution of 0.1° latitude/longitude, and compared the results with monthly M BC and P-weighted C MBC at Ny-Ålesund and Barrow (Figure 11). The biomass burning emission fluxes reached their maximum during July and August, corresponding to the seasonal pattern of C MBC at Barrow (Figure 5b), whereas the monthly P-weighted C MBC at Ny-Ålesund was much lower during those two months. This result suggests that the effect of biomass burning on C MBC is much smaller at Ny-Ålesund than at Barrow, possibly because of transport of cleaner air masses in the lower free troposphere over Ny-Ålesund, although we made no quantitative analysis of the potential effect of biomass burning in this study. Note that daily C MBC values were sometimes MORI ET AL. The two global model calculations by Mori et al. (2020) showed that the correlation between monthly C MBC and M BC near the surface at Barrow was relatively poor, whereas the monthly BC column amount and C MBC were well correlated, suggesting that monthly C MBC is more strongly controlled by M BC in the free troposphere over Barrow than by M BC near the surface. Thus, the observed poor correlation between monthly C MBC and M BC at Ny-Ålesund may also arise from differences in the variations of monthly M BC in the free troposphere and in the PBL. It should also be noted that the model calculations had a horizontal spatial resolution of about 100-200 km (Matsui et al., 2018;Oshima et al., 2020). This resolution would smooth the variability in C MBC and M BC , possibly leading to stronger correlations than those observed. This point is important in comparing model calculations with the in-situ data.

Precipitation and BC Wet Deposition Flux
Median monthly P at Ny-Ålesund was low in summer and high in winter, whereas at Barrow it was high from late spring to early fall ( Figure 5c). During winter, the median monthly P value at Ny-Ålesund was about eight times greater than at Barrow, whereas in summer the P values at Ny-Ålesund and Barrow were comparable. Annual median P values were about 3 times greater at Ny-Ålesund than at Barrow (Figure 5c and Table 6).
The seasonal variations of median monthly F MBC had very different patterns at Ny-Ålesund and Barrow, corresponding to their differences in C MBC and P (Figure 5d). The seasonal patterns of F MBC and F NBC were similar for each site ( Figure S1c). At Ny-Ålesund, median monthly F MBC was highest in winter, whereas it was highest in summer at Barrow. The amount of BC wet deposition per unit area was 0.71 mg m −2 at Barrow during summer , which is comparable to our estimate for Ny-Ålesund during winter (0.58 mg m −2 ). The annual median F MBC values at Ny-Ålesund (65 μg m −2 month −1 ) and at Barrow (62 μg m −2 month −1 ) were also comparable ( Figure 5d and Table 6).

Size Distributions
The annually and seasonally averaged BC mass size distributions in hydrometeors were very similar at Ny-Ålesund and Barrow, and the annually averaged BC number size distributions at both sites were also similar ( Figure 12). Annual median values of the parameters MMD, CMD, σ gm , σ gc , f 600 , and m BC at Ny-Ålesund agreed with those at Barrow within 10% (Table 6). These similarities are striking, considering the differences between Ny-Ålesund and Barrow in the concentrations of BC (M BC and C MBC ), meteorological conditions (P), emission sources, and transport pathways and the large distance between the two sites ( Figure 1). In this section, we attempt to explain these features and relate the size distributions of BC in hydrometeors at Ny-Ålesund and Barrow with those measured in the atmosphere at Ny-Ålesund.

Microphysical Properties of BC and Supersaturation
As described in Section 4.5.2, the BC size distributions in air and falling snow at Ny-Ålesund in spring 2017 were very similar, showing that RE of BC within the 70-850 nm D BC range was size-independent because of the efficient activation of BC-containing particles into cloud droplets. This size independence is consistent with the microphysical properties of BC and supersaturations observed by aircraft observations in the Arctic, as summarized below.
PAMARCMiP aircraft observations in spring 2018 showed that BC particles at 0.2-5.2 km altitudes north of 78°N were thickly coated, with median shell-to-core ratios of 1.35-1.45 in the 186-236 nm D BC range . Thickly coated BC-containing particles were also dominant below 4 km over the Canadian High Arctic region in summer 2014 (72.5°N-77.5°N, 100°W-80°W) and spring 2015 (80°N-85°N, 105°W-60°W) in observations of the NETCARE aircraft campaign (Kodros et al., 2018). These results strongly suggest that BC particles are thickly coated in general in the Arctic.
The degree of activation of aerosol particles into cloud droplets has also been studied by aircraft observations in the Arctic (Leaitch et al., 2016). The NETCARE aircraft campaign in summer 2014 measured number concentrations of cloud droplets and CCN and showed high supersaturation (0.3%-0.6%) in ambient air at 0.06-6.0 km altitudes in the Canadian High Arctic (Leaitch et al., 2016). The observed high supersaturation is comparable to that observed at Zeppelin Observatory .
The aircraft observations during PAMARCMiP 2018 and NETCARE 2015 documented an altitude dependence of BC size distributions in spring (Schulz et al., 2019;Yoshida et al., 2020), in which the MMD of BC at high latitudes (>70°N) gradually decreases from 200 nm at 1 km to 150 nm at 5 km.

Altitudes of BC Wet Removal and Altitude Dependence of BC Size
Global model calculations by Mori et al. (2020) showed that BC wet removal occurs mainly at altitudes below ∼2 km at Barrow throughout the year. The altitudes of BC wet removal were calculated to be similar at Ny-Ålesund (not shown).
The change in MMD for BC size distributions from the PBL to the lower troposphere (<2 km) was small (190-200 nm) over the Canadian High Arctic in summer 2014 and spring 2015 (Schulz et al., 2019).

BC Size Distributions in the Lower Troposphere
Combining these observations and model calculations (Sections 5.4.1 and 5.4.2), it appears likely that the shapes of the BC size distributions in hydrometeors reflected those in the lower troposphere at Ny-Ålesund and Barrow. Therefore, it is likely that the average shapes of BC size distributions from the surface to the lower troposphere were also similar at the two sites.
Note that BC size distributions in the lower troposphere may not be uniform at lower latitudes. Mori et al. (2019) investigated the latitudinal dependence on the size of BC particles deposited on Arctic snow in the latitude band 60°N-80°N, where local BC emissions are negligible during the snow accumulation period, and reported that BC size distributions in snowpack shifted toward larger sizes at lower latitudes.

Processes That Control BC Size Distributions in the Arctic
The annual median (25th percentile, 75th percentiles) f 1000 values were 0.09 (0.03, 0.21) at Ny-Ålesund and 0.09 (0.06, 0.15) at Barrow, indicating that a substantial BC mass was in the super-micron range (D BC > 1,000 nm) in hydrometeors; therefore, f 1000 was also similar in the lower troposphere throughout the year. Super-micron BC particles have also been found in the Arctic and Antarctic in hydrometeors, surface snow, and snowpack (Kinase et al., 2020;Marquetto et al., 2020;Mori et al., 2019Mori et al., , 2020Sinha et al., 2018).
Generally, larger BC particles should be removed more efficiently due to their presumably higher CCN activities. However, the ubiquitous existence of super-micron BC is apparently inconsistent with this supposition. We need to reconcile this feature with our understanding of the processes that control BC size distributions and mixing states. In this section, we consider three of these processes on the basis of observations of microphysical properties and wet removal of BC at urban and remote sites in East Asia Moteki et al., 2019;Yoshida et al., 2020).

Ground-Based Measurements of BC in East Asia
Previous studies have measured the size distributions of BC in air and rainwater for 29 precipitation events in urban Tokyo during the summers of 2014-2015 Moteki et al., 2019;, for six events on Fukue Island in the spring of 2019 , and for 11 events at Cape Hedo in Okinawa during the spring of 2016 , as shown in Figure 13. Fukue and Cape Hedo are in the East China Sea about 250-600 km and 800-1,000 km from the coasts of the Asian continent (China and Korea), respectively ( Figure 28 of Kondo, 2015). They are strongly influenced by Asian outflows and little influenced by local BC emission sources (Kondo, 2015;Shiraiwa et al., 2008). Typical transport times of air parcels from the Asian continent in spring are about 1-2 days to Fukue and about 2-3 days to Hedo Shiraiwa et al., 2008). Figure 13d shows the RE at all three sites, and the median values of BC shell-to-core ratios for D BC = 186 nm are summarized in Table 7. In Tokyo, the RE increased with D BC , indication of thin coating. This size dependence of RE was much weaker at Fukue and Hedo, where BC particles were thickly coated. At Hedo, RE was independent of D BC , similar to the case at Ny-Ålesund (section 4.5.3).
Process 1 RE in an air parcel becomes size-independent by BC aging during transport, unless fresh BC particles are added to the parcel.

Transport of BC to the Arctic PBL and Lower Troposphere
Transport of BC to the Arctic in the PBL is most efficient in winter and spring, when the edge of the polar dome shifts as far south as 40°N (AMAP, 2015; Barrie, 1986). North of 40°N, BC is emitted mainly at 40°N-60°N and is transported in the cold PBL over long distances. BC can also reach the Arctic lower troposphere by ascending over the polar dome (AMAP, 2015;Browse et al., 2012). These BC particles generally are continuously coated by non-BC (inorganic and organic) species during transport through condensation and coagulation processes, unless fresh BC particles are added to the air parcel through significant injections or mixing events. At some stage, the BC size distributions in a parcel reach a stable shape, even after precipitation events, because RE becomes size-independent, although the BC mass concentrations change. In fact, BC in the Arctic (>70°N) is thickly coated (Section 5.4.1). Thus, BC size distributions would tend to have stable shapes in the Arctic PBL and lower troposphere, and super-micron BC particles would be ubiquitous there. Here, we assumed that wet deposition predominates over dry deposition for the supermicron BC particles because aerosol particles as large as a few micrometers have low settling velocities (Seinfeld & Pandis, 2016

Process 2
The BC size distribution stabilizes from the surface to the lower troposphere at higher latitudes, although vertical mixing may decrease the altitude dependence of the size distribution.

Aircraft Measurements in East Asia
Aircraft observations over the East China Sea have shown that the size distributions of BC in the free troposphere shift to smaller sizes with increasing altitude, as a result of wet removal over the East Asian continent Moteki et al., 2012). The majority of these BC particles were uplifted from the PBL into the free troposphere over large sources in China. Over China, BC particles in the PBL have younger average ages than at Hedo because of continued BC emissions in China, which leads to size-dependent RE, as observed in Tokyo (Figure 13d). At higher altitudes, BC undergoes greater wet removal, on average, and the average BC size decreases with altitudes . Size-dependent RE accounts for the observed altitude dependence of the BC size distributions.
Process 3 Smaller BC particles are transported more efficiently to higher altitudes, when they are uplifted from the PBL to the free troposphere over large continental sources.

Transport of BC in the Arctic Free Troposphere
BC size distributions shift toward smaller sizes with increased altitude (1-5 km) at >70°N (Section 5.4.1). BC particles emitted south of the Arctic front can reach the Arctic middle or upper troposphere across surfaces of constant potential temperature (AMAP, 2015). Process 3 makes it likely that the altitude dependence of BC size distributions observed in the Arctic is mainly inherited from continental BC sources at lower latitudes, although the shape of the size distributions may depend to some extent on the type of BC source (Schwarz et al., 2008).
Our discussion is based on observations of BC in East Asia. However, processes 1-3 occur in all regions at various rates. An important point is that BC particles in source regions are linked with those in the Arctic by transport with time scales longer than the time scale of BC aging. It is very difficult to calculate the microphysical properties of BC by numerical models; however, qualitative understanding of the related processes will be useful in interpreting the observed results.

Seasonal Variation of BC Size Distributions
Precipitation at midlatitudes is generally highest in summer (e.g., Figure 8b), and thus BC size distributions would shift to smaller sizes more effectively in summer than in other seasons. This seasonal variation would propagate to the Arctic by long-range transport, causing the seasonal variations of BC size distributions in hydrometeors there.
In general, BC particles are more strongly influenced by biomass burning in summer than in other seasons (Section 5.2), and BC particles emitted from biomass burning have somewhat larger mean diameters than those from fossil fuel combustion (Kondo, Matsui et al., 2011;Schwarz et al., 2008). However, the influence of seasonal variations of BC wet removal on BC size distribution appears to be larger than the influence of biomass burning.

Effect of Atmospheric Circulation in the Arctic
Air parcels circulate in the Arctic (>70°N), although they continually enter and exit the Arctic (Hirdman et al., 2010;Huang, Gong, Sharma, et al., 2010). This circulation may occasionally send similar air parcels over Ny-Ålesund and Barrow. The similarity of the BC size distributions at Ny-Ålesund and Barrow (Figure 12) may be influenced by the presence of similar types of air parcels, although the parcels arrive at different times. The variability in BC size distributions at the two sites, associated with the differences in sampling times is smoothed by time-averaging, which is done in producing Figure 12.

Effect of Coalescence of Cloud Droplets
After airborne BC-containing particles form cloud droplets, the droplets grow by water vapor condensation and by collisions between them (Straka, 2009). As a large droplet collects smaller droplets, it accumulates the BC particles in the small droplets by coalescence. If such a collector droplet evaporates without reaching the ground, the BC particles it contains can coagulate into a larger BC particle that is released to the atmosphere, as theorized by Moteki and Mori (2015) and demonstrated in laboratory studies by Mori et al. (2016).
BC size distributions may shift to larger sizes as this process is repeated during long-range transport from source regions to the Arctic. Although our discussion concerns liquid droplets, the processes of coalescence and evaporation apply similarly to ice particles. It is very difficult to quantify this effect given the complexity of these related processes in the atmosphere; however, the effect is limited to cloudy and precipitation days, whereas processes 1-3 occur readily in any sunlit region and therefore can be presumed to be more effective.

Summary and Conclusion
Wet deposition of BC is a key process controlling the spatial and temporal distribution of BC mass concentrations in air and snow. However, the scarcity of relevant measurements limits our understanding of BC distributions in hydrometeors and BC wet deposition in mixed-phase clouds in the Arctic. To better document BC wet deposition in the Arctic, we measured BC mass concentrations in air (M BC ) and in hydrometeors (C MBC ) with accuracies of 15% and 16%, respectively, at Ny-Ålesund from August 2012 to October 2019, along with meteorological parameters and cloud-base heights. We then compared our measurement parameters at Ny-Ålesund (M BC , C MBC , F MBC , P, and BC size distributions in hydrometeors) with those at Barrow from 2013 to 2017 published by Mori et al. (2020).
Cloud-base height ranged from about 1-2 km with a clear seasonal variation, being high in winter and low in summer. The similar cloud-base temperatures at the two sites suggests that a portion of hydrometeors including BC originated in or passed through mainly mixed-phase clouds.
Median monthly M BC and P-weighted C MBC at Ny-Ålesund were highest in spring and lowest in summer. These two measurements were well correlated (r 2 = 0.78) in months with at least 40 data points during the study period. Thus, on a median monthly basis, the BC mass deposited per unit of precipitated water increases with increasing M BC .
Wet deposition of BC (median monthly F MBC ) showed marked seasonal variation, being highest in winter (96 µg m −2 month −1 ) and lowest in summer (28 µg m −2 month −1 ), associated with the seasonal patterns of C MBC and P. Wet deposition during winter accounted for about 35% of the annual wet deposition amount, reflecting higher precipitation in that season.
At Ny-Ålesund, the average BC size distributions in hydrometeors showed little temporal variation except in summer, and BC size distributions in hydrometeors on snow and rainy days were also similar. Size distributions of BC in hydrometeors and in ambient air measured in spring 2017 were very similar, suggesting that BC-containing particles with diameters larger than 70 nm were efficiently activated as CCN at Ny-Ålesund.
The median monthly M BC at Ny-Ålesund and Barrow showed similar seasonal patterns, being highest during winter-spring and lowest in summer. By contrast, the median monthly P-weighted C MBC at these sites showed different seasonal patterns, being low at Ny-Ålesund but high at Barrow during summer even as M BC at the surface was low at both sites. Global model calculations have shown that the enhancement of C MBC at Barrow is due to biomass burning . Considering this, our results suggest a lesser influence of biomass burning on C MBC at Ny-Ålesund in summer, likely due to differences in the transport pathways of air parcels affected by biomass burning. Median monthly F MBC at Ny-Ålesund and Barrow had different seasonal patterns due to different seasonal patterns of C MBC and P at these sites.
Median monthly P-weighted C MBC and M BC at Barrow were moderately correlated (r 2 = 0.17) except for July and August. The lower r 2 value is partly due to the smaller number of data from Barrow than from Ny-Ålesund.
Annual median values of M BC , C MBC , P, and F MBC at Ny-Ålesund were 6.6 ng m −3 , 1.9 µg L −1 , 34 mm month −1 , and 65 µg m −2 month −1 , respectively. At Barrow, annual median values of M BC , C MBC , P, and F MBC were 12 ng m −3 , 5.4 µg L −1 , 11 mm month −1 , and 62 µg m −2 month −1 , respectively. Annual median M BC and C MBC values were lower at Ny-Ålesund than at Barrow by factors of about 2 and 3, respectively. On the other hand, the annual median P values were about 3 times higher at Ny-Ålesund than at Barrow. As a result, the two sites had comparable annual median F MBC values.
BC particles observed at Ny-Ålesund in spring 2017 were thickly coated, indicating high CCN activity. In addition, previous aircraft observations have shown that BC particles are thickly coated in the Arctic PBL and free troposphere.
Annual and seasonal average BC size distributions in hydrometeors at Ny-Ålesund were very similar to those at Barrow. Considering the high CCN activities of BC-containing particles, the similarities suggest that BC size distributions from the PBL to the lower troposphere were also similar, on average, at these sites. This is a striking feature considering the large differences in environmental conditions and the distance between the two sites.
Measurements in East Asia have shown that the RE of BC becomes insensitive to BC size within 3-4 days after emission due to coating of BC by non-BC species. North of 40°N, BC is emitted mainly at 40°N-60°N and is transported to the Arctic in the PBL and lower troposphere over long distances in winter and spring. The measurements imply that BC particles continue to be coated by non-BC species during transport. At some stage, the BC size distributions attain a stable shape because RE becomes size-independent, and they tend to become uniform as mixing proceeds. This mechanism also explains the ubiquitous presence of super-micron BC, the seasonal variation of BC size distributions, and the mixing states of BC observed in the Arctic. Coalescence of BC by evaporation of cloud droplets may tend to slightly increase the size of BC particles.
Previous aircraft observations have shown that BC size distributions shift toward smaller sizes with increasing altitude between 1 and 5 km. BC particles emitted south of the Arctic front can reach the Arctic middle or upper troposphere across surfaces of constant potential temperature. It is likely that the altitude dependence of BC size distribution observed in the Arctic is mainly inherited from lower latitudes closer to the continental BC sources.
The reliable datasets of M BC and F MBC at Ny-Ålesund and Barrow will be useful for constraining and testing climate models. The BC size distributions in hydrometeors at Ny-Ålesund and Barrow will be useful in estimating the effect of BC on snow albedo in wide regions of the Arctic. More observations of M BC in the lower troposphere, C MBC , and F MBC will further improve our quantitative understanding of the spatial and temporal distributions of BC deposited in the Arctic.

Appendix A: Controlling Factors of Variance of Wet Deposition Mass Flux of BC
Assuming that monthly P-weighted C MBC and P are independent variables, the uncertainty in monthly F MBC (σ FMBC ) is given as where σ CMBC and σ P are 1σ uncertainties of monthly P-weighted average C MBC and monthly P during 2012-2019, respectively. By dividing both sides in Equation A1 by (F MBC ) 2 , the relative uncertainty of monthly F MBC is represented by the relative uncertainties of monthly C MBC and P: We compared the relative uncertainties of monthly P-weighted C MBC and P by using their averages and 1σ. The two relative uncertainties were comparable, suggesting that the variance of monthly F MBC was controlled by both C MBC and P, as might be expected. We also estimated the relative uncertainty of daily F MBC during 2012-2019, showing that both daily C MBC and P strongly control daily F MBC .