The incidence of wildfires in the Arctic and subarctic is increasing; in
boreal North America, for example, the burned area is expected to increase
by 200–300 % over the next 50–100 years, which previous studies suggest
could have a large effect on cloud microphysics, lifetime, albedo, and
precipitation. However, the interactions between smoke particles and clouds
remain poorly quantified due to confounding meteorological influences and
remote sensing limitations. Here, we use data from several aircraft
campaigns in the Arctic and subarctic to explore cloud microphysics in
liquid-phase clouds influenced by biomass burning. Median cloud droplet
radii in smoky clouds were
The incidence of wildfires in the Arctic and subarctic is increasing dramatically (Flannigan et al., 2009; Moritz et al., 2012; Stocks et al., 1998), and in some areas, such as boreal North America, it is expected to grow by 200–300 % over the next 50–100 years (Balshi et al., 2009). Already, periods of intense wildfires can increase regional aerosol concentrations in the Arctic twofold (Warneke et al., 2010), and the impact of smoke is increasingly being recognized as a strong contributor to Arctic haze (Hegg et al., 2009, 2010; McConnell et al., 2007; Shaw, 1995; Stohl et al., 2006, 2007). Increases in biomass burning aerosols could have a large effect on cloud dynamics (Earle et al., 2011; Jouan et al., 2012; Lance et al., 2011; Lindsey and Fromm, 2008; Rosenfeld et al., 2007; Tietze et al., 2011); in turn, smoke-derived changes to cloud microphysics may result in changes to precipitation and regional heating that are strong enough to affect dwindling regional sea ice (Kay et al., 2008; Kay and Gettelman, 2009; Lubin and Vogelmann, 2006; Vavrus et al., 2010).
However, the interactions between smoke particles and Arctic clouds are
poorly quantified, in part due to the confounding effects of meteorology and
surface conditions (e.g., Earle et al., 2011; Jackson et al., 2012; Jouan
et al., 2012), and in part due to satellite sampling constraints over the
Arctic, such as caused by the presence of many low contrast regions,
multi-layer clouds (Intrieri et al., 2002), and reduced sunlight. One common
way in which aerosol–cloud interactions (ACIs) are quantified is by assessing
how a cloud property changes relative to some aerosol tracer or, in this
case, biomass burning aerosol tracer (BB
The ACI term as defined by Eq. (1) was originally described as the “indirect effect” (IE) (Feingold et al., 2001, 2003). Here, similarly to McComiskey et al. (2009), we use “ACI” instead of “IE” to differentiate the fact that the metric in Eq. (1) is more directly associated with aerosol-driven changes to cloud microphysical responses than with radiative forcing.
The maximum value of ACIs as derived from Eq. (1) is 0.33. An ACI value of
0.33 corresponds with the 1.0 maximum possible change in ln
One study convincingly demonstrated that smoke reduces cloud droplet
effective radius and enhances cloud albedo in Arctic liquid clouds (Tietze
et al., 2011). In that study, modeled BB
However, despite being able to conclusively demonstrate a smoke cloud albedo effect, Tietze et al. (2011) noted that they might have underestimated the magnitude of satellite-derived ACI values because of difficulties constraining aerosol concentrations and locations. They cite a study by Costantino and Breón (2010), where it was demonstrated that not co-locating aerosol–cloud layers in the vertical column dramatically lowered ACI estimates from 0.24 to 0.04 over marine stratocumulus clouds influenced by African biomass burning. This bias seems to be apparent in many ACI estimates globally; from a literature search, McComiskey and Feingold (2012) revealed that remote-sensing-derived ACI values worldwide are lower than those derived from in situ, modeling, and/or ground-based studies. They also showed that in addition to errors in the co-location of clouds and aerosols, the comparatively low spatial resolution of remote sensing observations can further enhance the low bias in ACI estimates.
In the Arctic, these biases can be substantial. In a study in northern
Finland, ACI estimates derived over the same general time period and
location from both ground-based and remote sensing methods were
To better understand the impacts that expected increases in smoke will have on the Arctic, it is important to better constrain remote sensing and model estimates of smoke-specific ACIs in the Arctic using in situ aircraft data. The biggest challenge in obtaining representative aircraft-based ACI values is the fact that they are more prone to uncertainties caused by the influences of poorly constrained meteorological factors (Shao and Liu, 2006) than other methods due to logistical limitations in sample size. We confront this issue in two ways. First, we focus on a case study day from the Arctic Research of the Composition of the Troposphere from Aircraft and Satellites (ARCTAS) campaign (Fuelberg et al., 2010; Jacob et al., 2010) in which several clouds were sampled under very similar conditions. We derive ACI estimates for all clouds that were either verifiably clean or are clearly influenced by biomass burning aerosols, and contrast the observed cloud properties. Second, to increase sample size, we consolidated data from four separate aircraft campaigns in the Arctic. In addition to ARCTAS, these data sets include: the First ISCCP (International Satellite Cloud Climatology Project) Regional Experiment Arctic Clouds Experiment (FIRE.ACE), which included portions flown by the University of Washington Convair-580 (UW FIRE.ACE) and the Canadian National Research Council Convair-580 (NRC FIRE.ACE) (Curry et al., 2000), and the Indirect and Semi-Direct Aerosol Campaign (ISDAC) (McFarquhar et al., 2011). We then compare these findings with those from the ARCTAS case study.
Sampling locations for the following campaigns: ARCTAS (light orange), NRC FIRE.ACE (dark orange), UW FIRE.ACE (dark blue), and ISDAC (light blue). The locations of clouds sampled are shown in Fig. 4.
The dates and flight locations of data used in this study are shown in Fig. 1, and the data used are listed in Tables 1–4. The ARCTAS, FIRE.ACE, and ISDAC data sets have each been extensively described previously (e.g., Curry et al., 2000; Fuelberg et al., 2010; Jacob et al., 2010; Korolev et al., 2003; McFarquhar et al., 2011; Rangno and Hobbs, 2001; Soja et al., 2008). However, to our knowledge, they have never been compared directly to each other. Here we note only briefly a few relevant points about the data sets and how they are inter-compared.
Instrumentation used in this study from the ARCTAS data set. Data were collected at 1 s resolution, unless noted otherwise.
Instrumentation used in this study from the ISDAC data set. Data were collected at 1 s resolution.
* For days when high-quality CDP data were unavailable, following Earle et al. (2011).
Instrumentation used in this study from the NRC FIRE.ACE data set. Data were collected at 1 s resolution.
Instrumentation used in this study from the UW FIRE.ACE data set. Data were collected at 1 s resolution.
First, during the ISDAC and FIRE.ACE flights, multiple passes inside clouds
were often obtained, and aerosols were intentionally sampled above- and
below-cloud. In contrast, during ARCTAS there was very limited resampling of
a given region and generally only one pass through a cloud was obtained.
This difference in sampling impacts our results only in that there are not
as many vertical profiles through the ARCTAS clouds as in the other
data sets. Second, the UW FIRE.ACE data set contains some gaps in positional
data (latitude, longitude, and altitude), which range most frequently
between 1 and 10 s, with rare instances of gaps
In ARCTAS, cloud liquid water content (LWC) was determined from droplet size
spectra gathered with the CAPS-CAS instrument (Baumgardner et al., 2001)
based on integrated volume droplet size distributions between
0.75 and 50
During the UW and NRC FIRE.ACE campaigns, LWC was determined from droplet
size spectra gathered from the Forward Scattering Spectrometer Probe (FSSP-100)
measurements for particles with diameters between 0.5 and 47 and 5 and 47
Comparison of LWC measurements (g m
During ISDAC, LWC was determined from cloud droplet probe (CDP) data. These data agreed within 15 % of the bulk probe values. Following Earle et al. (2011), FSSP data were used on days when high-quality CDP data were unavailable; the FSSP data are estimated to agree with CDP data to within 20 %. Note that similarly to ice particles (e.g., Korolev et al., 2011), very large droplets may shatter on any of the cloud droplet probe tips. This may introduce some potential artifacts when droplet sizes are very large (e.g., for some of the reference measurements available in FIRE.ACE and ISDAC).
For comparability with ARCTAS clouds, the presence of liquid clouds in the
FIRE.ACE and ISDAC data sets was determined by simultaneous measurements of
LWC
There is no consistent definition for cloud phase in the literature. In
remote sensing studies for example, cloud phase is usually determined by
cloud radiative properties – thus, clouds with some mixed particles can be
included in liquid- or ice-phase classifications if they are mostly
liquid or mostly ice (e.g., Baum et al., 2012; Platnick et al., 2003). Due
to instrumentation limitations, aircraft studies sometimes also define a
cloud with small fractions of ice particles as being a “liquid” cloud
(e.g., Korolev et al., 2003). Alternatively, distinct portions of a cloud
may be classified as different phases if a primarily liquid portion of a
cloud is far away (
Here, we define liquid cloud phase by the lack of any ice particles in the CPI data throughout the entire cloud transect, based on a roundness criterion (Lawson et al., 2001). When possible (i.e., in the NRC FIRE.ACE and ISDAC data sets), we verified that there was no detectable ice water along the cloud transects. This relatively stringent definition of liquid phase clouds is used to describe as best as possible the liquid-phase end-member cloud characteristics. Because aircraft cloud transects can only sample a portion of a cloud, we must assume that the portion of the cloud sampled is representative of the rest of the cloud. This may introduce uncertainties, particularly in persistent large-scale stratus clouds. Nonetheless, as discussed in Sect. 3.1, we believe that errors from this assumption are not likely to have a large impact on our results.
We used aircraft vertical profiles to assess cloud droplet effective radius
(
In some instances in the multiple-campaign analysis, the same cloud or very
similar clouds were sampled more than once, often intentionally, either
through an entire vertical cloud transect or through a portion of a cloud.
In order to reduce the potential for pseudo-replication in the analysis,
transects that were deemed to be from the same cloud or from very similar
clouds were averaged to provide one aggregated profile or
LWC among aggregated clouds was generally similar (within 30 % of each
other). However, in some cases it was more variable; in one biomass burning
aggregation, the set of eight related cloud transects had LWCs ranging from
0.12 to 0.54 g m
For this work, distinguishing smoke-influenced conditions from background cloud
conditions is critical. During ARCTAS, background conditions were selected
by a combination of in-cloud gas concentrations (average CO
The 123 ppbv CO cutoff value represents the upper quartile range of time
periods with concurrently low CO, CH
A comparison of background concentrations of biomass burning and
pollution tracers as previously reported to those in the ARCTAS-B data set in
air masses that would be defined as background using only the CN
ARCTAS biomass-burning-influenced air masses were classified following
the procedure of Lathem et al. (2013), where BB-influenced air masses have
concentrations of
During the two FIRE.ACE campaigns, the combination of relevant high-quality and/or high-resolution aircraft chemical data for completely characterizing air mass sources was not collected, and remote sensing products useful for air mass classification were also unavailable. As a result, biomass-burning-derived haze events were indistinguishable from anthropogenic pollution events in the FIRE.ACE data sets. Therefore, we only use FIRE.ACE clouds sampled under unpolluted background conditions for inter-comparison with the other data sets.
Because within-cloud gas concentrations were not available, we used average
near-cloud (as defined above) aerosol concentrations to define
background conditions in the FIRE.ACE data. To reduce the risk of any
potential humidification effects, we excluded near-cloud air masses that had
any observations of cloud particles in the CPI or that had LWC values
To classify background air masses, we used the Passive Cavity Aerosol
Spectrometer Probe (PCASP) aerosol concentrations (CN
To be classified as background, air masses had to have CN
A biomass burning classification was assigned in ISDAC data when a cloud had contact with discernable amounts of biomass burning aerosols, as determined by single particle mass spectrometer, SPLAT II (Zelenyuk et al. 2009, 2015), based on the mass spectral analysis of individual aerosol particles (Fig. 2). This method has been similarly employed to determine biomass burning influence in the ISDAC data set previously (Earle et al., 2011; McFarquhar et al., 2011; Shantz et al., 2014).
ISDAC 2008 aerosol and flight characteristics near and in selected
clouds influenced by biomass burning from 19 April (left) and 20 April
(right). Flight characteristics shown include:
As mentioned before, the impact of smoke aerosols on cloud droplet
activation was assessed by looking at aerosol–cloud interactions of
biomass burning aerosols on cloud droplet number. The ACI values were
derived from changes in cloud droplet number relative to measured biomass
burning tracers, BB
Carbon monoxide (ppbv) during the 1 July 2008 ARCTAS-B flight as a
function of
As previously mentioned, ARCTAS was the only campaign where biomass burning
gaseous tracers were directly quantifiable in-cloud (here we use BB
Because the in-cloud CO and CH
There are some limitations of the ACI approach. First, a systematic bias can
be introduced when aerosol and cloud properties are averaged or co-located
in low spatial or temporal resolution data sets (McComiskey and Feingold,
2012). This particular systematic bias is generally not a large concern for
in-cloud aircraft studies such as this one where gas and/or aerosol
measurements and
A third potential problem is the risk that a snapshot of a cloud in time is not representative of the net cloud properties over its lifetime (Duong et al., 2011). Currently, only models can fully characterize cloud lifetime properties, but interpreting the model output can be challenging for other reasons. Within aircraft in situ data, this source of sampling error is best minimized in aircraft in situ data by resampling throughout the cloud's life cycle. Resampling was sometimes, but not always, carried out for individual cloud cases presented here, and was not specifically carried out throughout the lifetime of the cloud. However, based on the results presented in Duong et al. (2011), the magnitude of this type of error is unlikely to have a large impact on our results, although we cannot with full confidence assess how cloud life stage might have impacted the way aerosols were interacting with the clouds.
The fourth limitation of the ACI method is that
Ambient conditions such as cloud type and presence of drizzle from an
overlying cloud deck were determined from available video, photos, flight
notes, and AVHRR images. Although in situ chemical and physical measurements
were primarily used to determine end-member situations (i.e., where only
smoke or only background air were the dominant sources of aerosols
interacting with clouds), in some cases we discuss out-of-cloud aerosols
with potentially more mixed sources. In these cases we supplemented chemical
and physical data with 5-day HYSPLIT back trajectories (Draxler, R. R. and
Rolph, G. D. HYSPLIT (HYbrid Single-Particle Lagrangian Integrated
Trajectory) model, accessed via the NOAA ARL READY website
(
Map of cloud sample locations from all campaigns. Red points indicate biomass burning samples, blue cases indicate background samples, and gray points indicate intermediate samples.
Across all clouds sampled during the four campaigns, there was substantial variation between cloud properties (Table 7) and the physical locations of the clouds (Fig. 4). For example, background clouds were primarily sampled over the open ocean and at higher latitudes, whereas the smoky clouds were primarily sampled at lower latitudes over land. For this reason, in addition to comparing median characteristics of all background and clean cases, we also focus on a case study where multiple clean and smoky clouds were observed under very similar meteorological and surface conditions (Sect. 3.1).
Median properties and ranges for all background and biomass burning cloud cases in the multi-campaign assessment.
On 1 July 2008 during the ARCTAS-B campaign, a variety of small cumuliform
clouds were sampled during flight 18 over inland Saskatchewan, Canada. The
physical characteristics of the clouds were very similar (Table 8), being
small (
Mean properties and ranges for the 1 July 2008 ARCTAS case study, including background, intermediate, and biomass burning cloud cases.
Despite being exposed to similar meteorological and surface conditions,
aerosol inputs to these clouds ranged significantly, with average CH
In Fig. 3, we show that CO
Based on seven samples from the ARCTAS-B 1 July 2008 case study,
here we show the relationships between ln(
Mean cloud droplet size distributions (
As shown in Fig. 5, smoke is clearly correlated with reduced cloud droplet radius in the seven clouds studied (with an average 59 % reduction relative to background clouds, Table 8). As expected, there was a concurrent increase in cloud droplet number (Fig. 5). Based on this increase, we compute a combined median ACI of 0.05 (bootstrapped 95 % confidence interval 0.04–0.06) across all tracers shown in Fig. 5.
Although linear regressions were not used to derive ACIs, we plot them for
each tracer in Fig. 5 to show the degree of variation between individual
tracer ACI values. Other researchers have previously noted differences in
calculated ACIs when these interactions are computed from different tracers
(e.g., McComiskey et al., 2009; Lihavainen et al., 2010; Zhao et al.,
2012), and these differences probably reflect a combination of measurement
error and how well a given tracer approximates the sub-population of
aerosols that are participating in cloud droplet activation (Lihavainen et
al., 2010). As plumes age, there may also be increasing uncertainty in
biomass burning aerosol co-location with gaseous tracers such as CO and
CH
ACI estimates can also sometimes be influenced or even overwhelmed by systematic differences in local meteorological conditions associated with cleaner versus more polluted clouds (Hegg et al., 2007; Shao and Liu, 2006). For the case study, that possibility is unlikely because of the relatively small area and time frame considered and the similar meteorological conditions in which the clouds were sampled.
However, because case study smoky clouds had a combination of very low LWC, very high aerosol concentrations from a fresh fire, and consequently, very small droplet sizes (Fig. 6), it is likely that smoky case study clouds were less sensitive to further additions of smoke aerosols than clouds with lower aerosol concentrations. Such nonlinear behavior is predicted when high CCN levels cause increased competition for water vapor, which in turn decreases cloud supersaturation and reduces the tendency to form additional drops (e.g., Moore et al., 2013; Morales et al., 2011; Morales and Nenes, 2010). Additionally, possible enhanced entrainment of outside air in smoky clouds compared to background clouds (Ackerman et al., 2004; Bretherton et al., 2007; Chen et al., 2012; Lebsock et al., 2008) could enhance droplet evaporation and further reduce ACI values from the expected adiabatic ACI maximum value at a given aerosol level.
Because in situ ACI derivations assume linearity in the response of
For these reasons, although the 1 July 2008 case is in some ways ideal, in
that the clouds were sampled in very similar environmental conditions, it is
not necessarily representative of typical cloud conditions in the Arctic.
The clouds were present relatively far south in the subarctic (52–56
To assess the impact of smoke on liquid clouds more generally, we compared background and biomass burning cloud properties sampled over the larger region shown in Fig. 4. This more expansive set of clouds includes a broader range of high-latitude meteorological conditions, making it more representative of overall conditions in the Arctic region. However, the greater heterogeneity also makes trends in the data more difficult to interpret, as we cannot describe in full detail the degree to which meteorological influences affected each cloud given the limitations of the data sets.
Same as in Fig. 5, but for data from the multi-campaign analysis.
As in Fig. 5, CO* indicates that background values of 99.2 ppbv have been
subtracted. For CH
Despite the uncertain meteorological influence, we see qualitatively similar
trends to those in the 1 July 2008 ARCTAS case study (Fig. 7). We find a 3.7
Observed smoke-driven reductions in liquid cloud droplet size and increases
in cloud droplet number in both the case study and the multi-campaign
analysis are in line with several other studies in the Arctic. Peng et al. (2002) found a similar difference in
As noted previously, because the aircraft could only sample transects of
clouds, we had to assume that the observed cloud phase was representative of
the whole cloud. In the case study, all clouds were sampled at temperatures
Based on model output by McComiskey and Feingold (2008) (their Fig. 2a), we
estimate that given the case study median ACI value of 0.05, the
smoke-derived cloud albedo effect on summertime local short-wave radiative
forcing could be between
In contrast to the subarctic, in the Arctic, high surface albedo will lessen the expected impact of the cloud albedo effect. Although future sea ice losses and associated reductions in surface albedo may affect the relative importance of the cloud albedo effect on Arctic clouds, others (e.g., Garret et al., 2004) have suggested that in the Arctic, a more important impact of reduced cloud droplet size may be greater long-wave opacity, which can lead to enhanced snowmelt. Relatedly, smaller droplets may affect cloud lifetime either by extending it via reduced precipitation (the “second indirect effect”; Ackerman et al., 2000; Albrecht, 1989) or by reducing it via enhanced water vapor competition and evaporation, as may have occurred in the case study.
Cloud droplet spectra from the 1 July 2008 ARCTAS case study clouds are
shown in Fig. 6. Although sample size is small, the presence of smoke
appears to narrow the droplet spectra from a dispersion of 0.84 in
background clouds to 0.55 in smoky clouds, as calculated by the ratio
between the standard deviation of the size distribution and the mean droplet
radius. This narrowing is likely to lessen the eventual probability of
precipitation (Tao et al., 2012), as it moves median droplet size further
away from the 28
Mean cloud particle size distributions (
Cloud droplet spectra from the multi-campaign clouds are shown for
comparison in Fig. 8. There is not as obvious a narrowing of spectra as for
the case study, but median droplet concentrations in smoky clouds never
reached above 28
As mentioned previously, large numbers of nucleation- and Aitken-mode particles are frequently observed in the spring and summer Arctic and subarctic (Engvall et al., 2008; Leck and Bigg, 1999; Ström et al., 2009; Zhao and Garrett, 2015). These particles are thought to have a marine origin via some combination of new particle formation from marine gases (Allan et al., 2015; Leaitch et al., 2013; O'Dowd et al., 2010; Tunved et al., 2013) and direct oceanic nanogel emissions (Heintzenberg et al., 2006; Karl et al., 2012, 2013; Leck and Bigg, 1999; Orellana et al., 2011). Chemical data from the ARCTAS data set also show the presence of numerous small particles with a natural background source (Fig. 9).
Log relationships between ARCTAS-B and ISDAC aerosol number
concentration and submicron scatter. In panels
Previous studies also suggest that the small particles can condense upon
larger particles (e.g., smoke) when such particles are present
(Engvall et al.,
2008; Leaitch et al., 2013; Tunved et al., 2013). This coagulation process
may explain why Arctic smoke aerosols have been shown to sometimes contain
organic components likely derived from smaller, non-biomass-burning
particles mixed with sulfates and marine particles (Earle et al., 2011;
Zelenyuk et al., 2010). To get some idea of how important the background
particles may be, we estimated the maximum mean aerosol volume change that
would occur if high concentrations of small background aerosols were to mix
with and condense upon diluted smoke particles. Concentrations of background
particles were estimated at 5000 cm
Interestingly, the small Arctic marine particles appear to be fairly
hygroscopic
(Lathem
et al., 2013; Lawler et al., 2014; Zhou et al., 2001), and they can be
surface-active (Lohmann and Leck, 2005). One study using ARCTAS data showed
that background aerosol values of the hygroscopicity parameter,
However, the nucleation- and Aitken-mode background particles are not ubiquitous throughout the year. They tend to accumulate mainly in the spring and summer, which is thought to be due to a combination of three factors: (1) there is more sunlight available for the photochemical reactions key to new particle formation (Engvall et al., 2008; Tunved et al., 2013), (2) reduced sea ice and enhanced primary production likely lead to greater emissions of marine precursor gases and nanogels (Leaitch et al., 2013; O'Dowd et al., 2010; Tunved et al., 2013), and (3) during Arctic summer there tend to be fewer larger particles such as smoke for these small particles to coagulate and condense upon. However, Arctic summertime smoke events do occur (e.g., Fuelberg et al., 2010; Iziomon et al., 2006) and may be increasing (Moritz et al., 2012). In the subarctic, wildfires peak in the summer (Giglio et al., 2006). Thus, although the influence of the small background particles on subarctic and Arctic smoke ACI values is probably minor, deviations from the linear ACI expectations derived here might occur during dilute summertime Arctic smoke events and in subarctic locations, for example when smoke is diluted over or near marine environments.
The challenge of separating the influence of meteorology and aerosol indirect
effects on clouds introduces relatively large uncertainty in our
understanding of how smoke impacts clouds. Using in situ aircraft data, we
quantified these impacts in both a subarctic cumulus cloud case study and in
a multi-campaign data assessment of clouds north of 50
Based on a previous model study by McComiskey and Feingold (2008), the ACI value of
0.05 from the case study suggests that smoke may reduce local summertime
radiative flux via the cloud albedo effect by between 2 and 4 W m
Smaller cloud droplets can have various consequences. Smoke-driven reductions or delays in precipitation may affect the distribution of aerosol and moisture deposition. Longer cloud lifetime could impact not only Arctic albedo but also long-wave radiation (Stone, 1997), and previous studies suggest that even small changes in the above parameters may affect sensitive Arctic sea ice (Kay et al., 2008; Kay and Gettelman, 2009; Lubin and Vogelmann, 2006; Vavrus et al., 2010). Additionally, changes in cloud cover might also have indirect effects on ocean photosynthesis and biogeochemistry (Bélanger et al., 2013). It is our hope that the improved quantification of smoke-derived ACI values will help quantify these impacts in future model studies.
One obvious limitation of our study is that we do not address the impacts of smoke on existing mixed- and ice-phase clouds. Additionally, we cannot account for the ways in which smoke might have affected the sample phase. For example, ice nuclei presence might facilitate the conversion of an otherwise liquid-phase cloud into a mixed-phase cloud that was excluded in this assessment. Alternatively, we could have included liquid clouds in our assessment that might otherwise have been present as mixed- or ice-phase clouds if not for the inhibition of freezing by soluble smoke compounds via the Raoult effect (discussed in Tao et al., 2012).
Finally, we have presented evidence to suggest that coagulation of the numerous nucleation- and Aitken-mode background particles frequently present in clean summertime Arctic air masses might increase the volume of diluted smoke aerosols by up to 2–15 %. Previous studies suggest that such interactions with background particles may increase smoke aerosol hygroscopicity, which in turn could cause deviations from the ACI value derived here. Future remote sensing or ground-based analyses may be able to more completely address the different impacts of dilute vs. concentrated smoke aerosols in Arctic clouds.
We first estimate the volume of smoke particles at dilute concentrations of
450 particles cm
The degree to which aerosol properties can be affected by the collection of
Arctic nucleation- and Aitken-mode background particles onto larger smoke and
pollution particles also depends in part on the size ranges and
concentrations of the background particles. These can be quite variable
(Engvall et al., 2008) (also see Fig. A1). To estimate average background
concentrations, we use the observed geometric mean ratio range in 6-year
Svalbard summertime data (Engvall et al., 2008), which indicated that
Aitken-mode particle concentrations were
Mean out-of-cloud aerosol particle size distributions for several
ARCTAS background aerosol events. Some days had multiple background aerosol
events; these are distinguished by color and the letters
Alternatively, we can estimate what the background aerosol volume might be if
particle concentrations were as high as 5000 cm
The authors would like to thank A. Aknan, B. Anderson, E. Apel, G. Chen, M.
Couture, T. Garrett, K. B. Huebert, A. Khain, A. Korolev, T. Lathem, P.
Lawson, R. Leaitch, J. Limbacher, J. Nelson, M. Pinsky, W. Ridgeway, A.
Rangno, S. Williams, S. Woods, and Y. Yang for data and/or advice or help
with various aspects of this project, and all others who were involved in
collecting and funding the collection of the data sets we have used. We
acknowledge the Atmospheric Radiation Measurement (ARM) Program sponsored by
the U.S. Department of Energy, Office of Science, Office of Biological and
Environmental Research, Climate and Environmental Sciences Division for
providing the ISDAC data set. The authors also gratefully acknowledge the NOAA
Air Resources Laboratory (ARL) for the provision of the HYSPLIT transport and
dispersion model and/or READY website (