The dependence of ice microphysics on aerosol concentration in arctic mixed-phase stratus clouds during ISDAC and M-PACE

Authors


Corresponding author: R. C. Jackson, Department of Atmospheric Sciences, University of Illinois, 105 S. Gregory St., Urbana, IL 61820, USA. (rjackso2@illinois.edu)

Abstract

[1] Cloud and aerosol data acquired by the National Research Council of Canada (NRC) Convair-580 aircraft in, above, and below single-layer arctic stratocumulus cloud during the Indirect and Semi-Direct Aerosol Campaign (ISDAC) in April 2008 were used to test three aerosol indirect effects hypothesized to act in mixed-phase clouds: the riming indirect effect, the glaciation indirect effect, and the thermodynamic indirect effect. The data showed a correlation of R = 0.78 between liquid drop number concentration, Nliq inside cloud and ambient aerosol number concentration NPCASP below cloud. This, combined with increasing liquid water content LWC with height above cloud base and the nearly constant vertical profile of Nliq, suggested that liquid drops nucleated from aerosol at cloud base. No evidence of a riming indirect effect was observed, but a strong correlation of R = 0.69 between ice crystal number concentration Ni and NPCASP above cloud was noted. Increases in ice nuclei (IN) concentration with NPCASP above cloud for 2 flight dates combined with the subadiabatic LWC profiles suggest possible mixing of IN from cloud top consistent with the glaciation indirect effect. The lower Nice and lower effective radius rel for the more polluted ISDAC cases compared to data collected in cleaner single-layer stratocumulus conditions during the Mixed-Phase Arctic Cloud Experiment is consistent with the operation of the thermodynamic indirect effect. However, more data in a wider variety of meteorological and surface conditions, with greater variations in aerosol forcing, are required to identify the dominant aerosol forcing mechanisms in mixed-phase arctic clouds.

1. Introduction

[2] The Arctic is currently experiencing rapid changes in temperature and sea ice coverage attributable to climate change [Intergovernmental Panel on Climate Change (IPCC), 2007]. Clouds play a critical role in determining the radiation budget through their role in a complex feedback involving the sea ice, clouds, aerosols, and the atmosphere [Curry et al., 1993; Curry, 1995]. A lack of understanding of cloud-aerosol interactions significantly contributes to the uncertainty in climate change prediction [IPCC, 2007]. Given the transport of anthropogenic pollution from industrialized regions to the Arctic [e.g., Kerr, 1981; Shaw, 1982; Barrie, 1986] and the role of advected biomass burning products [Warneke et al., 2009], it is necessary to improve the understanding of cloud-aerosol interactions in order to refine arctic climate predictions.

[3] For all-liquid clouds, Twomey [1974] proposed that increases in the concentration of cloud condensation nuclei (CCN) increase the number concentration of cloud droplets and decrease the effective radius of droplets (rel) under conditions of constant liquid water content (LWC); this is commonly known as the first indirect effect. The increased total surface area of cloud droplets leads to the reflection of more solar radiation back to space, reducing the amount reaching the surface, providing a net cooling effect. Longwave emissivities of thin arctic stratus also increase with CCN concentration, providing a net warming effect in the dark winter months when shortwave radiation is negligible [Garrett et al., 2002; Lubin and Vogelmann, 2006]. Therefore, increasing CCN can either warm or cool the Arctic depending on the time of year, having possible significant impact on sea ice-cloud albedo feedbacks.

[4] The second indirect effect [Albrecht, 1989] postulates that the reduced droplet sizes associated with the increases in CCN inhibit droplet growth by collision-coalescence, reducing the frequency and/or amount of precipitation, increasing cloud lifetime and also consequently increasing the reflection of solar radiation. Pinto et al. [2001], Peng et al. [2002], and Garrett et al. [2004] have observed decreasing rel and drizzle concentration with increasing CCN in maritime arctic clouds consistent with the Twomey [1974] and Albrecht [1989] hypotheses. Earle et al. [2011], however, observed that although the droplet concentration increased with aerosol concentration in liquid-dominated clouds, rel did not decrease, due primarily to variations in cloud thickness with CCN concentration for the arctic stratus they considered.

[5] Shupe et al. [2005] observed that all-liquid clouds occurred only 19% of the time during the Surface Heat Budget of the Arctic (SHEBA) mission, with ice clouds occurring 38% of the time and mixed-phase clouds accounting for the other 43% of the observations. Two different types of mixed-phase clouds have been observed in the Arctic: single-layer clouds [i.e., Hobbs and Rangno, 1998; Rangno and Hobbs, 2001; Shupe et al., 2001; McFarquhar et al., 2007], and multilayer clouds where multiple liquid layers exist between layers of ice [Hobbs et al., 2001; Intrieri et al., 2002a]. The latter type occurred nearly half of the time during SHEBA [Shupe et al., 2005].

[6] The nature of interactions between aerosols and mixed phase clouds is not well understood. Homogeneous nucleation requires no ice nucleus (IN) to initiate freezing and only occurs at temperatures at or below about −38°C. For typical temperatures of arctic mixed phase clouds, heterogeneous nucleation mechanisms requiring aerosols to act as IN are more likely to occur. For the Arctic, Gultepe et al. [2001] found no strong relationship between ice crystal and aerosol concentrations during the First International Cloud Climatology Project Regional Experiment - Arctic Clouds Experiment (FIRE-ACE). Three cloud-aerosol interactions listed by Lohmann and Feichter, 2005] hypothesized to act in mixed-phase clouds are illustrated in Figure 1 and detailed here: (1) a glaciation indirect effect whereby increases in the fraction of aerosol particles acting as contact ice nuclei increase the number concentration of ice crystals and increase the amount of ice phase precipitation [Lohmann, 2002]; (2) a riming indirect effect whereby increases in CCN concentration result in decreases in the sizes of liquid cloud droplets and a less efficient riming process that reduces the mass of the ice crystals and hence the ice water content (IWC) [Borys et al., 2003]; and (3) a thermodynamic indirect effect whereby decreases in median droplet sizes associated with increases in CCN decrease the ice crystal number concentration by reducing the number of drops large enough to initiate secondary ice crystal production. Such secondary production mechanisms include contact nucleation of drizzle-size drops [Rangno and Hobbs, 1991], rime splintering, momentary bursts of high supersaturation (>10%) in air around a freezing drop, shattering of freezing drops, and/or fragmentation of ice crystals via droplet-crystal and crystal-crystal collisions [Rangno and Hobbs, 2001].

Figure 1.

Conceptual diagram of the (a) glaciation indirect effect, (b) riming indirect effect, and (c) thermodynamic indirect effect. Blue ellipses denote cloud droplets, black ellipses denote CCN or IN, and snowflakes represent ice crystals.

[7] Prior studies have suggested that arctic stratus could contribute to a net warming effect of 20 W m−2 during the fall and spring season [Dong et al., 2001], and to a total forcing of 30 W m−2 in fall, spring, and winter [Intrieri et al., 2002b]. However, the radiative influence depends on the microphysical properties of the clouds. Model simulations by Harrington and Olsson [2001] showed that increases in ice crystal concentration could reduce surface heat fluxes by as much as 80 W m−2. Thus, to determine the radiative effects of ice and mixed phase clouds, the dependence of cloud ice microphysical properties on aerosol concentration and composition must be known because in the concentrations and compositions of arctic aerosols vary considerably throughout the year [Shaw, 1982; Barrie, 1986; Quinn et al., 2002].

[8] Lance et al. [2011] observed narrower liquid drop size distributions and lower ice concentrations in polluted clouds compared against clean clouds during the Aerosol, Radiation, and Cloud Processes affecting Arctic Climate (ARCPAC) campaign. Data collected during a concurrent field study, the Indirect and Semi-Direct Aerosol Campaign, conducted by the Department of Energy Atmospheric Radiation Measurement program, Environment Canada, and other partners in April 2008 in the vicinity of Barrow, Alaska [McFarquhar et al., 2011a], are examined in this study to determine the conditions under which the aforementioned indirect effects operate in arctic clouds. Past studies of arctic stratus, including FIRE-ACE [Curry et al., 2000], the Surface Heat Budget of the Arctic Ocean (SHEBA) experiment [Uttal et al., 2002], and the Mixed Phase Arctic Clouds Experiment (M-PACE) [Verlinde et al., 2007] collected valuable cloud and aerosol data in arctic stratus. However, M-PACE did not collect airborne total aerosol concentrations. The aerosol conditions were generally clean during M-PACE and during May 1998 when boundary layer mixed-phase clouds were present in FIRE-ACE, providing little contrast of aerosol conditions when mixed-phase stratus were present. Thus, observations of polluted boundary layer mixed-phase stratocumulus collected during ISDAC are more conducive to the investigation of aerosol-cloud interactions.

[9] During ISDAC, the National Research Council of Canada (NRC) Convair-580 research aircraft flew five sorties through single-layer mixed phase stratus and stratocumulus on three different days during April 2008. The single-layer nature of these clouds provided an environment free from complications of seeder-feeder mechanisms from ice falling from above the liquid layer [Cotton and Anthes, 2011; Hobbs et al., 2001; Lawson et al., 2001]. The ISDAC data set is unique for investigating cloud-aerosol interactions because the NRC Convair-580 was equipped with 41 different cloud and aerosol probes sampling the complete range of sizes of cloud and aerosol particles. The redundancy and intercomparison of quantities from varying cloud probes allowed for more confidence in the generated microphysical measurements.

[10] In this study, the importance of the three different indirect effects postulated to occur in mixed-phase clouds was tested. The dependence of cloud microphysical properties, namely the total number concentrations of liquid drops (Nliq) and ice crystals (Nice), the number distribution functions of ice (Nice(D)) and water (Nliq(D)) particles, IWC, and rel on ambient aerosol and IN concentrations above and below cloud was examined. To further investigate how variations in aerosols affect cloud properties, data collected during ISDAC in more polluted conditions over broken sea ice were compared with those collected during M-PACE when four flights were made in single-layer stratus in pristine conditions over open water.

2. Case Overview

2.1. Sea Ice Characteristics

[11] Five flights conducted in, above, and below single-layer mixed-phase stratocumulus in the vicinity of Barrow were analyzed: two flights on 8 April 2008, one flight on 18 April 2008 and two flights on 26 April 2008. Figure 2 shows the corresponding flight tracks of the NRC Convair-580, excluding high altitude transit portions of flights (i.e., altitudes above 2 km for 8 and 26 April) and regions identified as cirrus or multilayer clouds for 18 April. The stratocumulus sampled was located over broken sea ice in the Beaufort and Chukchi Seas on 8 April and 26 April, and over the land on 18 April, with the sea ice concentration estimated from the Special Sensor Microwave Imager/Sounder (SSMIS) on the Defense Meteorological Satellite Program (DMSP) F17 satellite [Maslanik and Stroeve, 1999]. Regions with sub-100% sea ice concentrations, most prominent near the coast, contained leads, cracks, and polynyas. As expected due to the springtime melt, the sea ice concentration was lower on 26 April than on 8 April. It is possible, and perhaps likely, that the microphysical properties of the clouds were affected by the surface conditions as suggested by previous observational [Gultepe et al., 2003] and modeling studies [Morrison et al., 2008]. However, given the 25 km and daily resolution of the sea ice data and the few cases sampled, it is difficult to segregate the data by sea ice extent.

Figure 2.

(a) Flight tracks through mixed phase single-layer stratocumulus in the vicinity of Barrow, Alaska on 8, 18 and 26 April 2008. Only time periods below 2 km for the 8 and 26 April case and the single-layer case of April 18 are shown, with time periods through cirrus and multilayer clouds excluded. (b) Sea ice concentrations and flight tracks over the region for 8 April with flight tracks superimposed. (c) As in Figure 2b but for 26 April (credit: National Sea Ice and Snow Data Center).

2.2. Overview of Aerosol Transport

[12] Back trajectories for 6 days (not shown) were calculated using the National Oceanic and Atmospheric Administration HYbrid Single-Particle Lagrangian Integrated Trajectory (HYSPLIT) model (R. R. Draxler and G. D. Rolph, HYSPLIT, 2011, http://ready.arl.noaa.gov/HYSPLIT.php; G. D. Rolph, Real-time Environmental Applications and Display sYstem (READY), 2011, http://ready.arl.noaa.gov) from the location of the NRC Convair-580, namely Barrow, Alaska on 8 and 26 April and the midpoint of the constant altitude flight legs on 18 April. Altitudes for the trajectories were chosen both above and below the location of the liquid cloud layer, as estimated from the in situ profiles. The 8 April air masses at 500 m (below cloud) and 1500 m (above cloud) originated from northern Canada. For a height of about 1000 m above the capping inversion (2500 m), the air mass came from western Alaska. For 18 April, the air mass at 3000 m, corresponding to below cloud base, was traced to northern Canada, but did travel through altitudes of less than 1000 m over Siberia on 16 and 17 April, Air masses at 4000 m (above cloud top) and at 5000 m (an altitude above the capping inversion) were traced to 5000 m above ground level in northern Russia. The air mass trajectory below cloud was consistent with observations of biomass burning plumes from northern Russia present in the vicinity of Barrow on 18 April [Warneke et al., 2009]. For 26 April, the air mass below (500 m) and above cloud (1500 m) originated from the Arctic Ocean, while the air mass about 1000 m above the inversion (2500 m) originated from the Pacific Ocean. Differences in air mass source regions between the three cases suggested that the aerosol characteristics and their properties, such as ice nucleating ability, could vary between the days.

2.3. Meteorological Overview

[13] On 8 April, a stratocumulus deck formed over Barrow on the southwest side of a ridge in surface easterly flow associated with a deep 1040 hPa high pressure system over the North Pole, and northeast of a weakening low pressure system over the Bering Strait. The deck persisted throughout 8 April, but dissipated as the ridge of high pressure moved west from Barrow on 9 April. On 18 April, the meteorological situation was different, with an omega block and a ridge present over central Alaska [McFarquhar et al., 2011a]. A shortwave trough propagated around the ridge, providing forcing for a cloud deck over land, about 60 km from the coast east-southeast of Barrow. The cloud deck sampled on 18 April occurred at higher altitudes and colder temperatures than that sampled on 8 April. On 26 April, a high pressure system was again present over the North Pole, similar to the conditions of 8 April. A stratocumulus deck formed over the thin sea ice north of Barrow and persisted for approximately 15 h. The preceding analysis established that conditions on 18 April were quite distinct from those of 8 and 26 April – synoptic patterns were quite different, the cloud level temperature was about 3–4°C colder, and the aerosol composition was dominated by remnants of biomass burning on 18 April instead of the organics sampled on 8 and 26 April [Earle et al., 2011].

3. Methodology

3.1. Instrumentation

[14] The complete set of instrumentation installed on the NRC Convair-580 was summarized by McFarquhar et al. [2011b]. Only the probes from which data are used in this study are discussed here. Many different instruments measured cloud particle size distributions (SDs) as a function of particle maximum dimension D. Probes that converted the amount of light scattered in the forward direction to particle size included a Cloud Aerosol Spectrometer (CAS) that measured particles with 0.53 < D < 50 μm, a Forward Scattering Spectrometer Probe (FSSP) for 0.5 < D < 47 μm, and a Cloud Droplet Probe (CDP) for 2 < D < 50 μm. Optical array probes that determined particle size from the occultation of a laser beam incident upon an array of photodiodes included the following: a 2D Stereo Probe (2DS), nominally sizing particles with 10 < D < 1280 μm; a 2D Cloud Probe (2DC) for 25 < D < 800 μm; a 2D Precipitation Probe (2DP) for 200 < D < 6400 μm; a 15 μm resolution Cloud Imaging Probe (CIP1) for 15 < D < 960 μm (for flights before 19 April 2008); and a 25 μm resolution Cloud Imaging Probe (CIP2) for 25 < D < 1600 μm available for all flights. Not all of the probes were installed and working on every flight. Data from the 2DC, 2DP, and CIP1/2 were processed at the University of Illinois using image reconstruction techniques [Heymsfield and Parrish, 1978] and filtering out shattered artifacts by rejecting all particles with interarrival times <10−4 s following Field et al. [2006]. The CDP data were also processed at the University of Illinois. Diameter correction algorithms for hollow liquid spherical particles were applied to the 2DS, 2DC, and CIP2 data [Korolev, 2007]. The 2DS data were processed by Stratton Park Engineering Company filtering out shattered artifacts using the algorithms presented by Lawson [2011]. All other cloud probe data were processed at Environment Canada.

[15] A Passive Cavity Aerosol Spectrometer Probe (PCASP-100X) measured ambient aerosol concentrations and SDs of particles with 0.1 < D < 3 μm. Data from the PCASP were only used when the CDP or FSSP number concentration was less than 1 cm−3, so that any influence of cloud particles on measured aerosol concentrations was minimized. A Continuous Flow Diffusion Chamber (CFDC) provided IN concentrations at varying supersaturations and temperatures [Rogers et al., 2001]. A bifurcating valve was manually set to switch sampling from either an aerosol inlet or from a Counterflow Virtual Impactor (CVI) line that allowed sampling of residuals from cloud droplets and ice crystals [Twohy et al., 1997]. Long averaging times during constant altitude flight legs were required to get a statistically significant sample from the CFDC. No CFDC data were available for the first flight of 8 April, and the CFDC recorded IN concentrations below the noise level 55% of the time for the second flight of 8 and both flights of 26 April. IN concentrations below the noise level of the CFDC were assumed to be zero in this study. A counter measuring CCN concentrations at 2 differing supersaturations that were varied throughout the flight was also installed. The PCASP data were processed at Environment Canada [Earle et al., 2011] and the CFDC data were processed at Texas A&M University.

[16] A King probe provided estimates of LWC [King et al., 1985]. A Cloud Spectrometer and Impactor (CSI) measured total water content (TWC) by evaporating particles with D > ∼5 μm and then using a hygrometer to measure the resulting vapor [Noone et al., 1988]. A Deep Cone Nevzorov probe measured LWC and TWC by heating a wire at a constant temperature contained inside a cone that collects cloud water and ice and measuring the energy required to maintain that temperature when either water, or water and ice evaporated on the wire [Korolev et al., 1998]. A Rosemount Icing Detector (RICE) detected the presence of supercooled water by measuring the voltage changes associated with water freezing on a vibrating cylinder. A Rosemount sensor measured ambient air temperature. The NRC Convair-580 was also equipped with vertically and horizontally pointing NRC Airborne (NA X) and W-band radars.

3.2. Phase Identification

[17] The algorithm used to determine the cloud phase was based on previous algorithms [Cober et al., 2001; McFarquhar and Cober, 2004; and McFarquhar et al., 2007]. The RICE voltage change threshold of 2 mV s−1 [Cober et al., 2001] and the standard deviation of the CDP/FSSP SDs were primarily indicators of the presence supercooled water. Visual inspection of the CPI, 2DC, 2DS, and CIP1/2 imagery was used to determine the presence of ice or drizzle. When mixed phase conditions were present, it was assumed that all particles measured by the CDP or FSSP with D < 50 μm were liquid following McFarquhar et al. [2007]. Visual inspection of CPI images distinguished whether particles with 50 < D < 125 μm were liquid (drizzle) or ice. Drizzle was present in this size range for 4.1% of all mixed phase time periods. The mean area ratio, defined as the cross-sectional area of the particle divided by the area of a circle with the same D as the particle [McFarquhar and Heymsfield, 1996] of particles in this size range was 0.89 ± 0.16 in drizzle, 0.54 ± 0.20 in ice, and 0.63 ± 0.25 in mixed-phase conditions with no drizzle for 50 < D < 125 μm. Therefore, particles with 50 < D < 125 μm were assumed to be ice if no drizzle was observed in CPI images. Drizzle was present for D > 125 μm in only 1% of all mixed phase cloud periods, so particles of D > 125 μm were assumed to be ice in the derivation of the bulk cloud properties.

3.3. Derivation of Cloud Microphysical Parameters

[18] Post-flight calibrations conducted at Environment Canada indicated that the CAS gave questionable concentrations, the FSSP consistently undersized particles, and the CDP gave drop sizes consistent with bead calibrations for ISDAC. Hence, the CDP was the primary instrument used to derive the SDs for D < 50 μm. The 30 s averages of King Probe LWC (LWCKing) were compared against those derived from the CDP SDs (LWCCDP) for all liquid and mixed-phase cases during ISDAC. The best fit equation describing these parameters (R = 0.99) is:

display math

[19] The median difference between LWCCDP and LWCKing was 21%, which is slightly larger than the stated uncertainty of 15% for the King probe [King et al., 1985]. It was not possible to determine whether the CDP or King probe was offset. For the second flight of 8 April, when the CDP did not successfully record data, the FSSP SDs were modified based on a post-project calibration to account for the undersizing and subsequently used to derive the liquid bulk cloud properties.

[20] For ice conditions, past studies [Gardiner and Hallett, 1985; Gayet et al., 1996; Field et al., 2003; McFarquhar et al., 2007; Korolev et al., 2011] have suggested that the shattering of large crystals on the shrouds or inlets of the FSSP or CAS may artificially amplify the measured concentrations of ice crystals with D < 50 μm. On the other hand, the CDP frequently records few, if any, particles in ice-phase conditions [e.g., Plummer et al., 2010]. Further, the irregular and poorly resolved shapes of small ice crystals [Um and McFarquhar, 2011] make it difficult to convert the amount of forward scattered light to a particle size because Mie theory does not apply to non-spherical ice crystals. The number concentrations measured by the 2DS and CDP differed by 185% for 10 < D < 50 μm, and by 41% and 68% for 20 < D < 50 μm and 30 < D < 50 μm respectively.

[21] The 2DS was used to derive N(D) for 50 < D < 300 μm in ice and mixed-phase conditions. From analysis of data collected during the Rain in Cumulus over the Ocean experiment (RICO), Lawson et al. [2006] observed that the 2DS detected particles of D < 150 μm when the 2DC saw few or none due to its slower response time. However, the response times of different versions of 2DCs and CIPs vary [Strapp et al., 2001], so any conclusion drawn from the probes used during RICO cannot be extended to other versions of the same probe. The small and poorly defined depth of field of the 2DC and CIP for small crystals [Baumgardner and Korolev, 1997] also results in large uncertainties in the derived concentrations. The 2DS, 2DC, CIP1, and CIP2 data were all processed using algorithms that remove shattered artifacts because Korolev et al. [2011] demonstrated that shattering can artificially amplify the concentrations of particles with D as large as 500 μm. Lawson [2011] showed that application of these algorithms is more effective than redesigned probe tips at removing artifacts from the 2DS data. However, Korolev et al. [2011] showed that such algorithms were less effective than the use of tips specifically designed to mitigate shattering artifacts in the 2DC. Thus, although there are some uncertainties in the derived concentrations, the state-of-the-art probes and processing algorithms used during ISDAC provide a better description of SDs in this size range than was possible during the previous M-PACE and FIRE-ACE campaigns.

[22] The ratios of N(D) for the different probes may be a function of true airspeed due to optical array probe response time limitations [Strapp et al., 2001], and a function of particle size for the reasons previously mentioned. Hence, Figure 3 shows N2DS(D)/N2DC(D) and N2DS(D)/NCIP2(D) for varying true air speeds in uniform 50 μm bin widths. The spread in N2DS(D)/N2DC(D) and N2DS(D)/NCIP2(D) with varying airspeeds for D < 300 μm is consistent with the previously discussed uncertainties in small particle measurements. Further, 0.5 < N2DS(D)/N2DC(D) < 1.1 and 0.5 < N2DS(D)/NCIP2(D) < 1.1 for D > 300 μm, indicating a systematic low bias in N2DS(D) relative to N2DC(D) and NCIP2(D) in this size range.

Figure 3.

Mean N2DS(D)/N2DC(D) and N2DS(D)/NCIP2(D) for all ice phase cases in ISDAC for varying true air speeds (TAS).

[23] Figure 4 shows the ratio of the number concentrations measured by the CIP2 to that by the 2DC for 300 μm < D < 800 μm and to that by the 2DP for 800 μm < D < 1400 μm as a function of total number concentration measured by the 2DP for D > 800 μm (N2DP) for all ice phase cases during ISDAC. The increased spread in NCIP2(300 < D < 800 μm)/N2DC(300 < D < 800 μm) and NCIP2(800 < D < 1400 μm)/N2DP(800 < D < 1400 μm) for N2DP < 0.001 cm−3 in Figure 4 is likely due to a statistical sampling error given the few large particles detected in 10 s periods. The lack of dependence of NCIP2(300 < D < 800 μm)/N2DC(300 < D < 800 μm) (R = −0.15) on N2DP indicates that the probe sizing and response were not problematic for ISDAC. In general, N2DC(300 < D < 800 μm) and NCIP2(300 < D < 800 μm) agreed within 25%, with a mean difference of 19%, suggesting that one cannot distinguish between the performance of these probes. However, to avoid time periods when concentrations are uncertain, only time periods when 0.5 < NCIP2(800 < D < 1400 μm)/N2DP(800 < D < 1400 μm) <2 or N2DP(800 < D < 1400 μm) and NCIP2(800 < D < 1400 μm) = 0 were used in the subsequent analysis.

Figure 4.

(a) 30 s averages of NCIP2(300 < D < 800 um)/N2DC(300 < D < 800 um) and (b) NCIP2(800 < D < 1400 um)/N2DP(800 < D < 1400 um) plotted as a function of N2DP and colorized by TAS. Solid lines represent mean values in a given N2DP.

[24] Mass closure tests, in which the IWC measured by the bulk probes was compared against the IWC derived from the SDs, were used to further assess the optimum probe for characterizing N(D) for 300 < D < 800 μm, and to assess techniques for calculating ice mass from two-dimensional images. The mass closure tests were applied to only a subset of ISDAC data due to an uncorrectable positive baseline offset in the IWC caused by electrical interference from an unknown source that impacted the Nevzorov probe for much of ISDAC. The CSI also sometimes experienced an uncorrectable negative baseline offset and occasionally did not detect IWC in the presence of 2DS/2DC/2DP/CIP2 imagery. Consequently, differences in the CSI and Deep-Cone Nevzorov probe IWC sometimes exceeded an order of magnitude. Manual inspection of data was used to identify times when data from the two bulk probes were consistent and free of uncorrectable effects. Only time periods in which the CSI and Nevzorov IWC differed by less than 50% were used in the mass closure tests. IWC for these times was typically greater than 0.03 g m−3, representing only 1.6% of the total ice-phase data set.

[25] Two different methods for estimating crystal mass from the two-dimensional crystal images and ice crystal Nice(D), were used in the closure tests. One method, ‘CPI-mD’, derives IWC from using

display math

where the habit distribution fk(Dj) is the fraction of crystals in the bin centered at maximum dimension Dj having crystal habit k, αk and βk are habit-dependent coefficients used to determine crystal mass m as a function of crystal maximum dimension using m = αkDjβk [Brown and Francis, 1995; Mitchell, 1996], and Nice(Dj) is the number distribution function for the bin j with midpoint Dj. The fk(Dj) were derived using an automated habit identification scheme applied to CPI imagery at 60 s resolution that sorts crystals into the nine habit classes listed in Table 1 based on their morphological characteristics [Um and McFarquhar, 2009]. The 60 s averaging was required to get an adequate sample size. Because the CPI has a smaller sample volume than the 2DC or CIP, the required averaging period was larger and the habit distributions were applied to each of the size distributions occurring within the 60 s period.

Table 1. Table of m-D Relations Used to Calculate IWC Using Equation (1)
Habitα (g cmβ)βReference
Sphere0.91* π/63 
ColumnD < 30 μm 0.91*π/6D < 30 μm 0.91*π/6Mitchell [1996] – “Hexagonal Columns”
30 < D < 300 μm 0.00166630 < D < 300 μm 1.91
D ≥ 300 μm 0.000907D ≥ 300 μm 1.74
Plate0.007392.45Mitchell [1996] – “Hexagonal Plates”
StellarD < 90 μm 0.00583D < 90 μm 2.42Mitchell [1996] – “Stellar crystal with broad arms”
D ≥ 90 μm 0.000270D ≥ 100 μm 1.67
DendriteD < 100 μm 0.00583D < 100 μm 2.42Mitchell [1996] – “Broad branched crystal”
D ≥ 300 μm 0.000012D ≥ 300 μm 1.80
Rosette and Budding RosetteD < 90 μm 0.000012D < 90 μm 1.52Mitchell [1996] – “Bullet rosettes, 5 bullets”
D ≥ 90 μm 0.00308D ≥ 100 μm 2.26
Small irregular and Big irregularD < 100 μm look-up table based on Gaussian random sphereD < 100 μm look-up table based on Gaussian random sphereNousiainen and McFarquhar [2004] and Brown and Francis [1995]
D ≥ 100 μm 0.00294D ≥ 100 μm 1.9

[26] A second method for computing the mass, denoted ‘BL06,’ used the projected mass-area relation of Baker and Lawson [2006] to determine IWC. The projected area input to the scheme was based on the area directly measured by the optical array probes. If the mass calculated by this method was greater than that of an ice sphere with maximum dimension Dj, the latter was used.

[27] Figure 5 compares 30 s averages of IWC derived from various SDs using both methods as a function of the bulk IWC measured by the CSI and Deep Cone Nevzorov probe. The different representations of the SDs used in the mass closure tests were as follows: (1) 2DS for N(50 < D < 300 μm), 2DC for N(300 < D < 800 μm), 2DP for N(D > 800 μm); (2) 2DS for N(50 < D < 300 μm), CIP2 for N(300 < D < 800 μm), 2DP for N(D > 800 μm); and (3) 2DS for N(50 < D < 800 μm), 2DP for N(D > 800 μm).

Figure 5.

The 30 s averages of IWC derived using (a) SD scheme 1, (b) SD scheme 2 and (c) SD scheme 3 as a function of IWC measured by either the CSI or deep-cone Nevzorov probe. Red points compare IWC computed with ‘BL06’ method with CSI IWCs, green compare ‘BL06’ method with Deep Cone IWC, blue compare ‘CPImD’ IWC with CSI IWC and yellow points compare ‘CPImD’ method with Deep Cone IWC. Colored lines denote best fit for each comparison and the black line is 1:1.

[28] The mean differences in IWC derived from SDs using the ‘CPI-mD’ using SDs schemes (1) and (3) were 23% and was 18% between schemes (2) and (3). These differences are within the measurement error of the bulk IWC probes, indicating that the mass closure tests cannot distinguish between the three representations of the SDs. Scheme (1) was therefore used to define N(D) due to the poorly defined depth of field of the 2DC and CIP2 for D < 300 μm. Because of the reasonable agreement between N2DC(300 < D < 800 μm) and NCIP2(300 < D < 800 μm), Scheme (2) was used for those times when the 2DC failed to record data. The mean difference between the CSI IWC and that derived from the scheme (2) SDs using the ‘BL06’ method was 103%, whereas it was only 46% for the ‘CPI-mD’ method. For scheme (1), these differences were 96% and 58%, respectively, again indicating better agreement between the ‘CPI-mD’ method and the bulk IWC than between the ‘BL06’ method and the bulk IWC. The average difference in bulk IWC and that related from the SDs is around a factor of 2. The lines of best fit for IWC derived from the SDs as a function of bulk IWC in Figure 5 also indicate that the ‘CPI-mD’ method gives values more consistent with the bulk IWC. For these reasons, the ‘CPI-mD’ method was used to estimate IWC from the SDs, and for deriving mass distribution functions.

[29] The following microphysical parameters were generated for all time periods: Nliq, Nice, Nice(D), the liquid drop SD Nliq(D), LWC, and IWC. The LWC was derived using

display math

where Dl is the median liquid drop diameter in size bin l with a width of ΔDl and assuming a density of water ρl = 1 g cm−3. Hansen and Travis [1974] defined the liquid drop effective radius rel as

display math

3.4. Technique for Comparing Cloud and Aerosol Properties

[30] During ISDAC, ramped ascents and descents of the NRC Convair-580 gave vertical profiles of cloud properties and provided information on aerosol properties above and below cloud. Constant altitude legs above, within, and below cloud gave additional information such as horizontal variability and statistically significant samples required for deriving IN concentrations. For 8, 18, and 26 April, a total of 41 total vertical profiles were obtained. For each vertical profile, aerosol concentrations above and below cloud were estimated from the PCASP (NPCASP). Quantities inside cloud were compared with NPCASP above or below cloud. Since the time required to obtain a statistically significant sample from the CFDC exceeds the length of time spent below and above cloud during the ascents and descents, the IN concentrations could not be derived in the same fashion. Therefore, relationships between IN and cloud properties were explored using data from constant altitude legs above cloud. For the purpose of this study, a statistically significant difference between profiles was the rejection of the null hypothesis that the median of two data sets is equal at a 95% significance level by a Mann–Whitney U test unless otherwise noted [Mann and Whitney, 1947]. For 41 data points, the probability distribution function of R between two random variables follows a t-distribution with 39 degrees of freedom giving a less than a 1% probability that the relationship between two data sets is due to random chance if |R| > 0.39.

4. Observed Relations Between Cloud and Aerosol Properties

4.1. Liquid Phase Aerosol-Cloud Interactions

[31] Figure 6 shows the mean T, NPCASP, LWC/Adiabatic LWC, specific humidity Qv, Nice, Nliq, LWC, and IWC as a function of normalized cloud altitude zn as

display math

with z representing altitude, zn = 0 corresponding to the liquid base at altitude zb, and zn = 1 corresponding to the cloud top at altitude zt. A threshold of Nliq > 1 cm−3 [Earle et al., 2011] was used to define the presence of liquid cloud. Values with zn < 0 represent altitudes below the liquid base which correspond to the presence of precipitating ice in mixed phase clouds. The adiabatic LWC was calculated assuming moist adiabatic ascent of a saturated parcel from cloud base. Strong capping inversions, typically present for arctic single-layer mixed phase stratocumulus [Tsay and Jayaweera, 1984], were also present for the ISDAC single-layer stratus as seen in Figure 6 for 0.8 < zn < 1.2. The inversion limited the ascent of updrafts in the cloud. Liquid occurring above the inversion is common in mixed-phase arctic stratus clouds and probably supported by moisture inversions near cloud top [Solomon et al., 2011; Sedlar et al., 2012]. However, in Figure 6, Qv decreases for zn > 0.8, demonstrating the lack of a moisture inversion and the presence of dry air near cloud top. The fact that the inversion begins to appear at zn ≈ 0.8 shows that dry air was possibly mixed in from above or direct condensation occurred by vertical circulations driven by radiative cooling [McFarquhar et al., 2011a; Solomon et al., 2011].

Figure 6.

Mean and standard deviation of (a) NPCASP, (b) T, (c) LWC/Adiabatic LWC, (d) Qv, (e) Nice(D > 50 μm), (f) IWC, (g) LWC, and (h) Nliq as a function of zn for all 41 vertical profiles in this study. The dashed line in (c) denotes 1:1 ratio of LWC/Adiabatic LWC.

[32] The subadiabatic profile of LWC, most evident for zn > 0.8 and zn < 0.3, could have resulted from entrainment of dry air from above cloud top and/or the growth of ice crystals at the expense of the liquid water. The NPCASP ranged from 50 to 250 cm−3 below cloud and from 150 to 1200 cm−3 above cloud. The strong correlation (R = 0.78) between Nliq and below cloud NPCASP in Figure 7c, the almost constant Nliq for 0.5 < zn < 0.8 and the increasing LWC with zn indicate that the aerosols measured in the accumulation mode by the PCASP below cloud were likely responsible for the nucleation of cloud droplets near base, in agreement with previous analysis of ISDAC data by McFarquhar et al. [2011a]. The much weaker and statistically insignificant correlation of R = 0.37 for Nliq and above cloud NPCASP in Figure 7a, and the decrease in specific humidity and increase of NPCASP for zn > 0.8 in Figure 6, again suggest that there was possibly either mixing of dry air above cloud top causing the evaporation of cloud drops or that some aerosols could have originated from above cloud top during mixing.

Figure 7.

Cloud mean Nliq and rel versus NPCASP (a and b) above cloud and (c and d) below cloud for the 41 different profiles flown on the 5 indicated sorties. Lines denote best fit of Nliq and rel as a function of NPCASP.

[33] If increasing aerosol concentrations resulted in lower droplet sizes, a necessary mechanism for the riming or thermodynamic indirect effect to occur, a negative correlation between rel and NPCASP would be expected. The correlation between rel and NPCASP below cloud shown in Figure 7d (R = −0.43) is much weaker than the correlation between Nliq and NPCASP below cloud. This is consistent with the findings of Earle et al. [2011], who showed increases in LWC and cloud thickness for liquid-dominated clouds with higher NPCASP below cloud base during ISDAC. LWC and NPCASP covary as pollution and humidity covary, which explains why rel was not robustly correlated with NPCASP above or below cloud. Since profile mean rel < 12 μm for all of the ISDAC vertical profiles, the dynamic range of rel is too limited to determine whether the thermodynamic indirect effect is operating for the conditions sampled during ISDAC.

[34] Figure 8 shows CCN concentrations as a function of supersaturation, averaged over constant altitude flight legs below cloud base for the five sorties. The ratio of CCN concentration to Nliq for supersaturations >0.15% averaged 1.44 for the first flight of 8 April, 1.72 for the second flight of 8 April, 2.27 for 18 April, 1.82 for the first flight of 26 April, and 1.41 for the second flight of 26 April. The fact that Nliq increases with height for zn < 0.3 and decreases with height for zn < 0.8 demonstrates that Nliq is not at its maximum value for all points in cloud. The CCN concentration below cloud would be expected to be equal to or greater than Nliq if CCN were being nucleated from cloud base, as the fraction of CCN that nucleate would depend on the supersaturation of the ascending parcel.

Figure 8.

Mean CCN below cloud as a function of supersaturation from all constant altitude legs flown below cloud. Solid bars represent ±1 standard deviation about the mean value.

4.2. Ice Phase Aerosol-Cloud Interactions

[35] Figure 9 shows vertical cross sections of radar reflectivity factor Ze derived from the NA W-band radar on 8 April along with relevant cloud and aerosol properties determined from in situ data on co-located constant altitude legs and porpoising legs. A single liquid layer of variable thickness and cloud top height, with precipitating ice throughout and below cloud was observed. The horizontal variability in Ze (Figure 9) and Doppler velocity (shown in McFarquhar et al. [2011a]) indicates the presence of inhomogeneities in cloud structure associated with vertical circulations driven by cloud top radiative cooling [Solomon et al., 2011], which no doubt limited the correlations between cloud and aerosol parameters in the vertical profiles. The close coupling of upward and downward velocities with widths of about 1 km, indicates the potential role of turbulent mixing in the cloud fields [cf. Korolev and Field, 2008; Shupe et al., 2008; McFarquhar et al., 2011a].

Figure 9.

Vertical cross section of Ze from the W-band radar for a cloud deck observed on the second flight of April 8. The blue shaded regions denote the approximate location of the liquid layer derived from the in situ profiles of LWC. Maroon values denote PCASP concentration measured above and below cloud, black values in mm are median mass diameter of ice crystals, and values in L−1 denote Nice(D > 50 μm). Values in °C denote temperature. The solid black line denotes flight track altitude. The dashed black line denotes the approximate location of the temperature inversion.

[36] The in situ profiles show that large ice crystals occurred throughout the cloud, with some reduction in Nice and IWC near cloud top. Because of the potential role of vertical mixing and turbulence in distributing ice throughout the cloud, no definitive statement can be made on where the ice originated. Nevertheless, the presence of cloud over the top of the capping inversion, the subadiabatic LWC and gradual reduction in Qv for zn > 0.8 in Figure 6, and the close coupling of updrafts and downdrafts and associated turbulence are all consistent with the possibility of dry air mixing into the cloud from above the top.

[37] To further examine how aerosols affect ice cloud properties, the relationships between ice cloud microphysics and aerosol concentrations above and below cloud were examined. Figure 10 compares the mean Nice(D ≥ 50 μm) inside cloud with the mean NPCASP above and below cloud. Despite the scatter in the data, there is a stronger correlation of Nice(D ≥ 50 μm) with NPCASP above cloud (R = 0.69) than with NPCASP below cloud (R = 0.37). When removing the 18 April case, which has much higher NPCASP, colder temperatures, different synoptic forcing, surface conditions, and aerosol composition [e.g., McFarquhar et al., 2011a; Brock et al., 2010], the correlation of Nice(D ≥ 50 μm) with NPCASP above cloud was even stronger (R = 0.75). These correlations are consistent with the observed subadiabatic LWC profiles near cloud top and the close coupling of updrafts and downdrafts within cloud, and further support the possibility of entrainment of aerosol from above cloud. However, a possible uncertainty is that the dynamic range of Nice(D ≥ 50 μm) was on the order of uncertainty of calculated Nice(D ≥ 50 μm) indicated in Figure 4, especially for 8 and 26 April. Observations over a wider range of Nice and NPCASP, as well as improved measurements of Nice are therefore required to test the generality of this result.

Figure 10.

Cloud mean Nice(D ≥ 50 μm) versus mean NPCASP (a) above cloud and (b) below cloud for all 41 vertical profiles on the 5 sorties indicated. Lines denote best fit of Nice(D ≥ 50 μm) as a function of mean NPCASP.

[38] Lance et al. [2011] found a direct relationship between Nice and Nliq(30 < D < 50 μm), and greater Nliq(30 < D < 50 μm) in cleaner clouds compared to polluted clouds. Therefore, Figure 11 compares Nliq(30 < D < 50 μm) with NPCASP below cloud and Nice(D ≥ 50 μm) with Nliq(30 < D < 50 μm). No correlation (R = −0.16) between Nliq(30 < D < 50 μm) and NPCASP below cloud or between Nice(D ≥ 50 μm) and Nliq(30 < D < 50 μm) (R = 0.20) is noted. The lack of correlation between Nice(D ≥ 50 μm) and Nliq(30 < D < 50 μm) is likely due to the fact that very pristine conditions were not encountered and hence rel never exceeded the threshold of 12 μm necessary for the secondary multiplication processes [Rangno and Hobbs, 1991, 2001; Lance et al., 2011]. The lack of correlation between Nliq(30 < D < 50 μm) and NPCASP below cloud is again consistent with the lack of correlation between rel and NPCASP due to changes in cloud LWC with NPCASP [Earle et al., 2011]. When 18 April is removed from the analysis, the correlation between Nliq(30 < D < 50 μm) and NPCASP below cloud is R = −0.01 and R = 0.55 between Nice(D ≥ 50 μm) and Nliq(30 < D < 50 μm) probably not due to aerosol below cloud.

Figure 11.

As in Figure 10, except (a) cloud mean Nice(D ≥ 50 μm) versus Nliq(30 ≤ D ≤ 50 μm) and (b) Nliq(30 ≤ D ≤ 50 μm) versus NPCASP below cloud on the 5 indicated sorties.

[39] The positive correlation between Nice(D ≥ 50 μm) and NPCASP above cloud would be expected if IN concentration increased with NPCASP and if IN were entrained from above cloud top, causing a glaciation indirect effect. Figure 12 shows the CFDC operating temperature, TCFDC, ambient temperature, the supersaturation with respect to water within the CFDC, SSw, and the IN concentration from the CFDC as a function of NPCASP for all constant altitude legs above cloud base for the five sorties. The IN were sampled at temperatures as much as 15°C lower than the observed cloud top temperatures, so it was not possible to determine which primary nucleation mechanism, if any, dominated. For observations on 8 and 26 April, representing 90% of the vertical profiles considered in this analysis, NPCASP above cloud was less than 400 cm−3, with TCFDC ranging from −25°C to −20°C. An increase in IN concentration with NPCASP above cloud noted on 8 and 26 April for cases with NPCASP < 400 cm−3 could explain the positive correlation between Nice and NPCASP seen in Figure 10 if IN are entrained at cloud top for the 8 and 26 April cases. However, NPCASP was greater than 400 cm−3 for all cases on 18 April with a nearly constant TCFDC of −30°C. The data collected on 18 April were at temperatures up to 10°C lower than those on 8 and 26 April, the aerosol composition was vastly different [Warneke et al., 2009; Earle et al., 2011], and the synoptic regime and surface conditions differed as well, making it difficult to compare trends observed on 18 April with trends observed on other days.

Figure 12.

(a) Mean IN concentration from the CFDC as a function of NPCASP for bins of width 30 cm−3, (b) CFDC operating temperature (TCFDC), ambient temperature and (c) CFDC recorded supersaturation with respect to water (SSw). All quantities derived from constant altitude legs above cloud. Vertical bars denote 1 standard deviation above and below the mean.

[40] In addition to IN concentration and composition, ice cloud microphysics can depend on meteorological conditions, surface fluxes, dynamical forcing, and other factors. Since the thermodynamic and riming indirect effects depend on increases in NPCASP leading to decreases in rel, differences in LWP need to be considered when examining indirect effects. Figure 13 shows the mean Nliq(D) and Nice(D) for all vertical profiles divided according to whether the above cloud NPCASP was ≤ or >200 cm−3, and according to whether the LWP was ≤ or > than 16 g m−2, the median LWP. The error bars were determined using the bootstrap technique [Efron and Tibshirani, 1993; McFarquhar and Heymsfield, 1997], in which alternate versions of the average SDs were generated by randomly drawing, with replacement, from the population of SDs meeting the stated criteria. The standard deviations in Nliq(D) and Nice(D) from the alternate average SDs are displayed as error bars. Table 2 shows the mean ratio of N(D) in above cloud NPCASP > 200 cm−3 compared to NPCASP ≤ 200 cm−3 for differing LWP. There were consistently higher ice concentrations for more the polluted clouds, with greater differences for the clouds with larger LWP.

Figure 13.

(a) Mean Nliq(D) and Nice(D) for vertical cloud profiles for differing ranges of NPCASP above cloud for clouds with LWP < 16 g m−2 and (b) LWP ≥ 16 g m−2. Vertical bars represent standard deviations of N(D) produced by the bootstrap technique.

Table 2. Ratio of Quantity When NPCASP Above Cloud >200 cm−3 to Quantity When NPCASP Below Cloud ≤200 cm−3 for Varying LWP
LWPNice(0.05 < D < 1 mm)Nice(D ≥ 1 mm)Nliq(30 < D < 50 μm)
≤16 g m−2 (all cases)4.151.062.26
>16 g m−2 (all cases)2.014.5221.8
≤16 g m−2 (8 + 26 April)1.992.425.61
>16 g m−2 (8 + 26 April)2.004.5219.0

[41] Figure 14 shows the mean Nliq(D) and Nice(D) for all vertical profiles divided according to whether the below cloud NPCASP was ≤ or >200 cm−3, and according to whether the LWP was ≤ or >16 g m−2. The error bars were produced in the same manner as for Figure 13. Table 3 shows the mean ratio of N(D) in below cloud NPCASP > 200 cm−3 compared to NPCASP ≤ 200 cm−3 for differing LWP. This indicates a reduction in the number of large drops for NPCASP > 200 cm−3 that is more prominent for the thinner clouds. Figure 14 and Table 3 show no consistent trend of Nice(D) versus NPCASP.

Figure 14.

(a) Mean Nliq(D) and Nice(D) for vertical cloud profiles for differing ranges of NPCASP below cloud for clouds with LWP < 16 g m−2 and (b) LWP ≥ 16 g m−2. Vertical bars represent standard deviations of N(D) produced by the bootstrap technique.

Table 3. Ratio of Quantity When NPCASP Above Cloud >200 cm−3 to Quantity When NPCASP Above Cloud ≤200 cm−3 for Varying LWP
LWPNice(0.05 < D < 1 mm)Nice(D ≥ 1 mm)Nliq(30 < D < 50 μm)
≤16 g m−2 (all cases)4.360.010
>16 g m−2 (all cases)0.810.470.94
≤16 g m−2 (8 + 26 April)No dataNo dataNo data
>16 g m−2 (8 + 26 April)0.830.541.13

4.3. Examination of Riming Indirect Effects During ISDAC

[42] To examine the potential impact of the riming indirect effect, Figure 15 compares the mean IWC inside cloud to the mean NPCASP above and below cloud for the 41 vertical profiles flown on three days with single-layer clouds during ISDAC. The lack of correlation between IWC and either the above or below cloud NPCASP (|R| < 0.39) suggests that the riming indirect effect did not impact the IWC. Rimed particles were imaged by the CPI on only one of the five sorties – the first flight of 26 April, and even then for only 5.6% of the ice crystal images with D > 50 μm. The mean and standard deviation of LWC on that particular day were 0.13 ± 0.09 g m−3, greater than that observed on other days (0.11 ± 0.09 g m−3 on the first flight of 8 April, 0.09 ± 0.07 g m−3 on the second flight of 8 April, 0.07 ± 0.03 g m−3 on 18 April, and 0.08 ± 0.06 g m−3 on the second flight of 26 April), suggesting why some riming was observed on this particular sortie, and not on the others. However, on the first flight of 26 April, NPCASP below and above cloud varied between only 170 and 210 cm−3, which did not provide a large enough range of aerosol concentrations for investigating the riming indirect effect. Thus, although the riming indirect effect did not have an important impact on the IWC for the single-layer cases sampled during ISDAC, its role still needs to be tested in clouds with sufficient LWC over a wider range of aerosol loadings.

Figure 15.

(a) Mean IWC in cloud versus mean NPCASP above cloud and (b) below cloud for all vertical profiles of clouds.

4.4. Comparison of ISDAC and M-PACE Cases

[43] There was a limited dynamical range of aerosol concentrations sampled during ISDAC, limiting investigations of aerosol indirect effects. For further investigation of aerosol indirect effects on single-layer arctic stratocumulus, four sorties conducted during M-PACE are compared against the single-layer ISDAC cases. The M-PACE cases used were one flight on 9 October 2004, two flights on 10 October 2004, and one flight on 12 October 2004. The surface, aerosol, and meteorological conditions for these cases have been summarized by Verlinde et al. [2007]. McFarquhar et al. [2007] details the instrumentation installed on the University of North Dakota Citation for M-PACE and the methodology by which cloud microphysical parameters were derived.

[44] To investigate potential ice nucleation and multiplication mechanisms, the clouds sampled during M-PACE and ISDAC were classified according to the scheme of Rangno and Hobbs [2001], which differentiates between five different cloud categories based on Nliq, Nice, rel, maximum threshold diameter Dt, and cloud top temperature Ttop:

display math
display math
display math
display math
display math

[45] Figures 16 and 17 show normalized frequency histograms of Nice(D ≥ 125 μm), Nliq, rel, Ttop, LWC, and IWC for the single-layer M-PACE and ISDAC cases. The M-PACE cases did not neatly fall into a single category. The M-PACE had rel consistent with category V and Nice consistent with category IV, while the ISDAC clouds fell into either category IV or, for 50% of the cases on the first flight of 26 April, category I. The rel categorizations indicate that secondary multiplication mechanisms such as rime-splintering, shattering of freezing drops, or contact ice nucleation of drizzle drops would likely occur [Rangno and Hobbs, 2001; Rangno and Hobbs, 1991] in the M-PACE cases, but not in the ISDAC cases. The Hallet-Mossop secondary ice crystal production process, operating at temperatures between −7°C and −4°C [Hallett and Mossop, 1974], probably did not operate because temperatures inside cloud exceeded −8°C only 3.9% of the time for ISDAC and 0% of the time for M-PACE.

Figure 16.

Normalized frequency histograms of (a) Nice(D ≥ 125 μm), (b) Nliq, and (c) rel, from the ISDAC single-layer cases and (d) Nice(D ≥ 125 μm), (e) Nliq, and (f) rel from the M-PACE single layer cases.

Figure 17.

Normalized frequency histograms of (a) Ttop, (b) LWC, and (c) IWC from the ISDAC single-layer cases and (d) of Ttop, (e) LWC, and (f) IWC from the M-PACE single layer cases.

[46] Arctic aerosol conditions in October are generally clean, with April being a transition season between polluted and clean conditions [Quinn et al., 2002; Garrett and Zhao, 2006]. This suggests that clouds sampled during M-PACE occurred in cleaner conditions than those sampled during ISDAC. However, because there were no direct measurements of aerosol concentration during M-PACE, the relationship between cloud parameters and aerosol concentration cannot be directly examined from these data. However, the CFDC recorded IN concentrations under the noise level 91% of the time for the M-PACE single layer cases, compared to 53% of the time for the ISDAC single-layer cases, consistent with cleaner conditions being prevalent during M-PACE. Thus, frequency distributions of cloud properties can be compared for the cleaner M-PACE and the more polluted ISDAC conditions. Figures 16 and 17 show these comparisons. In these figures, ice crystal concentrations with D > 125 μm are plotted because there were no 2D-S measurements acquired during M-PACE, and hence, the concentrations of crystals with 50 μm < D < 125 μm are highly uncertain.

[47] Figure 16 shows that the mean and standard deviation of Nice(D ≥ 125 μm) of 2.52 ± 6.45 L−1 during M-PACE were higher than those of 0.27 ± 0.26 L−1 during ISDAC. The mean and standard deviation of rel were 9.49 ± 2.30 μm for M-PACE and 4.97 ± 3.17 μm for ISDAC. The mean and standard deviation of Nliq were 46 ± 30 cm−3 and 93 ± 81 cm−3 for the M-PACE and ISDAC cases, respectively. A Mann-Whitney U test at a 95% significance level showed that these differences are statistically significant. These results are consistent with the reduced aerosol and CCN concentrations during M-PACE being associated with higher rel, providing conditions more favorable for secondary ice crystal production via shattering of freezing drops, contact nucleation of drizzle drops, shattering of freezing drops, and/or fragmentation of ice crystals via droplet-crystal and crystal-crystal collisions [Rangno and Hobbs, 1991, 2001]. This is also consistent with the finding of Lance et al. [2011] that cleaner conditions during ARCPAC were associated with enhanced ice crystal concentrations, but contradictory to the increase in Nice(D ≥ 50 μm) with increasing NPCASP above cloud shown in Figure 10. Since it is likely that the liquid drops are nucleating from CCN below cloud, it is also possible that if LWC is held constant, then CCN below cloud, rather than IN above cloud, are related to Nice in a manner consistent with the thermodynamic indirect effect. However, because there was considerable variability in surface and meteorological conditions and aerosol compositions in the limited number of cases considered. More in situ and remote sensing studies of aerosol and single-layer arctic mixed-phase clouds are required to make definitive statements about how the concentrations of aerosols, IN, and CCN affect cloud properties.

[48] There were other factors that might cause the variations in Nice(D ≥ 125 μm) and IWC. Figure 17 shows that the mean and standard deviation of LWC was 0.19 ± 0.12 g m−3 for M-PACE and 0.10 ± 0.13 g m−3 for ISDAC, a statistically significant difference according to a Mann-Whitney U test at a 95% significance level. This difference is consistent with increased surface moisture fluxes from the wider availability of open water during the fall compared to spring. Modeling studies initialized with combinations of ISDAC and M-PACE surface and aerosol conditions, and evaluated against observations are required to isolate the roles of aerosol and surface forcing in determining arctic stratus properties.

5. Summary

[49] Cloud and aerosol properties measured above, within, and below single-layer mixed-phase stratus and stratocumulus during ISDAC and M-PACE were used to assess the roles of three aerosol indirect effects; namely, the glaciation indirect effect, the riming indirect effect, and the thermodynamic indirect effect. The variation of ice crystal concentration Nice, liquid drop concentration Nliq, liquid drop effective radius rel, liquid water content LWC, and ice water content IWC with ice nuclei IN, cloud condensation nuclei CCN, and total accumulation mode aerosol concentration NPCASP above and below cloud were examined to determine the relative importance of the indirect effects. The main conclusions of this study are as follows:

[50] 1. A best estimate of cloud microphysical properties (rel, Nice(D), Nliq(D), IWC, LWC) was determined using data from the myriad cloud probes installed on the NRC Convair-580 during ISDAC. By comparing size distributions from different probes under varying conditions, assessing post flight calibrations, and conducting mass closure tests, it was determined that a combination of the CDP, 2DS, CIP/2DC and 2DP best represented the size distributions. A comparison of the bulk IWC measured by the deep-cone Nevzorov probe and a Cloud Spectrometer and Impactor against that derived from the size distributions showed that IWC calculated using m-D relationships [Brown and Francis, 1995; Mitchell, 1996] weighted according to size-resolved habit distributions provided a better estimate of IWC than that derived using the relation between mass and crystal morphology of Baker and Lawson [2006] for these cases. The selection of probes used here should be considered as one possible combination that gives a reasonable estimate of N(D) for ISDAC. There may be different combinations of probes that give equally good estimates of N(D).

[51] 2. The liquid drop number concentration Nliq was well correlated (R = 0.78) with NPCASP below the liquid cloud base, and weakly correlated (R = 0.37) with NPCASP above the cloud. These correlations, the nearly constant vertical profile of Nliq,, and the increases of LWC with height above cloud base indicates that cloud droplets were nucleated near cloud base and grew as they ascended through cloud. The liquid drop effective radius rel was not as well correlated with NPCASP below (R = −0.43) and above cloud (R = −0.24) as was Nliq due to increases in LWC and cloud thickness with increasing NPCASP.

[52] 3. The Nice(D ≥ 50 μm) was well correlated (R = 0.69) with NPCASP above cloud top. This correlation was present regardless of the thickness of the sampled clouds. The Nice(D ≥ 50 μm) was not as well correlated (R = 0.37) with NPCASP below the liquid cloud base. These trends could be explained by increases in IN with NPCASP above cloud for 8 and 26 April. The increased Nice(D ≥ 50 μm) yet decrease in IN on 18 April may have been caused by the colder temperatures on that day. The correlation between Nice(D ≥ 50 μm) and NPCASP above cloud top, the subadiabatic profile of LWC, and the presence of turbulence induced by the close coupling of updrafts and downdrafts are consistent with the possibility of aerosols in dry air entrained from above cloud affecting the ice crystal concentration through the glaciation indirect effect for 8 and 26 April.

[53] 4. IWC was not correlated with NPCASP above (R = −0.21) or below cloud (R = −0.32) for ISDAC. The lack of correlation between IWC and NPCASP and the lack of rimed particles on 4 of the 5 sorties imply the riming indirect effect did not impact the cloud microphysical properties for the conditions observed during ISDAC.

[54] 5. The properties of single-layer mixed phase clouds sampled in more pristine conditions during M-PACE were compared against those sampled in more polluted conditions during ISDAC. Means and standard deviations of Nice(D > 125 μm) were 2.52 ± 6.45 L−1 and 0.27 ± 0.26 L−1 for M-PACE and ISDAC, respectively. Means and standard deviations of Nliq were 46 ± 30 cm−3 and 93 ± 81 cm−3 for M-PACE and ISDAC, respectively. The increase in Nice and decrease in Nl in the cleaner conditions is consistent with the action of the thermodynamic indirect effect, but is different than that shown by the strong correlation of Nice(D ≥ 50 μm) with NPCASP above cloud top. It should also be noted that the mean and standard deviation of LWC were 0.19 ± 0.12 g m−3 for M-PACE and 0.10 ± 0.13 g m−3 for ISDAC, with the increased LWC in M-PACE consistent with the presence of open water during the fall. This variation in LWC with varying surface conditions could also have affected the cloud properties and suggests that the dominant aerosol effect may depend on the range of aerosol, surface and meteorological conditions sampled.

[55] With the information available, it was not possible to categorically state how much the variation in aerosol conditions during ISDAC and M-PACE contributed to the difference in cloud conditions, as opposed to differences in surface fluxes or meteorology. Nevertheless, there were indications that the mixing of aerosols at the tops of clouds may cause a glaciation indirect effect in some circumstances. In other circumstances, aerosols may cause a thermodynamic indirect effect. There was no evidence of a major role for the riming indirect effect, but observations over a wider range of conditions are required to show the generality of this finding. Future modeling studies, initialized with combinations of ISDAC and M-PACE surface and aerosol conditions, and evaluated against observations should also be performed to isolate the roles of aerosol and surface forcing.

Acknowledgments

[56] This work was supported by the Office of Biological and Environmental Research (BER) of the U.S. Department of Energy (DE-FG02-02ER63337, DE-FG02-07ER64378, DE-FG02-09ER64770 and DE-SC0001279) as part of the Atmospheric Systems Research and Atmospheric Radiation Measurement (ARM) Airborne Facilities. Data were obtained from the ARM program archive, sponsored by the U.S. DOE, Office of Science, BER, Environmental Sciences Division, as well as from Environment Canada. The authors acknowledge Kenny Bae for providing preliminary versions of some processing algorithms for this study. The authors also acknowledge Walter Strapp at Environment Canada for his aid in the CDP and FSSP calibrations, as well as for providing the cloud microphysical data and feedback on this study. The authors would like to acknowledge Paul DeMott and three anonymous reviewers for their feedback on this study.

Advertisement