Rapidly rising temperatures and loss of snow and ice cover have demonstrated the unique vulnerability of the high Arctic to climate change. There are major uncertainties in modelling the chemical depositional and scavenging processes of Arctic snow. To that end, fresh snow samples collected on average every 4 days at Alert, Nunavut, from September 2014 to June 2015 were analyzed for black carbon, major ions, and metals, and their concentrations and fluxes were reported. Comparison with simultaneous measurements of atmospheric aerosol mass loadings yields effective deposition velocities that encompass all processes by which the atmospheric species are transferred to the snow. It is inferred from these values that dry deposition is the dominant removal mechanism for several compounds over the winter while wet deposition increased in importance in the fall and spring, possibly due to enhanced scavenging by mixed-phase clouds. Black carbon aerosol was the least efficiently deposited species to the snow.
In recent decades drastic changes have been observed within the Arctic, including a rapid increase in surface temperatures and loss of sea ice and snow cover (Rigor et al., 2000; Stroeve et al., 2005; Hartmann et al., 2013). Not only have these changes had adverse consequences for local populations and ecosystems, it has been suggested that their impacts may be significant at the global scale (Law and Stohl, 2007; AMAP, 2011). Light-absorbing compounds, the most widely studied of which being black carbon (BC) particles, can have a particularly significant impact on the Arctic atmosphere and snow systems through the absorption of solar radiation and subsequent warming and snowmelt (Bond et al., 2013). While the Arctic atmosphere has been previously explored spatially, temporally, and compositionally (e.g., Hartmann et al., 2013), Arctic snow and the mechanisms linking snow to the atmosphere have been the subject of only a relatively small number of studies (AMAP, 2011) despite the enormous amount of research conducted on the Arctic haze phenomenon (Quinn et al., 2007). Seasonal observations of fresh snow samples are particularly uncommon (e.g., Davidson et al., 1993; Toom-Sauntry and Barrie, 2002; Hagler et al., 2007) and previous explorations of snow deposition and scavenging mechanisms have been largely reliant on short-term or aged snowpack sampling (e.g., Bergin et al., 1995), ice cores (e.g., Legrand and De Angelis 1995), modelling, and laboratory tests.
Aerosols entering the Arctic atmosphere, either generated locally or transported from elsewhere, can be removed by atmospheric transport or deposition. Deposition of particles follows two mechanisms: dry deposition, whereby particles are deposited to the ground by impaction, gravitational settling, and Brownian motion; and wet deposition, whereby particles are scavenged by hydrometeors and deposited through precipitation. Wet deposition is further split into two scavenging mechanisms: in-cloud scavenging, which removes particles from the cloud layer during precipitation formation, and below-cloud scavenging, which removes particles from the atmospheric column through which precipitation falls. Gaseous compounds also undergo similar scavenging processes (Seinfeld and Pandis, 2006).
The rate of dry deposition is dependent on the properties of the depositing
particle, the surface onto which deposition occurs, and the air–surface
boundary layer (Sehmel, 1980; Zhang and Vet, 2006). Dry deposition
velocities of accumulation-mode particles, the dominant mass-weighted mode
of particles observed in the non-summer Arctic (Sharma et al., 2013), to
snow have been modelled and observed over a range of 0.01 to 0.60 cm s
The goals of this paper are to present a new dataset in which the chemical composition of freshly fallen snow was measured through a fall–winter–spring period at a high Arctic field station. By combining these data with simultaneous measurements of ambient aerosol, the efficiency of deposition of individual species from the atmosphere to the snow can be evaluated under a set of broad assumptions. While this paper presents the measurement dataset in detail and focuses on the depositional and scavenging mechanisms that can be inferred from it, a subsequent publication will identify potential pollutant sources based on the snow compositional data. To our knowledge, this is the first time that the composition and flux of freshly fallen snow has been analyzed at high temporal frequency throughout an entire cold season in the high Arctic. All data from this study will be available upon conclusion of the NETCARE project via the Government of Canada Open Data Portal.
Snow samples were collected at Environment and Climate Change Canada's (ECCC)
Neil Trivett Global Atmosphere Watch Observatory at Alert, Nunavut, from
14 September 2014 to 1 June 2015 as part of the Network on Climate and
Aerosols Research (NETCARE) initiative to create a temporally refined and
broadly speciated dataset of high Arctic snow measurements. Alert is a remote
outpost in the Canadian high Arctic, at the northern coast of Ellesmere
Island (82
The collection of fresh snow samples reduces the impact of snow sublimation and/or melt as well as the movement of chemical species between snow and air, which can be a concern for snowpack sampling; however, some bi-directional exchange between snow and atmosphere is unavoidable within natural snowpack and still expected to smaller extent on the snow table. Also, the collection of samples from a snow table eliminated the difficulty in distinguishing the fresh stratigraphic snow layer from aged layers below, a source of uncertainty for traditional surface snow sampling. This ability to assign a well-defined deposition area and time period to each sample was an advantage over traditional sampling campaigns of aged snowpack. However, both this and traditional snow collection techniques are prone to the uncertainty introduced by the redistribution of snow by winds. Measurements of snowfall accumulation were not available for the collection site. Snow depths measured at the Alert ECCC station indicate that the snow collected on the tables may have underestimated the total snowfall volume by a factor of approximately 1 to 10; however, the meteorological station and collection site were separated by over 6 km with a 50 m difference in elevation, and there was significant disagreement between operator records of weather and that indicated by the meteorological station (see Sect. S4.2 for details). Thus, it was unclear whether this disagreement was the result of snow loss from the snow table or the natural spatial variability in precipitation, and no correction was applied to the collected snow depth. Furthermore, it should be noted that dry deposition via the filtration of air as it is pumped through the snowpack (as described in Harder et al., 1996) may differ between snow on a snow table and that on the surface.
Alert station operators recorded the collection conditions for each sample. Atypical snowfall events were noted: diamond dust events, small crystalline snowfalls, and blowing snow events, periods when high winds potentially resuspended snow from the ground. Operators also made note of any unusual weather conditions such as fog or blizzard conditions. Local ground-level meteorological conditions were monitored by the Alert ECCC stations, approximately 6 km NNE of the collection site (station IDs 2400306, 2400305, and 2400302; retrieved November 2015 from climate.weather.gc.ca). In addition to ground-level meteorological information, vertical profiles were monitored via 6 to 12 h radiosondes. The radiosonde data were used to estimate mixing height and cloud height over the campaign. Mixing height was taken as the lowest altitude corresponding to an inflection point in the potential temperature. When the potential temperature gradient did not change from negative to positive within the lowest 3 km, no mixing height was found. The vertical humidity profiles were used to identify cloud height as the lowest altitude, within 3 km of the surface, with 100 % relative humidity. When 100 % humidity was not reached, this criterion was relaxed to 95 %. Details of meteorology data are provided in Sect. S4.2.
All snow samples were kept frozen prior to analysis, throughout storage and shipping. A broad suite of analytes was quantified using replicate snow samples from each snowfall: BC, major ions, and metals. Detailed procedures are provided in the Sect. S2.
Refractory BC quantification was completed via single-particle soot photometry (SP2) as per McConnell et al. (2007). Briefly, melted and sonicated snow samples were atomized via Apex-Q nebulizer and dried particles with 0.02 to 50 fg BC were quantified via SP2. Observed BC mass distributions did not suggest significant underestimation of the total BC mass due to this size cut-off. A quality control standard and an analysis blank were analyzed for every batch of 17 samples.
Overview of fresh snow composition and inferred fluxes during the 2014 to 2015 winter season.
Notes: BC is black carbon; MS is methanesulfonate; ACE is acetate; PRP
is propionate; FOR is formate.
< no. indicates measurement is below MDL. n/a
Major ions were measured via ion chromatography (IC) at ECCC, as per Toom-Sauntry and Barrie (2002). Briefly, melted samples were quantified using a Dionex IC: DX600 for anions and cations and ICS2000 for organic acids. Aliquots of these samples were also used for pH analysis (Denver pH analyzer). Equipment was calibrated daily and quality control runs completed every 10 samples.
Metals analysis was completed via inductively coupled plasma mass
spectrometry (ICP-MS) at the University of Toronto. Briefly, melted samples
were filtered to separate insoluble and soluble metals (considered as that
which was retained or passed through a 0.45
Quality assurance is of the upmost importance in the analysis of dilute
Arctic samples. Instrument accuracy was confirmed through the analysis of
certified reference materials. The uncertainty of each measurement was
estimated based on analysis detection limits and reproducibility; details are
provided in Sect. S2 (as per Reff et al., 2007; Norris et al., 2014). Also,
the signal-to-noise ratio (S/N) of each analyte was calculated to indicate the
strength of each measurement, with a S/N value over one considered to be
strong (Norris et al., 2014). Regular analysis of blanks was used for
background subtraction and to define method detection limits (MDL) as 3
standard deviations of the blank levels. Beyond typical preparation blanks,
which used DIW in the place of snow meltwater, field blanks were also
analyzed. Once per month, a set of empty sample bottles was brought to the
snow table, opened, and resealed without collection. These field blank
bottles were stored and shipped with the regular samples and rinsed with DIW
to quantify any contamination throughout the sampling process. Any influence
from the local Alert base camp was identified using local wind records and
the activity logs of the base camp personnel. The only analytes that showed a
potential influence from base camp winds were crustal metals, with Pearson's
correlation coefficients (
Ground-level atmospheric monitoring data from the Alert Global Atmospheric Watch Observatory were provided by ECCC (see Sect. S3 for details). Atmospheric BC was monitored hourly by SP2 (Droplet Measurement Technology) (as per Schroder et al., 2015) and major ions by IC of 6 to 8-day high-volume filters of total suspended particles (Hi-Vol) (as per Sirois and Barrie, 1999). Both the SP2 and Hi-Vol were operational throughout the campaign with coverages of 92 and 94 %, respectively.
The Lagrangian particle dispersion model FLEXPART (Stohl et al., 2005) was
used to determine the source region of air masses that were measured over
Alert. This model has previously been shown to be an effective tool for the
prediction of transport pathways into and within the Arctic (e.g., Paris et
al., 2009). The simulations were driven using meteorological analysis data
from the European Centre for Medium-Range Weather Forecasts with a horizontal
grid spacing of 0.25
Each sample for this study was collected fresh after a known time and over a known area. Given that the snow tables were exposed to the ambient atmosphere for the entirety of each collection period, the measured deposition is considered to represent the total deposition (wet and dry) for said period; however, it is known that surrogate surfaces do not provide an exact proxy for the deposition, which would be seen to a natural snow surface (Ibrahim et al., 1983; Davidson et al., 1985a; Hicks, 1986). There are two additional caveats to this assumption. Firstly, dry deposition at the beginning of each period would fall directly on the exposed clean table rather than onto previously deposited snow. It is unknown what impact these different surface characteristics could have had on the initial deposition rate and collection efficiency. Thus, there is additional uncertainty in the capture of initial dry deposition to the bare table. Secondly, strong winds can disturb and redistribute the snowpack and cause snow to be blown off and/or onto the snow table. Alert operators recorded four occasions when the snowpack was observed to be resuspended due to high winds and these were excluded from the presented results. The dates of these blowing snow events are noted in Table S1 as are missed collections.
The observed snow mixing ratios and fluxes are summarized in Table 1 and Fig. 1 for measured analytes with a strong S/N. Mixing ratio is reported as parts per billion by mass (ppb) with the exception of pH. Flux is reported on a per day basis to take into account the differing collection period lengths; however, it should be noted that this length corresponds to the entire collection period (i.e., the number of days between clearing the snow table), not just the length of time when snow was actually falling. A full record of the measured deposition over the campaign is provided in the Supplement (Tables S1–S6) along with the associated uncertainties and notes of atypical collection conditions. It should be noted that although IC measurements are provided as the measured ions throughout the discussion, these analytes may not necessarily exist in the dissociated ionic form in the environment. Also, the metal measurements provided in Table 1 are total values, insoluble and soluble. The soluble fractions differed by analyte and by date and are provided in the Supplement (Tables S4–S6). The metal measurements can be roughly classified into three categories: predominantly insoluble analytes Fe and Al (> 50 % insoluble over full campaign); variably soluble/insoluble analytes Co, V, As, Cu, Pb, Mn, K, and Mg; and predominantly soluble analytes Ca and Na (< 50 % insoluble) (in order from least to greatest average soluble fraction), excluding analytes with insufficient soluble or insoluble measurements above MDL.
Measured snow mixing ratio (line) and uncertainty (shaded area) of key analytes during 2014 to 2015 campaign.
A review of existing Arctic snow measurements found the measured median
mixing ratios to fall within expected ranges (see Table S5 for details);
however, it should be noted that the referred data represent a variety of
collection and analysis techniques. In general, measurements of this campaign
showed salt species and non-crustal metals to be at the lower end of the
typical range while SO
As discussed above, the collected snow samples provide information on the
total deposition of material to the surface over a given time and area. In
order to elucidate the mechanisms controlling this bulk deposition, a
simplistic model for flux, Eq. (1), was adopted to describe the measured
deposition:
The measured snow flux (
Effective deposition velocities were calculated for chemical species
measured in both snow (SP2 and IC) and atmospheric (SP2 and Hi-Vol) samples.
Figure 2 shows effective deposition velocities calculated as the ratio of
total summed snow flux and average atmospheric concentration measured over
the same period. Both a 6-day resolution, as per the Hi-Vol sampling
frequency, and monthly resolution are provided. The calculated effective
deposition velocities ranged from 0.001 to 10 cm s
Effective deposition velocity at monthly (solid) and approximately 6-day (dashed) frequencies with 6-day uncertainties (shaded area). Missing values indicate periods with snow and/or atmospheric measurements below detection limits.
Monthly effective deposition velocities were used to contrast deposition mechanisms by aerosol composition. A monthly resolution provides insight into the general deposition regime of each analyte, highlighting the impact of bulk deposition characteristics rather than event-specific variability. The variability between aerosol of different composition and the influences of seasonal changes within the Arctic system are simpler to identify without the interference of variability across event-specific conditions. A monthly analysis also facilitates future comparison with modelled results which may not replicate individual events. January and February 2015, were excluded from the monthly analysis because blizzard and high wind conditions were believed to have caused significant losses of snow from the snow tables during these months (based on operator reports), which would lead to underestimation of these snow flux values. The effective deposition velocity is best suited to analysis across periods of equal length and precipitation volume, since both of these parameters are inherently included when the wet deposition efficiency is converted to an equivalent deposition velocity. With the exception of January and February, the total monthly snow precipitation over the campaign was relatively constant, with a relative standard deviation of 20 %.
Figure 3 shows that the typical deposition characteristics varied by analyte,
with median effective deposition velocities ranging from 0.03 to
1.1 cm s
Monthly effective deposition velocities by composition (points, excluding January and February). The median of each analyte (bar) and full range with uncertainty (error bar) are also shown. Also shown is the typical range of dry deposition velocity for accumulation-mode particles to snow by others (Davidson et al., 1987; Petroff and Zhang, 2010).
The variability observed across analytes may be the result of variations in
aerosol properties. First, the measured chemical species differ in terms of
dominant phase: BC, ammonium (NH
Seasonal variation in precipitation
Particle nucleation affinity may also be a significant contributor to the
observed differences in bulk deposition. The lowest velocities were observed
for BC, NH
Furthermore, salt and crustal particles may experience enhanced deposition
since they typically consist of coarser particles than BC, SO
Shared temporal trends in effective deposition velocity can be observed in
Fig. 2. A general trend of heightened deposition in the fall and spring can
be observed across all analytes. In particular, BC, Na
Several factors controlling deposition experience seasonal variations, which may have contributed to the observed inter-monthly variability. Six properties of the Arctic system with seasonal trends were considered as possible influences on the observed velocity trend: precipitation, temperature, mixing height, cloud height, dominant aerosol source region, and sunlight availability, as shown in Fig. 4. The precipitated snow-water equivalent depth was calculated from the snow mass and table area of each sample. Temperature was monitored at local ground-level meteorological stations over the campaign (Table S9) and sunlight estimated from location and time of year. The dominant aerosol source of each month was described using the southern limit to transport and mixing/cloud heights were estimated from radiosonde data, as described above.
Temperature, transport, and sunlight can be seen to follow similar seasonal
trends with fall/spring peaks. Precipitation, mixing height, and cloud height
exhibit episodic peaks and a less significant seasonal trend (intra-monthly relative
standard deviation was a factor of 1.5 to 2 times higher than
inter-monthly, whereas these values were approximately equal for temperature,
transport, and sunlight). When compared to effective deposition velocities,
BC, SO
The effective deposition velocities for warmer and colder months were
separated using a
Effective deposition velocities split by season. The effective deposition velocities are separated into two time periods: warmer months, S/O/My (September, October, and May), and colder months, N/D/Mr/A (November, December, March, and April).
With the exception of C
The unexpected discrepancy in the deposition of Ca
Thus, the observed effective deposition velocities suggest that most analytes
and particularly those expected to exist primarily as particle phase were
preferentially scavenged during the warmer S/O/My months, possibly due to the
presence of mixed-phase clouds and the associated CCN activation of these
chemical species or enhanced below-cloud deposition of those compounds
typically associated with larger particles. However, the change in source
profile typically experienced during these months along with other seasonal
changes in aerosol processing and altitudinal distribution might have also
contributed to the observed S/O/My enhancement. In particular, records of
volcanic activity show that the Icelandic volcano Bárðarbunga was
active August 2014 through February 2015 (Global Volcanism Program, retrieved
March 2016 from
To help characterize the chemical state of the rapidly changing high Arctic,
an intensive campaign of fresh snow sampling at Alert, Nunavut, was
completed and snow quantified for a broad suite of analytes. Comparison of
these snow measurements with coincident atmospheric measurements allowed
estimation of monthly effective deposition velocities describing the total
dry and wet deposition in the range of about 0.02 to 8 cm s
All data is provided in the Supplement.
Organization of the snow collection campaign was led by Sangeeta Sharma with the assistance of Andrew Platt and sample collection by Mike Elsasser. Snow SP2 analysis was completed by Joseph R. McConnell and Nathan Chellman, snow IC analysis was led by Desiree Toom with the assistance of Alina Chivulescu, and snow ICP-MS was led by Katrina M. Macdonald with the assistance of Ying Duan Lei. Analysis of radiosonde data was completed by David Tarasick. Ambient atmospheric monitoring of inorganic aerosols was completed by Desiree Toom and monitoring of BC by Sarah Hanna with the assistance of Allan K. Bertram. FLEXPART simulations were completed by Heiko Bozem and Daniel Kunkel with data analysis assisted by K. Macdonald. Data interpretation was led by Katrina M. Macdonald with input and comments by all authors. Greg J. Evans and Jonathan P. D. Abbatt provided oversight for the project, including input on the manuscript.
The authors declare that they have no conflict of interest.
Funding of this study was provided as part of the Network on Climate and Aerosols Research (NETCARE), Natural Science and Engineering Research Council of Canada (NSERC), the government of Ontario through the Ontario Graduate Scholarship (OGS), and Environment and Climate Change Canada. This project would not have been possible without the collaboration of many skilled individuals: Richard Leaitch at Environment Canada and Catherine Philips-Smith and Cheol-Heon Jeong at the University of Toronto. Edited by: D. J. Cziczo Reviewed by: two anonymous referees