Spatial and temporal variability of aerosol particles in Arctic spring



The objective of this work is to investigate the variability in the aerosol particle number concentration in Arctic spring. The Indirect and Semi-Direct Aerosol Campaign (ISDAC) was conducted during April 2008 in the vicinities of Fairbanks and Barrow, Alaska. Aircraft-based measurements of total aerosol particle number concentration (Na) in the size range of 0.12–3 µm diameter were obtained using a passive cavity aerosol spectrometer probe (PCASP-100X). The analysis considers Na during cloud-free periods in biomass burning (BB) and non-BB aerosol loading scenarios, the latter including background cases and cases with elevated concentration in layers. The BB cases had air masses originating mainly from Russian and Asian forest and crop fires, whereas the non-BB cases originated predominantly from Arctic or oceanic regions. The average Na for all non-BB cases was 127 cm−3, while that for all BB cases was Na = 720 cm−3. These estimates do not, however, capture the details of aerosol particle layers encountered during most flights. Variability in Na was considered for constant altitude (horizontal) flight legs ranging from 50 to 650 km in length, as well as for vertical flight profiles up to 7 km above sea level. When aerosol particle layers were encountered, Na rapidly increased from 20 to 550 cm−3, and reached up to 2200 cm−3 within air masses dominated by BB plumes. The observed variability in Na may have important implications for estimating cloud microphysical properties as well as estimates of particle properties used in global climate model simulations, because averaging over large space- or time-scales may not represent real atmospheric conditions. The analysis demonstrates the difficulty in interpreting average aerosol particle characteristics along longer flight legs, particularly during cases with higher particle loading that varies over shorter distance scales and time periods. Copyright © 2012 Royal Meteorological Society and Crown in the right of Canada.

1. Introduction

Aerosol particles influence the radiation budget of the atmosphere directly, by scattering or absorbing solar radiation, or indirectly, by acting as cloud condensation nuclei (CCN) or ice nuclei (IN) that alter cloud microphysical properties, such as cloud liquid water path and effective size (e.g. Twomey, 1974; Albrecht, 1989; Liou and Ou, 1989; IPCC, 2007). These direct and indirect effects are not well understood, especially in Arctic regions, where the impact of aerosol particles on liquid, mixed-phase and ice clouds is not clear (IPCC, 2007). In the past 100 yr, the Arctic has warmed nearly twice as much as the global average, and future projections suggest that this trend will continue (IPCC, 2007). Accordingly, research interest in Arctic regions has increased in recent years; however, aircraft measurements of aerosol particles are not extensive due to the cost, remoteness and harsh working conditions associated with field campaigns in these regions. For example, in the past few decades there have been several Arctic field campaigns with aerosol particle measurements on-board aircraft (Table 1); however, these measurements are not well-distributed in space and time, and a complete Arctic aerosol particle dataset is still not available (Treffeisen et al., 2005).

Table 1. A summary of total particle number concentration (Na) ranges and mean values from aircraft studies in Arctic spring/summer. Dashes (—) in this table indicate missing data from the literature.
ProjectDateKey ArcticMinimumAltitudeRange inNumberMean NaBackgroundReferences
acronym locationsparticle(m-MSL)Na (cm−3)of(cm−3)clean (c), 
   diameter size (M = maximum)flights polluted (p), 
   range (µm)  for finding Mean Na all (a)* 
  • *

    ‘all’ simply means that polluted conditions were not filtered out of background conditions.

  • It was noted in Brock et al.(1989, 1990) that most of the particles observed were in the accumulation mode size range (i.e. diameters of 0.1–2 µm), which is in the range of the ISDAC PCASP measurements.

  • Determined over vertical profiles

  • §

    Determined over a range of altitudes when horizontal flight legs where flown

  • During ISDAC, ‘c’ is non-BB cases and ‘p’ is BB cases

Not givenAugust 1985Greenland0.005∼200cBrock et al.,
  Canadian  300–6001a1989
  (C.) Arctic  M = 31001p 
AGASP IIMarch–AprilAlaska, Alert,0.170–1500150–2258∼200cLeaitch et al.,
 1986Arctic Ocean 1500–3500100–1758∼150c1989
Not givenMarch–AprilC. Arctic,0.005150–9001150aBrock et al.,
 1986Greenland 100–3001230a1990
LEADEXApril 1992Alaska0.0130–2200134–4288206cHegg et al.,
Not givenApril 1994Siberia,0.004150–3001cDreiling and
  C. Arctic 100–10003aFriederich,
    2000M = 1000A fewp1997
FIRE- ACEApril–JulyAlaska, Arctic0.135Lower alt.0–10006215 ± 264aGultepe and
 1998Ocean Higher alt.0–12006348 ± 235aIsaac, 2002
    Mid-alt.M = 1200A fewp 
ASTAR 2004May 2004Svalbard0.01< 2500273–10246529aEngvall et al.,
    > 2500341–1011 587a2008
    < 2500176–10767436a 
 June 2004  > 2500273–718 443a 
    < 2400132–8616338a 
    > 2400194–817 398a 
ISDACApril 2008Alaska0.1238–70670–66211152 ± 111cFigure 4(a)
    132–70460–26544612 ± 477pFigure 4(b)
    510–6793§ 4–73011127 ± 84cFigure 5(a)
    1744–6913§ 2–22006720 ± 539pFigure 5(b)

A phenomenon known as ‘Arctic haze’ has been observed during Arctic spring (e.g. Brock et al., 1989; Shaw, 1995; Quinn et al., 2007, and references therein). This refers to aerosol particles and gases, largely of anthropogenic or biomass burning (BB) origin, that have been transported to the region and are typically observed as distinct layers in the atmosphere. The removal rate of pollution particles is slowed down in cold polar regions, especially in the springtime (Shaw, 1995), and thus haze layers tend to persist in those regions. Moreover, the atmosphere is stable in the Arctic and layers at different altitudes are not well-mixed and therefore the aerosol number concentrations can vary significantly from one region to the next. The Arctic haze phenomenon is most prevalent in the winter and spring and becomes less frequent as the summer progresses (Quinn et al., 2002). Because aerosol particles can influence the radiation budget and therefore climate, it is important to study particulate matter during the spring, when the solar radiation increases after long dark winter nights and when particle concentrations are higher than other times of the year.

Accurate representations of aerosol properties within general circulation models (GCMs) are needed to improve climate change simulations. General circulation models are used to predict climate trends of parameters such as temperature, precipitation and surface heat fluxes. Aerosol effects on ice particle number concentration (Ni) are a large unknown in GCMs. Aerosol indirect effects on climate via liquid clouds are better understood, but still highly uncertain (IPCC, 2007), and are generally initiated in GCMs in one of two ways: (i) incorporating particle properties such as size distribution and chemical composition into a detailed microphysical parameterization (e.g. Abdul-Razzak et al., 1998) that determines cloud droplet number concentration (Nd) as a function of vertical air velocity and total aerosol particle number concentration (Na) in a rising air parcel; and (ii) using constant bulk parameters, such as Na (from clear cloud-free air) and Nd (from in-cloud measurements), determined from observations (e.g. Jones et al., 1994; Menon et al., 2008). With regard to the latter, Gultepe and Isaac (1999) studied relationships between Na and Nd for five field projects and found Nd to be particularly sensitive to averaging scales smaller than 10 km. They suggested that models would have much larger uncertainties if the averaging scale is not considered, especially in cases where there are breaks in the cloud. Current GCMs use large grid spaces, typically tens (e.g. Girard and Bekcic, 2005) to hundreds of kilometres (e.g. Schmidt et al., 2006). It is therefore important to consider the averaging scales for aerosol particle properties used in parametrizations that define Nd, as the use of inaccurate averaged values could affect the results of climate model simulations.

The analysis presented here is based on aircraft in situ data collected during the US Department of Energy (DOE) Indirect and Semi-Direct Aerosol Campaign (ISDAC) in Alaska during April 2008. The goal of this work is to reduce the uncertainty in climate model simulations by improving understanding of the variability of cloud-free Na measurements that can be used in bulk parametrizations of Nd and Ni (Gultepe and Isaac, 1999). The four main objectives of this work are as follows: (i) to examine Na variations over horizontal flight legs for cases with lower Na (non-BB cases); (ii) to characterize the aerosol concentration variability of BB cases; (iii) to examine Na variations along vertical profiles, as well as the meteorological conditions that may have led to these variations; (iv) to explore relationships between Na variability and averaging scales. These objectives are accomplished using detailed analysis of aircraft in situ observations collected by probes mounted on the National Research Council of Canada (NRC) Convair-580 aircraft.

2. Aircraft in situ instrumentation and analysis

2.1. In situ observations

In situ aircraft measurements were collected in Alaska from 1 to 30 April 2008 during ISDAC. This intensive field campaign utilized airborne measurements from instruments deployed on the Canadian NRC Convair-580 (McFarquhar et al., 2011). Transit flights were conducted from/to Fairbanks, where the aircraft was based due to logistics, and to/from Barrow, where project flights were typically performed in the vicinity of the DOE North Slope of Alaska site. A total of 28 flights were conducted during the campaign, comprising approximately 100 h of flight time.

The horizontal leg flight data used here from 17 of these flights are summarized in Table 2 (non-BB cases) and Table 3 (BB cases). The majority of measurements presented here were made in the free troposphere at altitudes higher than 1000 m above mean sea level (m-MSL) (i.e. above the boundary layer height or inversion layer base), but there may be boundary layer effects for the occasions when measurements were made below 1000 m-MSL, such as aircraft take-off, landing, or missed approaches. Flights occasionally passed over the Arctic Ocean; earlier in the campaign, the ocean surface was predominantly ice-covered, but later in the campaign there were more open-water leads and polynyas. In the case of open water surfaces, aerosols originating from the ocean can affect the measurements within the boundary layer (Gultepe et al., 2003); however, this should not influence the measurements significantly at the altitudes considered here.

Table 2. Summary of aerosol particle data during horizontal, constant altitude flight legs with no biomass burning influence during ISDAC in April 2008. All data with liquid, ice and/or precipitation have been excluded. The total aerosol particle number concentration, Na, is from the PCASP-100X. All numbers indicated are the 10 min averages (calculated with 1 s data, thus there are 600 data points in each average shown here) for that horizontal leg within that time period (for some flights, there are multiple 10 min clear air horizontal legs). Standard deviations (± SD) are provided for Na and RH values.
Time, date (UTC)Flight numberNa (cm−3)RHw (%)RHi (%)T (° C)Altitude (m-MSL)
0135–0155, 2 April10128 ± 2645 ± 363 ± 5−326186
   80 ± 2353 ± 7 73 ± 10−336202
1810–1823, 4 April11 62 ± 2066 ± 387 ± 4−285936
1907–1917, 4 April11 31 ± 1769 ± 598 ± 8−366541
0229–0239, 5 April12 74 ± 35 53 ± 15 75 ± 21−356748
1943–1957, 5 April13 3.5 ± 1.964 ± 262 ± 2 2.4 510
1531–1619, 8 April15 55 ± 49 53 ± 13 78 ± 19−406356
   96 ± 9436 ± 8 53 ± 12−396356
   84 ± 8137 ± 9 54 ± 14−396359
1700–1710, 8 April15223 ± 6883 ± 593 ± 5−111200
0026–0117, 9 April17 61 ± 1236 ± 451 ± 6−366096
   58 ± 1441 ± 7 58 ± 10−366093
   76 ± 6339 ± 7 56 ± 10−366106
   76 ± 1628 ± 540 ± 6−376102
   66 ± 46 46 ± 20 67 ± 29−386103
2041–2051, 26 April30166 ± 1884 ± 393 ± 4−10868
2347–2357, 26 April31130 ± 3455 ± 261 ± 2−101130
2358, 26 April to 0008, 27 April31120 ± 1557 ± 662 ± 6 −91246
0024–0034, 27 April31161 ± 1783 ± 193 ± 1−11 936
0156–0206 , 27 April31171 ± 2085 ± 295 ± 3−12 920
0445–0456, 27 April32254 ± 65 30 ± 11 44 ± 17−376087
2022–2115, 28 April33196 ± 3263 ± 488 ± 5−346586
  185 ± 5256 ± 278 ± 4−336585
  111 ± 2847 ± 365 ± 4−326587
0350–0523, 29 April34250 ± 4930 ± 443 ± 5−376775
  156 ± 3525 ± 136 ± 1−366776
  100 ± 2827 ± 439 ± 6−376774
  183 ± 7127 ± 538 ± 7−376778
  232 ± 69 49 ± 21 70 ± 30−366779
  184 ± 3663 ± 389 ± 4−356779
  141 ± 2742 ± 560 ± 7−366779
  163 ± 4645 ± 265 ± 4−366786
   220 ± 15538 ± 755 ± 9−376793
Table 3. Summary of aerosol particle data during horizontal, constant altitude flight legs with biomass burning influence during ISDAC in April 2008. See Table 2 caption for details.
Time, date (UTC)Flight numberNa (cm−3)RHw (%)RHi (%)T (° C)Altitude (m-MSL)
1541–1552, 18 April22 457 ± 13526 ± 136 ± 2−356913
1744–1754, 18 April22 966 ± 6147 ± 856 ± 9−183998
2039–2049, 18 April23 992 ± 5734 ± 741 ± 8−194109
2100–2120, 18 April23 935 ± 5112 ± 114 ± 2−163909
  1045 ± 82 14 ± 0.516 ± 1−153811
2331–2342, 18 April23 217 ± 165 71 ± 11 99 ± 16−346633
2053–2121, 19 April25 178 ± 3930 ± 742 ± 9−346813
   78 ± 2829 ± 340 ± 4−356817
2205–2215, 19 April25 542 ± 21530 ± 8 39 ± 10−275581
2223–2233, 19 April25 822 ± 14616 ± 820 ± 9−204474
2242–2252, 19 April251181 ± 56221 ± 524 ± 5−122931
2258–2308, 19 April25 865 ± 83940 ± 744 ± 7−102301
0045–0105, 20 April261878 ± 56243 ± 346 ± 3 −71736
   500 ± 46847 ± 451 ± 5 −81744
1741–1812, 24 April27 454 ± 30064 ± 485 ± 5−296703
  864 ± 8263 ± 8 83 ± 10−296705
   872 ± 10456 ± 675 ± 8−296711
2232–2242, 24 April28443 ± 67 77 ± 16 92 ± 20−183933

Measurements used in this work are Na from a Particle Measuring Systems (PMS) passive cavity aerosol spectrometer probe (PCASP-100X; size range 0.12–3 µm geometric diameter over 15 channels), ice crystal number concentration (Ni) from a PMS two-dimensional cloud (2D-C) optical probe (size range 25–800 µm), and Nd from a PMS forward scattering spectrometer probe (FSSP-100; size range of 2–47 µm). Each probe was mounted under the aircraft wing and sampled at 1 Hz. The FSSP-100 measurements were corrected for probe dead time and coincidence errors following the approach of Baumgardner et al.(1985). Wind speed and direction measurements were obtained by a Rosemount 858Y five-hole pressure probe. Measurements of temperature (T) and relative humidity with respect to ice (RHi) and water (RHw) were obtained using a Rosemount T sensor and an EG&G chilled mirror dewpoint temperature (Td) hygrometer. The equivalent potential temperature (θe) was calculated from measurements of T,Td and pressure.

2.2. Analysis

This work involves the study of clear air measurements of Na, which have been used in studies to relate Na to Nd or Ni (e.g. Jones et al., 1994; Menon et al., 2008). These studies neglected the variability of Na in the horizontal and vertical domains, and assumed a single diagnostic relationship in climate simulations. In the present analysis, the aerosol particle measurements consider sizes > 0.12 µm, representing potential IN and CCN. All observations considered here represent cloud-free and precipitation-free conditions. In order to ensure that there were no liquid or mixed-phase clouds present, measurements of Nd from the FSSP-100 were examined; and when Nd exceeded 5 cm−3, the aerosol particle data were not considered in this analysis. To ensure that there were no ice crystals present, the raw Ni measurements from the PMS 2D-C probe were examined, and aerosol particle data were not included when Ni exceeded 5 L−1. Note that ‘raw’ ice crystal concentrations are based solely upon the number of particles passing through the laser, and do not consider if they are in-focus, too small, or too large to be sized. For this reason, the threshold is assumed as a high value that represents a reference for possible ice crystal occurrence.

To better understand the importance of the variability of Na and how it changes with the averaging interval, three types of case studies from constant altitude flight legs were selected for comparison. Here, 1 s observations of Na are compared with arithmetic averages and standard deviations (SD) over different time-scales (section 3.1). Arithmetic averages (as opposed to geometric averages) of Na are used in this work to be consistent with the mean values of Na used in climate model simulations and previous parametrizations. The arithmetic SD values provide a measure of the variability in Na for each case study. Also included are vertical profile case studies, which demonstrate the variations in Na in the vertical domain. To aid in understanding where the air masses originated from, the HYSPLIT (Draxler and Rolph, 2010; Rolph, 2010) back trajectory model was utilized.

Following this analysis, vertical profiles are presented in section 3.2 using cloud-free periods from the entire field campaign. The data were separated into two categories, non-BB and BB cases. This separation was originally determined based on composition measurements (e.g. Earle et al., 2011); when the fraction of BB particles was high, such as during plumes of BB smoke from the west (Warneke et al., 2009) measured over several days (18–24 April 2008), Na was high, and therefore in this work, Na is used to separate the case studies into BB and non-BB. All other cases from before and after the BB cases are considered non-BB cases because there was a smaller BB particle fraction in the chemical composition (Zelenyuk et al., 2010; Earle et al., 2011) and maximum Na values are significantly lower (approximately ≤ 500 cm−3) than those observed during the BB-influenced period. Individual profiles within each of these two categories were averaged together by altitude.

Last, Na from time series at constant altitudes for all aerosol loading scenarios are averaged over various time-scales (representing space-scales by considering the aircraft true air speed) in section 3.3, again separating the data into non-BB and BB cases. Extreme values, represented by the 95th and 5th percentiles of Na, are used to demonstrate how layers of higher particle loading influence Na for different averaging scales.

3. Results

3.1. Case studies

The variability of Na is examined along constant altitude flight legs and vertical profiles for three different scenarios: non-BB cases with Na < 200 cm−3, representing typical background scenarios observed during the field campaign (section; non-BB cases with elevated Na observed in distinct layers (section; and BB cases (section 3.1.2.). A map of Alaska indicating the location of each constant altitude flight leg and vertical profile case is shown in Figure 1. Vertical profiles of Na, T, θe, wind speed (and wind direction indicated with arrows) are shown in Figures 2(a)–2(c), with 5-day back trajectories from HYSPLIT (Draxler and Rolph, 2010; Rolph, 2010) for selected altitudes within the vertical profiles shown in Figure 2(d). The Na observed during constant altitude flight legs are shown in Figure 3.

Figure 1.

Map of Alaska showing the aircraft flight tracks for the case studies presented in section 3.1 (three vertical profiles and three horizontal flight legs). Note that the Flight 25 profile falls along the same latitude/longitude line as the horizontal leg from the same flight. This figure is available in colour online at

Figure 2.

Vertical profiles of aerosol particle total number concentration (Na), air temperature (T), potential temperature (θe) and wind speed (with wind direction indicated with arrows, with the starting point of the arrow indicating the wind speed value). Note that instantaneous values of wind speed are shown every 5 s for readability, whereas 1 s data are shown for Na,T and θe. Note that θe is shown in degrees Celsius so that it is comparable to T. Vertical profiles for three different flights are shown: (a) Flight 11 (1736–1752 UTC 4 April), a low concentration (non-BB) case with no stratified aerosol particle layers; (b) Flight 15 (1628–1652 UTC 8 April), a low concentration (non-BB) case with distinct aerosol particle layers; (c) Flight 25 (2153–2206 UTC 19 April), a BB case. Also shown are (d) 5-day back trajectories using the HYSPLIT online model (Draxler and Rolph, 2010; Rolph, 2010) for a variety of endpoint heights for each case, with the location reflecting the vertical profile locale (i.e. Fairbanks or Barrow). This figure is available in colour online at

Figure 3.

Time series (UTC) of Na during horizontal flight legs and average Na values for different time-scales (1, 5 and 10 min) for: (a) a background non-BB case on 2 April 2008 (Flight 10), with the averaging scales of 5 min indicated by segments numbered 1 to 4 (also indicated is the 20 min averaging scale for the entire flight leg); (b) a non-BB case with elevated aerosol particle concentration in layers on 9 April 2008 (Flight 17), with selected 1 min averaging scales indicated by segments numbered 1 to 5 (also shown is the 40 min averaging scale); and (c) a BB case on 19 April 2008 (Flight 25) shown at three constant altitude flight legs (A, B and C), with overall averages indicated for each altitude, as well as selected 1 min time intervals. Error bars indicate one standard deviation of average Na values. This figure is available in colour online at

3.1.1. Low concentration cases (non-BB cases) Background cases

Figure 2(a) shows the vertical profile from ISDAC Flight 11 (4 April), flown over land near Fairbanks (Figure 1). In this case, minimal variations in Na are observed in the vertical domain, ranging from a minimum Na of 10 cm−3 at around 5200 m-MSL to a maximum of about 110 cm−3 at around 3500 m-MSL. Over the entire vertical profile, the average Na was 54 ± 22 cm−3. Note that unstable atmospheric conditions would lead to mixing throughout the profile, decreasing the variability observed in the aerosol concentration. The vertical profile of equivalent potential temperature, θe, (positive slope throughout) shows that the atmospheric conditions were stable over the entire profile. The temperature steadily decreased with height, and hence there were no temperature inversions during this profile that may lead to variations in Na. The profiles of θe and T are similar for the vertical profile case studies discussed in the following sections, showing stable conditions and no temperature inversions. The winds were quite calm during this profile, with a maximum wind speed of 10 m s−1 and a minimum of around 2 m s−1. The wind direction, indicated with the wind speed arrows, indicated a generally southerly direction, changing from southeast to southwest throughout the profile. The back trajectories for this flight (Figure 2(d)) show that the air mass travelled from the southwest (Pacific Ocean) and remained at higher altitudes (not shown), which can explain the low Na given the absence of significant anthropogenic sources along this trajectory.

A 20 min time series of Na along a constant, high altitude flight leg on 2 April (Flight 10, background case) is shown in Figure 3(a) (note that there was not a full vertical profile out of cloud for Flight 10, which is why the preceding vertical profile was from Flight 11). This portion of the flight was flown over land in the vicinity of Barrow (Figure 1) at an average altitude ± SD of 6188 ± 26 m-MSL. This represents a typical background aerosol concentration case observed during the project, with low Na and small variations in Na of 104 ± 34 cm−3 when averaged over the full 20 min interval. Superimposed on the 1 s data in Figure 3(a) are 1, 5 and 10 min averages of Na, with the 5 min averages labelled 1 through 4. The Na did not vary to a great extent over time segments 1 and 2 (likewise for time segment 3 compared with 4). However, comparing the first two segments to the latter two indicates some variation. During segments 1 through to 4, the average ± SD of Na for each segment are 132 ± 23 cm−3, 124 ± 27 cm−3, 84 ± 27 cm−3 and 77 ± 17 cm−3, respectively. Although Na is low for this case, a decrease in Na of 42% from time segment 1 to segment 4 is relatively large, and shows that even in background cases, significant relative changes in Na can occur. Low concentration cases with elevated concentration layers

During Flight 15 (8 April), distinct aerosol particle layers were clearly identified in a vertical profile over the coast near Barrow, as shown in Figure 2(b). This vertical profile was flown partially over land and partially over the Beaufort Sea (Figure 1) and showed a minimum Na of 5 cm−3 at 4000 m-MSL and a maximum Na of 520 cm−3 at 5700 m-MSL. The thickness of the layers is around 250–260 m in the vertical dimension. The average Na over the entire vertical profile was 171 ± 108 cm−3. Wind speeds fluctuated from a minimum of around 10 m s−1 to 15 m s−1 and wind directions were easterly at all levels of the profile. Back trajectories for these layers (Figure 2(d)) indicate that the particulate layers at 5100 and 5700 m-MSL show similar back trajectories, crossing over Alaska prior to looping over the Arctic Ocean. The altitudes of these 5-day back trajectories (not shown) remained above 3000 m-MSL. Although the exact source of these air masses is not known, one may assume from these trajectories that the aerosols have circulated the Arctic for days, and therefore may be the result of long-range transport. At the 2000 m-MSL back trajectory endpoint, the air mass traces back to the central Canadian Arctic (but the air mass did not drop below 2000 m-MSL according to the trajectory analysis) and the vertical profile below this level shows consistently higher concentrations (compared with the extreme variations in concentrations observed at higher altitudes), with Na steadily increasing towards the surface, suggesting a more regional source.

Figure 3(b) shows the time series of Na along a constant, high altitude flight leg over land during a transect between Fairbanks and Barrow at 6100 ± 6 m-MSL altitude on 9 April (Flight 17; see Figure 1), demonstrating distinct aerosol particle layers (e.g. at 0052, 0054 and 0111 UTC) in the horizontal domain. The aerosol particle layers are characterized by Na variations from about 20 cm−3 to 550 cm−3, similar to those observed in the vertical profile at altitudes > 4000 m-MSL. Because of these strong variations, the time intervals chosen for calculating the mean values of Na and associated SD are important. One minute averages, shown in Figure 3(b), still capture some of the aerosol particle structure (although the magnitude of the aerosol spikes are lost), but these details are increasingly obscured as the averaging interval is increased from 1 min to 5 and 10 min. The 40 min average over the whole horizontal flight leg was Na = 70 ± 40 cm−3. Selected 1 min time segments are labelled 1 through 5 in Figure 3(b), with 1 min averages of Na = 60 ± 11 cm−3, 177 ± 129 cm−3, 32 ± 12 cm−3, 83 ± 14 cm−3 and 171 ± 71 cm−3, respectively. Each of these 1 min averages is quite different from the overall 40 min average. This shows that averaging Na in the presence of layers of higher particle loading can lead to significant variations in bulk estimates of Na depending on the specific time interval selected.

3.1.2. Biomass burning influence (BB cases)

During ISDAC, BB plumes were encountered that were attributed to forest fires in Siberia and agricultural burning in Kazakhstan and southern Russia (Warneke et al., 2009). The vertical profile (Figure 2(c)) during Flight 25 (19 April), flown near Barrow over the Arctic Ocean (Figure 1) showed Na as low as 30 cm−3 at 5200 m-MSL and peaked at 2100 cm−3 at altitudes of 2400–2700 m-MSL. The thickness of the polluted layer was around 590 m during this vertical profile. The mean and SD of Na over the vertical profile was 559 ± 530 cm−3, with significant variability indicated by the SD. There were stronger wind speeds in this case than the low concentration vertical profiles (Flights 11 and 15), with a minimum wind speed of around 15 m s−1 and a maximum of close to 30 m s−1, from a southwesterly direction at all levels of the profile. The back trajectory ending at 2700 m-MSL (Figure 2(d)) traces back 5 days to the regions of forest and crop fires indicated by Warneke et al.(2009), who also reported increases in trace gases such as carbon monoxide, acetonitrile and benzene that are indicative of BB plumes during this period. Together with aerosol composition analysis (Earle et al., 2011), this provides supporting information that the plumes of increased Na observed in this work were BB emissions. At higher altitudes (5000 m-MSL), Na was lower, with a different back trajectory (dropping to no lower than 2000 m-MSL during the previous 5 days; not shown) that was mainly over the Pacific Ocean, with few potential sources of anthropogenic emissions.

Time series of Na sampled along horizontal flight legs at various altitudes for the same flight are shown in Figure 3(c). These flight legs were flown over the coast near Barrow, mainly over the ocean, but occasionally over land (Figure 1). This BB case shows significant variation in Na, from a minimum of 100 cm−3 to a maximum of over 2000 cm−3, demonstrating the significant concentration variations in BB plumes that were also observed in the vertical profile (Figure 2(c)). For the uppermost altitude (labelled A in Figure 3(c)), the overall average of Na is 675 ± 267 cm−3. At this altitude, there is significant variation in Na; for example, the 1 min averages labelled A1 and A2 in Figure 3(c) have Na = 914 ± 79 cm−3 and 197 ± 55 cm−3, respectively. For the lower two altitudes (B and C in Figure 3(c)), there were even more extreme variations, with Na reaching over 2500 cm−3 near the end of interval C. The overall 13 min average for C, for example, was Na = 1168 ± 888 cm−3; however, the 1 min averages of Na over C1 and C2 were 126 ± 28 cm−3, and 2258 ± 116 cm−3, respectively, demonstrating the extreme variations observed over these constant altitude flight legs. The large variations that are observed in Na over seconds to minutes are lost in the averages over 5 and 10 min.

3.2. Vertical profiles for entire study

This section summarizes results for vertical profiles from the entire project dataset, with data separated into non-BB and BB air masses. Figure 4(a) shows 24 vertical profiles of Na for 11 flights in non-BB air masses, in which the maximum Na was generally below 500 cm−3. Averages, as well as SD, were determined every 250 ± 5 m-MSL, from an altitude of 250 up to 6500 m-MSL. The average Na values for altitudes above and below 3500 m-MSL (chosen as an intermediate height for all profiles) were 146 ± 126 cm−3 (gradually decreasing with higher altitude) and 157 ± 98 cm−3, respectively, with the overall average Na = 152 ± 111 cm−3. The significant scatter indicated by the SD suggests that the variability was large.

Figure 4.

Vertical profiles of 1 s and 250 m-averaged aerosol particle number concentration as well as the average ± standard deviation lines for (a) non-BB cases and (b) BB cases between 18–24 April 2008

Figure 4(b) shows 10 vertical profiles of Na from four flights during the BB cases of 18–24 April 2008. During this period, the aerosol particle conditions changed drastically from day to day, flight to flight, and even within individual flights. For altitudes higher than 3500 m-MSL, the average of all BB profiles was Na = 519 ± 373 cm−3, while the average for altitudes lower than 3500 m-MSL was Na = 723 ± 555 cm−3. The average Na for all profiles over the full altitude range was 612 ± 477 cm−3, with Na ranging from 5 cm−3 up to a maximum of 2500 cm−3, especially around 2000 m-MSL (where the BB plume is particularly pronounced in some vertical profiles). This large variability has been observed previously in Arctic regions (Gultepe and Isaac, 2002).

Back trajectories from HYSPLIT (Draxler and Rolph, 2010; Rolph, 2010) were determined for three endpoint heights (2000, 4000 and 6000 m-MSL, not shown) for each vertical profile date/time and location. Earlier flights in the campaign (Flights 10, 11, 12, 15 and 17) showed air masses originating predominantly from the Arctic and oceanic regions, which explains the low Na observed. Occasionally, following the BB cases (Flights 30–34), the air masses moved over Siberia or Russian regions, but the air masses were predominantly from the Arctic and oceanic regions, where few local sources exist. In most of those cases, the air masses remained aloft (not shown) over the last 5 days and when the air masses reached ground level, it was in oceanic regions, which explains the low concentrations observed during Flights 30–34 (and the smaller fraction of BB particles (Zelenyuk et al., 2010)). Air masses measured during Flight 18 originated in Russian/Siberian regions, but the air masses did not reach lower levels; however, slightly elevated Na (compared with flights prior to this time), suggests possible long-range transport. During the BB cases (18–24 April), the back trajectories crossed over Asia and Russia (and reached ground level) more often than the non-BB cases, increasing the chances of encountering ground-based pollution. As discussed by Warneke et al.(2009), there were crop and forest fires in these regions during this period, leading to the higher Na observed during Flights 22–28.

3.3. Na averaging scale for all flight legs

Averages of Na over horizontal, constant altitude flight legs throughout the entire field campaign were calculated for Δt = 1, 2, 5, 7, 10 and 15 min of flying time (when these time spans were available); all flight legs considered in this analysis are compiled in Tables 2 and 3. By considering the average aircraft true air speed (∼ 85 m s−1) during the above time intervals, the distances travelled were about 7, 35 and 75 km for the 1, 5 and 10 min intervals, respectively. While comparing the scale of atmospheric variability and that of models along one dimension, it is important to keep in mind that these averages may not be applicable to three-dimensional models. This is a major issue when transferring the knowledge from aircraft observations to a model-based analysis. In Figure 5, mean values of Na are plotted against the distance over which the mean was calculated. The colour bar indicates the altitude—the cooler colours are for higher altitudes and thus colder temperatures, warmer colours are for lower altitudes and thus warmer temperatures. There were more cloud-free, extended horizontal flight legs at higher altitudes (colder temperatures), as there were often high altitude transit flights between Fairbanks and Barrow. Figure 5(a) shows analysis for non-BB cases (also shown in Table 2) and Figure 5(b) includes the BB cases (Table 3).

Figure 5.

Mean aerosol particle number concentration as a function of the averaging scale for all cloud-free horizontal flight legs that were 10 min or longer in Tables 2 and 3. The colour bar shows the altitude at which the measurements were made. The black solid lines represent mean aerosol particle number concentrations calculated using 1 s data. The dashed lines are fits through the 95th and 5th percentiles that were determined for each 20 km segment (shown with solid triangles). Shown here are the (a) non-BB cases and (b) BB cases. Equations for the fits shown in (a) for the 95th and 5th percentile (Nac95 and Nac5, respectively), in terms of the averaging distance, d, are Nac95 = −0.8058 × d + 284 and Nac5 = 0.6119 × d + 4.45, with correlation coefficients of R2 = 0.88 and 0.57, respectively. Equations for the fits through the 95th and 5th percentiles in (b) (Nab95 and Nab5; dashed lines) are Nab95 = −8.768 × d + 1853.9 and Nab5 = 3.5404 × d with correlation coefficients R2 = 0.89 and 0.57, respectively.

An average value of Na = 127 ± 84 cm−3 was determined from the non-BB 1 s data. For the majority of the non-BB cases, Na = 127 cm−3 (shown as a horizontal line in Figure 5(a)) is a good estimate for all averaging scales tested here. The 95th and 5th percentiles were determined for each 20 km segment (shown as solid triangles) and fits through these points are indicated by dashed lines. Here, 20 km segments (rather than 1 to 15 min segments) were chosen so that the calculations would be equally spaced with distance. With increasing averaging distance, d, the 95th percentile decreases towards the mean, and the 5th percentile increases towards the mean.

The BB cases show a large spread in Na values over different averaging scales, resulting from the strong aerosol particle concentration variations in the BB plumes. The average Na determined from 1 s data (13 cloud-free flight legs during six flights) was 720 ± 539 cm−3 (with Na = 720 cm−3 shown as a horizontal line in Figure 5(b)). The 95th and 5th percentiles for each 20 km segment are shown in Figure 5(b) (triangles). At averaging scales > 80 km, there were fewer data points for the BB cases, and hence, greater fluctuations in Na, especially in the 5th percentile. There is a much more pronounced decrease towards the mean for the 95th percentile, and likewise, a pronounced increase in the 5th percentile towards the mean with increasing averaging distance. Biomass burning cases, such as those presented here, with significant variations in Na, should be considered when averaging Na (as well as the upper and lower limits) to be used in climate simulations.

4. Discussion

A summary of average Na values from previous field campaigns during Arctic spring and summer is provided in Table 1. Leaitch et al.(1989) found an average Na of around 200 cm−3 at low altitudes and 130–150 cm−3 at higher altitudes during a springtime aircraft field campaign using similar instrumentation as employed during ISDAC. Gultepe and Isaac (2002) found Na = 215 ± 264 cm−3 for lower altitudes and Na = 348 ± 235 cm−3 for higher altitudes, with the SD clearly showing the variability of Na in Arctic spring. For non-BB vertical profile observations during ISDAC, there was not much difference between the average Na at higher (146 ± 126 cm−3) or lower (157 ± 98 cm−3) altitudes (over the entire profile, the average Na was 152 ± 111 cm−3) over 11 flights (Figure 4(a)). For ISDAC non-BB horizontal leg observations, Na = 127 ± 84 cm−3 was found to be the average concentration for 17 flights (Figure 5(a)), which was at the lower end of the measurements discussed in the literature (Table 1). Many of the average Na values provided in the literature do not give SD, and thus details on the variability during those studies are not known.

Pollution plumes observed during previous Arctic aircraft campaigns are also summarized in Table 1. Maximum Na values of 3100 cm−3 (Brock et al., 1989; altitude of this maximum not given), 4600 cm−3 at 200 m altitude (Hegg et al., 1995), 1000 cm−3 at around 2000 m altitude (Dreiling and Friederich, 1997) and 1200 cm−3 at mid-level altitudes (Gultepe and Isaac, 2002) were observed. The above pollution episodes, with Na maxima ranging from 1000 to 4600 cm−3, show the strong variability in Na measurements from case to case. During ISDAC, Na reached approximately 2500 cm−3 in the horizontal leg for Flight 25 (Figure 3(c)), 2100 cm−3 in the vertical profile for Flight 25 (Figure 2(c)) and 2600 cm−3 (Figure 4(b)) during the BB case vertical profiles, which are within the range of maximum measurements from the literature.

Climate models often use simple parametrizations with respect to Na, with large grid spacings up to 500 km in size (e.g. Schmidt et al., 2006). During ISDAC, measurements were collected every second (measurement frequency of 1 Hz), corresponding to a distance of approximately 100 m. This value can vary for aircraft field campaigns, with typical values ranging from 50 to 200 m depending on the speed of the aircraft. This work shows how the details of Na variability are lost as the averaging scale increases. Figure 5(a) shows that the mean Na of 127 cm−3 for non-BB cases would be a reasonable estimate for the mean aerosol particle number concentration in the Arctic for background cases, but by making this kind of assumption, one would lose the vertical or horizontal structure observed for aerosol particle layer cases (Figures 2(b) and 3(b)), and would certainly greatly misrepresent conditions during BB cases (Figures 2(c) and 3(c)). For such episodes, the estimated mean Na was 720 cm−3 (Figure 5(b)) but this fails to capture the fine details of the aerosol particle number fluctuations, which are important for cloud formation, microphysical properties and radiative effects. The 95th and 5th percentiles show the extremes in Na variability (Figure 5) and decrease/increase towards the mean with increasing averaging scale for both non-BB and BB cases, but the decrease/increase towards the mean are much more significant in the BB cases. The 95th and 5th percentiles should be considered in climate simulations, as they represent upper and lower limits of Na in Arctic spring as observed during ISDAC. It is possible that these upper and lower limits may have dramatically different implications for aerosol effects on the radiation budget. The direct radiative forcing in the Arctic has been shown to vary from a cooling of −1.5 W m−2 to a warming of +1.2 W m−2 depending on the aerosol particle chemical and optical properties (e.g. Hegg et al., 1996; Iziomon et al., 2006; Quinn et al., 2008), and still represents a large unknown for climate modellers. The aerosol indirect effect is an even larger unknown in climate change assessment studies because of lack of knowledge concerning nucleation processes (IPCC, 2007). Further studies are required to demonstrate how variations in the aerosol number concentrations detailed in this work would affect the direct and indirect radiative forcings.

5. Conclusions

The variability of aerosol particle number concentrations (Na) was investigated for cloud-free and precipitation-free aircraft flight legs over Alaska and the Beaufort Sea in April 2008. Significant variability in Na in both BB and non-BB cases was observed throughout the project. The main conclusions of this work are summarized as follows.

  • (1)Significant variability in Na was observed even for background (non-BB) cases, with relatively low particle number concentrations. For a horizontal flight leg during a background case, the observed variability in Na averaged over 5 min time segments ranged from 77 cm−3 to 132 cm−3 (a 42% difference). A vertical profile in similar aerosol conditions showed Na varying from 10 cm−3 to 110 cm−3. Back trajectories indicated that the air travelled from the southwest (Pacific Ocean), where there is an absence of significant anthropogenic sources.
  • (2)Distinct layers of elevated aerosol particle concentration were observed during a horizontal flight leg (non-BB case), characterized by Na variations up to an order of magnitude (< 50 to 500 cm−3) over very short time-/distance scales (5 min or ∼ 30 km). Air masses with distinct aerosol particulate layers were also observed in the vertical, with Na ranging from < 50 to 550 cm−3 and layers that were 250–260 m thick in the vertical.
  • (3)Non-BB flights prior to the BB cases had air masses originating predominantly from the Arctic and oceanic regions. Occasionally, non-BB flights that followed the BB cases showed back trajectories over Siberian or Russian regions, but the air masses were generally from Arctic and oceanic regions, where few local sources exist.
  • (4)Distinct BB cases were observed with Na ranging from 125 to 2260 cm−3 in the horizontal. Biomass burning plume layers were also observed in vertical profiles, with 590–820 m thickness in the vertical at around 2000–2500 m-MSL altitude, and Na varying from < 30 to 2100 cm−3. This suggests that the vertical resolution in climate simulations should be better than 500 m.
  • (5)During the BB case, the back trajectories crossed over Asian and Russian regions when crop and forest fires were burning (Warneke et al., 2009), leading to plumes with a large fraction of BB particles (Earle et al., 2011) and with higher Na.
  • (6)The mean Na determined for all horizontal legs during the non-BB cases was 127±84 cm−3, whereas that for the BB cases was 720±539 cm−3. The mean Na determined for all vertical profiles during the non-BB and BB cases were 152 ± 111 cm−3 and 612 ± 477 cm−3, respectively, with the SD indicating significant variations. Hence, these average values fail to capture the fine details of the aerosol particle number fluctuations in both non-BB and BB cases.
  • (7)The marked changes in Na in all air masses considered suggest that the averaging scales used to determine representative aerosol particle number concentrations for GCMs need to be chosen with caution.


Thanks to Greg McFarquhar, Steve Ghan, Walter Strapp, Alexei Korolev, W. Richard Leaitch, Anne Marie Macdonald, Mohammed Wasey, Rob Reed, Mark Couture, Stewart Cober and the National Research Council of Canada (NRC) piloting and technical staff. The authors gratefully acknowledge the NOAA Air Resources Laboratory for the provision of the HYSPLIT transport and dispersion model and/or READY website ( used in this publication. Funding for this work was provided by the Office of Biological and Environmental Research of the US Department of Energy (Grant No. DE-FG02-08ER64554) through the Atmospheric Radiation Measurement (ARM) program and the ARM Aerial Vehicle Program with contributions from the DOE Atmospheric Sciences Program (ASP) and Environment Canada. Data were obtained from the ARM program archive, sponsored by the US DOE, Office of Science, Office of Biological and Environmental Research, Climate and Environmental Sciences Division. Some additional funding was also provided by the European COST-722 fog initiative project office. Thanks also to two anonymous reviewers for helpful comments that improved this manuscript.