Abstract

Supercooled large drops (SLD) can be a significant hazard for aviation. Past studies have shown that warm-rain processes are prevalent, or even dominant, in stratiform clouds containing SLD, but the primary factors that control SLD production are still not well understood. Giant aerosol particles have been shown to accelerate the formation of the first drizzle drops in some clouds and thus are a viable source of SLD, but observational support for testing their effectiveness in supercooled stratiform clouds has been lacking. In this study, new observations collected during six research flights from the Alliance Icing Research Study II (AIRS II) are analyzed to assess the factors that may be relevant to SLD formation, with a particular emphasis on the importance of giant aerosol particles. An initial comparison of observed giant aerosol particle number concentrations with the observed SLD suggests that they were present in sufficient numbers to be the source of the SLD. However, microphysical calculations within an adiabatic parcel model, initialized with the observed aerosol distributions and cloud properties, suggest that the giant aerosol particles were only a limited source of SLD. More SLD was produced in the modeled clouds with low droplet concentrations, simply by an efficient warm-rain process acting at temperatures below 0°C. For cases in which the warm-rain process is limited by a higher droplet concentration and small cloud depth/liquid water content, the giant aerosol particles were then the only source of SLD. The modeling results are consistent with the observed trends in SLD across the six AIRS II cases.

1. Introduction

Sand et al. (1984) describe the 2 most hazardous flights of 100 research-aircraft icing encounters as having involved ice accretion on the underside of the wings, associated with the collection of large supercooled drops having diameters between 40 and 300 μm. Most icing prevention mechanisms on aircraft protect the leading edges of the wings, because smaller supercooled cloud droplets tend to collide there or are otherwise swept around and avoid impact with the wing. However, larger supercooled drops may deviate from the airflow around the wings because of their greater inertia, be collected along the top or underside of the wings where icing prevention mechanisms are lacking, and accumulate on the aircraft, sometimes with disastrous results (e.g., Marwitz et al. 1997).

Politovich (1989) coined the term supercooled large drops (hereinafter SLD), bringing attention to those drops with diameters greater than 50 μm that contribute to this specific aircraft icing hazard. She demonstrated that traditional assessments of icing severity based on cloud conditions, such as supercooled liquid water content (LWC) and mean volume droplet diameter, were inadequate to represent the hazard to aircraft when SLD were present, even when SLD number concentrations were as low as 10 L−1 for exposure times as short as 10–15 min. The lower limit on SLD concentrations that is hazardous to aircraft is currently unknown.

Two general scenarios have been identified for creating SLD. In the first scenario, frozen hydrometeors can completely melt as they fall through a warm (temperature greater than 0°C) layer in the atmosphere, then subsequently “supercool” (cool to temperatures less than 0°C but remain liquid) as they fall through a colder layer below. Multiple field studies have reported a second scenario, whereby SLD are formed by a “supercooled warm rain process” (Huffman and Norman 1988), that is, condensation upon favorable nuclei in the atmosphere, followed by collision and coalescence of liquid droplets to create drizzle-sized particles, all occurring over a depth of the atmosphere having temperatures that are less than 0°C. Some observational studies have reported that as many as 75% of freezing-precipitation events may form in this kind of environment (Strapp et al. 1996; Cober et al. 2001b).

The factors controlling SLD formation by the warm-rain process are not well understood. Multiple hypotheses have emerged to explain SLD formation in stratiform wintertime clouds, and it is possible that some factors act in some cases but not in others. Song and Marwitz (1989) noted that the SLD cases reported by Politovich (1989) had very low droplet concentrations, like those found in maritime stratiform clouds, suggesting that SLD formed by a very efficient warm-rain process (fewer drops compete less for the available vapor and grow much more quickly to sizes capable of initiating collisions and coalescence, given enough time). Several additional studies have noted low droplet concentrations observed in clouds containing SLD (Rasmussen et al. 1995; Murakami et al. 1992; Bernstein et al. 2004; Ikeda et al. 2007), as well as larger cloud liquid water contents that would also accelerate SLD formation. Pobanz et al. (1994) hypothesized that shear-induced mixing at the cloud top and the associated entrainment could accelerate large drop growth (by reducing the local droplet number, allowing those drops remaining to grow with limited competition). Korolev and Isaac (2000) hypothesized that isobaric mixing across a temperature inversion could create very high supersaturations that could accelerate the growth of smaller droplets into SLD.

In the current study, another possible source of SLD is investigated: giant aerosol particles. Numerous studies [see the review by Beard and Ochs (1993) and references therein, Lasher-Trapp et al. (2001), Blyth et al. (2003), etc.] have suggested that giant and/or ultragiant aerosol particles, having radii greater than 1 and 10 μm, respectively, can be important for the first drizzle formation in warm cumulus clouds. Other numerical modeling studies of stratocumulus and stable stratus clouds have found similar results (Feingold et al. 1999; Geresdi and Rasmussen 2005). These particles, because of their substantially larger size than the more prevalent cloud condensation nuclei (CCN), need less time to grow by condensation to sizes at which they can initiate collisions and coalescence, proceeding very rapidly to form drizzle.

Novel observations of giant aerosol particles (GA) were collected during SLD events within the larger framework of the Alliance Icing Research Study II (AIRS II), an international field campaign based in Mirabel, Quebec, Canada, during winter 2003/04, designed to improve understanding and forecasting of aircraft icing (Isaac et al. 2005). These observations are used in the current study to investigate the importance of GA to SLD formation during six of the AIRS II flights.

2. Data analysis

All data used in this study were collected with the National Center for Atmospheric Research (NCAR) C-130 aircraft during AIRS II. Flights were primarily over the eastern Great Lakes region of the United States and Canada, a climatologically favored region for aircraft-icing events, and specifically SLD events (Bernstein et al. 2007). Of the 14 C-130 flights during AIRS II, 6 had atmospheric soundings consistent with a supercooled warm-rain process and at least one cloud that contained SLD (i.e., SLD formation in the absence of a melting layer above or within the cloud). These cases also had CCN, GA, and supercooled cloud sampling in the same air mass and were screened [using the data from the Scanning Aerosol Backscatter Lidar (SABL) mounted on the C-130 aircraft] to eliminate any cases that may have been “seeded” by upper-level cloud decks.

Satellite images and SABL data were used to investigate the appearance, evolution, and structure of the clouds. The clouds containing SLD could be classified as stratocumulus (Fig. 1), except for those occurring at higher altitudes that might be described as altocumulus stratiformis. They were often “patchy,” with intermittent, nearly clear areas evident. Some of the clouds in Fig. 1 appear very thin and, one might assume, benign, but all of these clouds contained SLD. Most of the cloud depths were less than 1 km, and often were 600 m or less, except for the 1 December case that had convective elements stretching nearly 2 km upward from the base.

Fig. 1.

Moderate-Resolution Imaging Spectroradiometer (MODIS) visible images during each of the six flights: (a) 5 Nov, (b) 11 Nov, (c) 25 Nov, (d) 1 Dec, (e) 3 Dec, and (f) 4 Dec 2003. Time in UTC is labeled at the top of each image, which was the nearest image to the time of cloud sampling listed in Table 1. Sampling area is denoted on each image by a dark circle.

Fig. 1.

Moderate-Resolution Imaging Spectroradiometer (MODIS) visible images during each of the six flights: (a) 5 Nov, (b) 11 Nov, (c) 25 Nov, (d) 1 Dec, (e) 3 Dec, and (f) 4 Dec 2003. Time in UTC is labeled at the top of each image, which was the nearest image to the time of cloud sampling listed in Table 1. Sampling area is denoted on each image by a dark circle.

a. In situ cloud and environment observations

On each day of operations, the C-130 aircraft conducted clear-air sampling at numerous altitudes in the vicinity of the clouds to sample the ambient aerosol particles. The aircraft was then directed into the stratiform clouds to discern the amount of supercooled water present, if any. The clouds were so shallow that all penetrations of a cloud were conducted at the same altitude, but the horizontal extent of the clouds and multiple passes allowed for long in-cloud samples. The cloud sampling was preferentially over large bodies of water, so that the aircraft could sample at and beneath the bases to assess the aerosol conditions there. All clouds penetrated on each day of operations were scrutinized for evidence of SLD (discussed in section 2c and the appendix), and those clouds containing SLD were used in this study.

Some environmental and cloud conditions for each cloud containing SLD on the six flights are summarized in Table 1. The maxima and averages were calculated during that part of the penetration when SLD were observed (time listed in the first column) at the penetration altitude (third column). The CCN number concentrations correspond to the values measured in clear air at the penetration altitude, taken from the total vertical profile of CCN measured at 1% supersaturation from the Desert Research Institute (DRI) CCN spectrometer (Hudson 1989) over multiple altitudes during each flight.

Table 1.

Observed thermodynamic and microphysical conditions within the supercooled stratiform clouds, or their environment, for six flights during the AIRS II field campaign.

Observed thermodynamic and microphysical conditions within the supercooled stratiform clouds, or their environment, for six flights during the AIRS II field campaign.
Observed thermodynamic and microphysical conditions within the supercooled stratiform clouds, or their environment, for six flights during the AIRS II field campaign.

The environmental characteristics for these clouds containing SLD varied considerably. The altitudes at which SLD were encountered varied from 1 to nearly 5 km MSL. On a given day, the lower-altitude clouds were more likely to have SLD (higher-altitude clouds often contained no SLD, probably because of the very cold temperatures resulting in more active ice nuclei, and lower liquid water contents), and thus most of the clouds in Table 1 occur at altitudes below 2 km MSL. The average observed in-cloud temperature among the cases varied only from −3° to −11°C, except for one cloud having SLD at −19°C. This narrow temperature range likely resulted from these cases having been selected to eliminate the possibility of a melting layer contributing to SLD (limiting the higher end of the range) and the temperature dependence of ice nuclei (limiting the lower end of the range). Water vapor mixing ratios among the cases varied from 1 to nearly 5 g kg−1, indicative of both the variation in altitude of the clouds and the drier conditions experienced on some days.

The duration of each icing event (first column) ranged from 4 to nearly 50 min. The intensity of the icing is represented using the cycling rate of the Rosemount Icing Detector (Baumgardner and Rodi 1989; Cober et al. 2001a), where higher rates are indicative of more rapid ice buildup on the rod that is exposed to the airstream. Cycling rates ranged from 0.2 to 2.6 min−1 over the clouds presented here; those with rates exceeding 1 min−1 fall into the range of cases investigated by Politovich (1989) that were considered to be hazardous to aircraft. Considering both the Rosemount probe cycling rate and the length of time during which the aircraft encountered SLD, the 5 and 11 November cases would be considered the most severe.

The maximum and average updraft speeds within the clouds at the penetration altitude also showed considerable variability, with some convective elements within the clouds reaching updraft speeds of 6–7 m s−1. Across a cloud penetration, the updrafts and downdrafts were often of equal magnitude, resulting in average speeds close to zero and indicative of the cellular structure that tended to exist in these clouds (Fig. 2).

Fig. 2.

Time series (1855–1905 UTC) of observed (top) FSSP number concentration and (bottom) vertical velocity for the 3 Dec 2003 cloud, showing only a portion of the cloud to highlight the structural detail in the cloud. At an aircraft speed of approximately 110 m s−1, 1 min corresponds to approximately 6.6 km.

Fig. 2.

Time series (1855–1905 UTC) of observed (top) FSSP number concentration and (bottom) vertical velocity for the 3 Dec 2003 cloud, showing only a portion of the cloud to highlight the structural detail in the cloud. At an aircraft speed of approximately 110 m s−1, 1 min corresponds to approximately 6.6 km.

Cloud LWC observed at the penetration altitude with a Commonwealth Scientific and Industrial Research Organisation (CSIRO) probe (King et al. 1978) ranged from 0.1 up to 0.6 g m−3. This probe underestimates the LWC when drops larger than ∼20 μm in diameter are encountered (Biter et al. 1987) and is not designed to operate in mixed-phase clouds. The magnitudes are thus questionable, although very near the calculated adiabatic values. Liquid water, ice water, and total water measurements from new instrument prototypes flown on the C-130 are being analyzed and will be presented in a future study (J. Hallett 2007, personal communication).

Droplet number concentrations within the clouds were estimated with a forward scattering spectrometer probe1 (FSSP) that detects and sizes particles in the range of 2–45-μm in diameter. The measured number concentration can be artificially enhanced in regions where ice is present (Gardiner and Hallett 1985), and so additional information must be used to discern the amount of unfrozen drops present. In most clouds, the cloud droplet number concentration could be established with a high level of confidence with the FSSP data, because images from the Cloud Particle Imager2 (CPI; Lawson and Cormack 1995) could help to identify sections in which only liquid particles were present. (An exception is the cloud sampled on 11 November, for which no such sections could be found with the CPI, and thus droplet concentration was estimated from a section of the cloud where the FSSP-derived LWC was nearly the same as that measured by the CSIRO probe.) The FSSP concentrations across a cloud penetration (Fig. 2) often exhibited the patchiness evident in many of the satellite pictures and corresponded with the cellular structure of the updrafts. For several cases, the cloud droplet number concentrations were very low, less than 200 cm−3, whereas others had higher values, exceeding 500 cm−3. Because the number of droplets in the cloud depends on the ambient aerosol—namely, the CCN—the variability in the droplet number concentrations is consistent with the observed CCN number concentrations measured in clear air at the penetration altitude at a supersaturation of 1% within the instrument (Table 1, last column).

b. Clear-air giant aerosol observations

1) Number concentrations

The GA in the clear air were estimated following the technique discussed in Lasher-Trapp and Stachnik (2007). The number concentrations of GA (and ultragiant particles) detected in the clear air by the FSSP and 260X probes were combined for each altitude flown, resulting in a vertical profile of GA number concentration representative of the sampled air mass for each flight.3 These number concentrations ranged from ∼0.3 to over 100 L−1 over the six flights and tended to decrease exponentially with altitude as found over all of the AIRS II flights by Lasher-Trapp and Stachnik (2007). The peak in the size distributions of the GA was typically at 2–5 μm in diameter. Ultragiant aerosol particles were detected by the 260X probe at the lowest altitudes on some flights; particles collected by the CPI and by impaction on glass slides (with a much larger sample volume) verified that these particles were present.

2) Comparison with CCN concentrations

Because some past studies have suggested that SLD may be correlated with lower droplet number concentrations in the clouds (inherently resulting from fewer CCN in the ambient air) and this mechanism might compete with the formation of SLD upon the GA, observations of CCN and GA number concentrations from the AIRS II were compared. CCN measurements from the DRI CCN spectrometer (Hudson 1989) across a range of supersaturation values (0.02%–1%) were compared with the GA number concentrations during the same sampling intervals. Albeit with considerable scatter, correlations of 0.6 (at the 99% confidence level) were found for the GA and the CCN when computed at supersaturation values of 0.06%, 0.08%, 0.1%, 0.2%, 0.3%, 0.4%, 0.6%, and 1% (Fig. 3). Although it is not a high correlation, it does suggest the tendency that in this dataset, air containing more CCN will also have more GA.

Fig. 3.

Correlation of observed GA to observed CCN (at 1% supersaturation) for the 14 AIRS II flights of the C-130 aircraft. The best-fit line is overlaid, with correlation coefficient R shown.

Fig. 3.

Correlation of observed GA to observed CCN (at 1% supersaturation) for the 14 AIRS II flights of the C-130 aircraft. The best-fit line is overlaid, with correlation coefficient R shown.

3) Chemical composition

For the last four flights by the C-130 during AIRS II, glass slides were exposed to the airstream outside the aircraft fuselage to collect GA in the clear air using the same technique as discussed by Colón-Robles et al. (2006). Photographs of some of these particles under an electron microscope are shown in Fig. 4. The average diameter of these particles that were collected at altitudes below 1 km AGL was 16 μm. A few particles detected by the CPI at these times were larger than several hundred micrometers.

Fig. 4.

Examples of microscope images of particles collected by impaction on glass slides (flight dates and scales are labeled).

Fig. 4.

Examples of microscope images of particles collected by impaction on glass slides (flight dates and scales are labeled).

A randomly chosen subset of 100 of these particles collected on the slides for each flight was analyzed with scanning electron microscopy/X-ray spectrometry to determine the composition (Table 2). The particles categorized as hygroscopic [sulfates and salts—see Twohy and Poellot (2005) for full description] varied considerably, ranging from 21% to 81% for a given flight, even when sampled on consecutive days over the same region (3 and 4 December over Georgian Bay). Surprising is that as many as 28% of the GA sampled at these low altitudes on a given flight consisted primarily of sodium chloride. On one of the slides during the 3 December flight, an 80-μm-diameter salt particle was collected. All of these samples were taken over Lake Huron or Georgian Bay, which are freshwater lakes, and the nearest bodies of saltwater are Hudson Bay and the Atlantic Ocean, nearly 600 and 1000 km away, respectively. On 25 November, nearly one-half of the sampled GA consisted of crustal dust, particles that are traditionally thought to be hydrophobic (and incidentally effective ice nuclei); other days had far fewer dust particles. The remaining particles were carbonaceous, which can be either hygroscopic or hydrophobic, depending on the detailed chemical type and structure.

Table 2.

Summary of chemical composition of GA particles collected on glass slides at low altitudes during AIRS II.

Summary of chemical composition of GA particles collected on glass slides at low altitudes during AIRS II.
Summary of chemical composition of GA particles collected on glass slides at low altitudes during AIRS II.

c. Derived estimates of SLD in the clouds

Quantification of SLD is a difficult task. Optical array probes (Knollenberg 1981) only provide shadow images to discern particle phase, and the 1D probe data that must be used to quantify SLD in the range of 50–100 μm in diameter4 do not contain shadow images. The CPI is useful for determining the phase of particles, collecting holographic images that clearly show the differences in ice particles and water drops, but there are uncertainties associated with its sample volume. Given these limitations, a technique was designed to emphasize measurements made with more confidence and to estimate a range of SLD during a cloud penetration to account for the uncertainties, as outlined in the appendix.

As with nearly every other physical characteristic of these clouds, there is also a large variation in the amount of SLD found over the six flights, ranging from 0.0004 to 31 L−1 (Table 3). The maximum drop size observed by the two-dimensional cloud probe (2DC) was 900 μm on 11 November, and almost every other case had drops larger than 200 μm (fourth column). There was a tendency for the clouds with the highest number concentrations of SLD also to have the largest supercooled drops (11 and 5 November and 1 December). The SLD accounted for less than one-third of the particles larger than 100 μm and was often less than one-fourth, indicating that the majority of the larger particles were ice. In some cases, there were “patches” of the cloud in which only liquid or only ice particles existed, but in others the phases were coexisting, having both ice and liquid particles in the sample volumes of the probes.

Table 3.

Large-particle data summary and derived SLD estimates for the cases listed in Table 1.

Large-particle data summary and derived SLD estimates for the cases listed in Table 1.
Large-particle data summary and derived SLD estimates for the cases listed in Table 1.

d. Relationships between SLD concentrations and other observations

1) Cloud dynamic and thermodynamic characteristics

The clouds with the fewest SLD, observed on 25 November and 4 December, had the weakest updrafts and some of the smallest cloud depths and values of LWC. The cloud with the most SLD (11 November) had the highest maximum updraft speed, the warmest cloud base, nearly the greatest cloud depth, and the largest measured LWC. Despite these extreme cases, however, the relationships between the maximum SLD observed and the maximum cloud updraft (Fig. 5a), and in-cloud temperature (Fig. 5c), did not show strong trends across the set of cases investigated here. Figure 5b does appear to show signs of increasing SLD with increasing cloud liquid water content, but the uncertainty in these measurements limits any strong conclusions.

Fig. 5.

Observed SLD (maximum estimate) vs (a) maximum updraft speed, (b) LWC, (c) in-cloud temperature, (d) maximum FSSP droplet number concentration, and (e) ambient CCN number concentration for all clouds listed in Table 1.

Fig. 5.

Observed SLD (maximum estimate) vs (a) maximum updraft speed, (b) LWC, (c) in-cloud temperature, (d) maximum FSSP droplet number concentration, and (e) ambient CCN number concentration for all clouds listed in Table 1.

2) Droplet concentration/CCN concentration

Trends in the data are also apparent when comparing the observed droplet number concentration (or CCN number concentration) and the maximum SLD observed (Figs. 5d,e). A lower droplet number concentration, as seen in maritime clouds, allows the droplets to grow larger by condensation (having less competition for the available vapor) and the variation of sizes to be greater than in a cloud with a much higher droplet concentration as seen in a polluted air mass (Pruppacher and Klett 1997; Hudson and Yum 2001; Yum and Hudson 2002). Cases with lower CCN/droplet number concentrations should thus tend to favor SLD production, by growing larger droplets capable of initiating collisions and coalescence more quickly than clouds with higher CCN/droplet number concentrations. Figures 5d and e suggest these relationships, except for one outlier, the 11 November case, that had the most SLD of any of the cases studied here but had higher droplet and CCN number concentrations than most of the other cases. If the 11 November case is eliminated, the SLD and droplet number concentration have a correlation coefficient of 0.88, but it reduces to 0.6 when the 11 November case is included.

3) Giant aerosol particles

A comparison between GA and SLD was conducted separately for each flight, because the amount of GA at the cloud altitude must be estimated from the trends at other altitudes in the clear air (Fig. 6). The assumption in this analysis is that the cloud droplets are forming upon GA at nearly the same altitude as the cloud. Assuming that the GA profiles continue the trends at the altitudes in between the sampled altitudes, the values of the GA do appear to lie within the range of the observed SLD at those altitudes, for four of the six flights. For two of the flights, the SLD concentrations are much less than the GA concentrations expected at those altitudes. In none of the cases do the SLD outnumber the GA, suggesting that GA may be a source of SLD but that not all of the GA become SLD.

Fig. 6.

Observed number concentrations of GA measured in clear air (red circles, with error bars based on Poisson counting statistics), and estimated range of SLD based on observations within cloud (blue squares, with range explained in the appendix), vs altitude for the six flights.

Fig. 6.

Observed number concentrations of GA measured in clear air (red circles, with error bars based on Poisson counting statistics), and estimated range of SLD based on observations within cloud (blue squares, with range explained in the appendix), vs altitude for the six flights.

Further investigation of the 25 November and 4 December clouds, for which the GA outnumbered the SLD, indicates that the chemical composition of the GA might also have been important. The chemical analysis of the GA collected on glass slides during the last four flights (Table 2) indicated that the GA on 25 November and 4 December consisted of fewer hygroscopic particles and thus may be less favorable for accelerated growth by condensation (and hence less favorable for accelerating the onset of coalescence) than if more of the GA were hygroscopic.

3. Parcel model calculations

The limited number of cases makes it difficult to establish what factors are dominant in controlling the amount of SLD; these factors are now explored further with the aid of a numerical model constrained by observations from the six AIRS II cases.

a. Model description

The adiabatic parcel model of Cooper et al. (1997) has been adapted for investigating the production of SLD and its dependencies pertinent to the six stratocumulus cases. A brief description of the model and the new adaptations follows, and Cooper et al. (1997) contains a detailed explanation of all physical processes and numerics. The model is initialized with CCN (assumed to be ammonium sulfate) in 256 logarithmically spaced bins. For each of the six cases, the observed CCN sampled at the cloud altitude over supersaturation ranges from 0.1% to 1% was fitted with a relationship of the form

 
formula

where s is the supersaturation in percent, N(s) is the cumulative number of CCN active at or below s, and C and k are constants; this relationship was used to represent CCN up to ∼0.1-μm radius in the model. The value of C in (1) was sometimes increased above that observed in order for the parcel model to match the observed maximum droplet number concentration. A Junge distribution represents CCN in the model from ∼0.1 to ∼1.0 μm in radius, and a similar power law with a larger slope represents the GA. Because the slope of the size distribution for the GA is unknown,5 two extreme examples are tested for each set of calculations: a smaller slope to include more GA and a larger slope to include fewer GA (Fig. 7), both anchored at the observed value of GA at a radius of 1 μm. Because the model cuts off the initialization of aerosol particles at a concentration of 10−28 cm−3, a very small slope allows much larger GA to be represented in the model (i.e., a tail out to larger sizes) than does a very large slope. A third set of model calculations cuts off the input aerosol distribution at 1 μm (essentially, no GA). These variations in input GA allow an assessment of the importance of the GA to the amount of SLD produced in the model and of the importance of the tail of the GA distribution.

Fig. 7.

Example of three different GA distributions used as input to the microphysical parcel model, in this example anchored at a radius of 1 μm to match the observed total GA number concentration observed for the 4 Dec cloud.

Fig. 7.

Example of three different GA distributions used as input to the microphysical parcel model, in this example anchored at a radius of 1 μm to match the observed total GA number concentration observed for the 4 Dec cloud.

Other initial conditions included the observed cloud-base temperature and pressure and a prescribed cloud-base vertical velocity that forces the parcel to ascend adiabatically. No entrainment is considered. Because the maximum LWC measured for each of the modeled cases is near the adiabatic value calculated from the estimated cloud-base height, the use of the adiabatic model appears to be justified. The microphysical calculations include activation of the CCN according to the maximum supersaturation calculated in the model, growth of each droplet bin by condensation, and growth by quasi-stochastic collection, where particles are reassigned to bins using a modified Kovetz and Olund (1969) scheme to limit artificial broadening of the drop sizes. Bins were added as necessary to maintain a fine resolution in the droplet sizes. Droplet sedimentation was not considered.

The model contains no representation of ice particles. Although the lack of ice in the model prevents any study of the interaction between the liquid and ice particles in these clouds, the CPI and 2D optical array probe observations showed that ice and SLD were often coexisting in close proximity, implying that the environment was saturated with respect to liquid water, which are the conditions assumed in the model. Ikeda et al. (2007) also found in upslope clouds in Oregon that where updraft speeds were on the order of a few meters per second both the warm-rain process and ice-crystal growth were supported by the large amount of condensate. Furthermore, although a study with a model capable of representing ice and warm-rain processes would be interesting, it is beyond the scope of this study. Ice nuclei measurements were not available on most of the days examined here (P. DeMott 2004, personal communication), and so the amount of ice nuclei that should be initialized in such a model is unknown; because fundamental knowledge about heterogeneous nucleation of ice itself is still lacking (e.g., Cantrell and Heymsfield 2005), such detailed modeling seems premature for comparing with the observations.

For this study, the model of Cooper et al. (1997) has been adapted to cycle upward and downward over the depths of the observed clouds, at the speed of the observed maximum updraft speed. As discussed earlier, the stratiform clouds contained cellular, convective elements with updrafts of several meters per second (Fig. 2; Table 1). Three-dimensional modeling studies of warm stratocumulus clouds by Stevens et al. (1996) and Feingold et al. (1998) showed how parcel trajectories might appear in such a cloud, cycling upward and downward in time. The parcels modeled in the current study are forced to cycle up and down over the observed depth of the cloud, with condensation occurring on ascent because of the increasing supersaturation, some evaporation occurring on descent, and droplet coalescence occurring on ascent and descent, until the model produces the minimum observed amount of SLD. In reality, the parcels would overshoot the cloud top on ascent and entrain dry air, resulting in negatively buoyant parcels that would also undershoot the cloud base. In the representation of forced cycling used here, where the parcel is confined to cycle upward and downward within the heights of cloud base and top, no dry air (or new CCN) is entrained. Here, the importance of the cycling is to represent (or limit) the amount of potential cloud water available for collection on ascent (or descent) of the trajectories within the stratocumulus cloud. The details of the entrainment at cloud top, mixing of that entrained air, and resulting effects on the droplet size distribution are not represented in the model and would affect the time required to produce the observed SLD, and thus the times found by the model are only useful for comparison among the different modeled cases. In all cases the vertical profiles of CCN measured in the clear air above the cloud tops were less than those measured below, indicating that entrainment would not have increased the droplet number concentrations within the real clouds but would have lowered them from the values represented in the model used here.

b. SLD calculations

There are three modes of droplet growth that can result in SLD by the warm-rain process acting under supercooled conditions: 1) droplets that originated on CCN smaller than the GA can grow by condensation to sizes large enough to collide and coalesce with each other and then grow further by collection to SLD sizes, 2) GA can grow by condensation to sizes large enough to collect droplets formed on the CCN smaller than the GA and thus can grow by collection to SLD sizes, and 3) GA, given enough time, can grow by condensation alone to SLD sizes (the largest GA in the distribution, although present in a trivial amount, are already larger than the threshold for SLD when initialized in the model). The model calculations were used to investigate the relative importance of these mechanisms in the six AIRS II cases.

One cloud from each of the six AIRS II flights was selected for modeling.6 The SLD were quantified by integrating over the bins containing droplets greater than 50-μm diameter at different times throughout the model run. The microphysical parcel model calculations were run for the time required to reproduce the observed number concentration of SLD with the “more larger GA” profile of Fig. 7. These time periods ranged from a minimum of 1200 s to a maximum of 4800 s, with the median value lying between 2400 and 3600 s.

1) The importance of GA to SLD formation

Figure 8 shows that the model produces SLD within the range observed for all six AIRS II cases, using the upper estimates of GA derived from the observations. When the amount of GA input into the model is varied, however, the six cases bifurcate into two categories: one in which the GA greatly influence the amount of SLD produced in the model and another in which the GA have less of an effect. The model produces significantly fewer SLD for the 25 November and 4 December cases when the number of GA are reduced or eliminated. Additional runs of the model excluding droplet collection (not shown) indicated that nearly all of the SLD produced was from condensational growth on the largest GA, those of 10-μm radius and larger (ultragiant aerosol). In contrast, the 5 November, 11 November, 1 December, and 3 December cases show comparatively little sensitivity to the amount of GA initialized in the model, although they do show some decrease when the GA are eliminated. The absolute amount of GA input into the model does not appear to control the differences among the results: the total number concentration of GA initialized in the model for 3 and 4 December is the same, and yet the 3 December case appears to be much less sensitive to the GA, whereas the 4 December case appears to depend greatly on the GA for SLD production.

Fig. 8.

Observed range of SLD for one cloud each day as listed in Table 3, and corresponding modeled SLD considering the three different GA profiles as shown in Fig. 7.

Fig. 8.

Observed range of SLD for one cloud each day as listed in Table 3, and corresponding modeled SLD considering the three different GA profiles as shown in Fig. 7.

It is indeed interesting that the two cases in Fig. 8 in which the GA appear to influence the amount of SLD the most (25 November and 4 December) are those in which the observational analysis of section 2 had shown the largest discrepancy in the GA and SLD concentrations (Fig. 6). The model calculations suggest that using the total number of GA to compare with the SLD, as presented in Fig. 6, was misleading. In the two modeled cases, the SLD formed only upon those GA of greater than approximately 10-μm radius (ultragiant aerosol), present in number concentrations far less than the total number of GA plotted in Fig. 6. The modeling also suggests that the agreement in the total number concentration of GA and the observed SLD shown for the other four cases in Fig. 6 is an odd coincidence—no other explanation has been found.

2) The importance of CCN/droplet number versus cloud depth/LWC

From Fig. 8, the modeling results suggest that the 5 November, 11 November, 1 December, and 3 December cases produced higher amounts of SLD, regardless of the amount of GA present. Thus, SLD must be forming by collection among droplets produced by CCN smaller than the GA. All but the 11 November case had lower CCN/droplet number concentrations (Table 1) than the 25 November and 4 December cases. Thus the relationship shown among these variables in Figs. 5d and e was strong but did not explain all of the AIRS II SLD cases.

The reason for the differences in the SLD production among these cases becomes clear by looking at the drop size distributions calculated in the model on the first ascent of the parcels from cloud base to cloud top (Fig. 9): a trade-off exists between the depth of the cloud (proportional to LWC in the adiabatic parcel model) and number concentration of CCN/cloud droplets. A cloudy parcel with higher CCN/droplet concentration requires more time (translating into a greater cloud depth for more condensational growth during ascent) to produce droplets large enough to initiate collection. The cases with higher CCN/droplet concentrations but smaller depths (25 November and 4 December) have very few collection events on the first ascent of the parcel (seen in Fig. 9 as a lack of “bumps” that are formed by collection events in the droplet size distribution). On the other hand, all of the other cases have a greater cloud depth (i.e., more time for condensational growth of the drops, equivalent to a higher LWC) over which the collection process can begin, and thus, even on the first ascent of the modeled parcels, numerous collection events have occurred. The 11 November case highlights the importance of this balance between the CCN/droplet number concentration and the cloud depth/LWC: although it has one of the higher CCN/droplet number concentrations, it has a depth that is more than 1 km greater than the 25 November or 4 December cases, allowing more droplet growth by condensation (as indicated by an LWC that is at least 0.5 g m−3 greater in Table 1), to sizes at which growth by collection can become effective.

Fig. 9.

Modeled drop size distributions for each of the six AIRS II cases, at a few tens of meters above cloud base and later at the top of the cloud after the first ascent of the parcel.

Fig. 9.

Modeled drop size distributions for each of the six AIRS II cases, at a few tens of meters above cloud base and later at the top of the cloud after the first ascent of the parcel.

The 11 November case is also special in that it had the warmest cloud-base temperature (∼0°C) of any of the cases examined here, and this also served to facilitate the warm-rain process. If the cloud-base conditions of the 11 November case were changed to match those of the 25 November case (7°C colder cloud base) or the 1 December case (6°C colder cloud base), all else being equal, the SLD was reduced from 14 to 0.008 L−1 or 0.01 L−1, respectively, because growth by condensation is reduced and the impact of the large number of droplets is more substantial. Thus, although the emphasis in this study has been upon microphysical factors that influence SLD production, clearly there are thermodynamic factors that are important as well, already noted in forecasting techniques (e.g., Bernstein et al. 2005) and well established in warm-rain theory (e.g., Pruppacher and Klett 1997). The larger amount of SLD observed in the 11 November cloud appears to have resulted from a combination of a warmer cloud base and a larger cloud depth, both of which allowed sufficient growth by condensation of a moderately high number of droplets to reach sizes effective for further growth by collection to create SLD.

3) Sensitivity tests

Modeling the same six cases but varying the input parameters to the model demonstrates the sensitivity of the SLD production in the model to various factors. In this set of calculations, the model was always run for 3600 s, regardless of whether the observed amount of SLD was produced, to facilitate comparison among the cases. All cases were run with the “more larger GA” profile of Fig. 7; results had similar trends when the “fewer larger GA” profile was used instead.

The sensitivity of the SLD produced in the model to various parameters further illustrates the importance of cloud depth/liquid water content and CCN/droplet number concentration in encouraging SLD production. In the deeper clouds for which the majority of SLD formed by collection among droplets formed on CCN smaller than GA (Fig. 10a), increasing the CCN/drop number concentration slows condensational growth and thus the onset of droplet collection and SLD production; decreasing the CCN/drop number concentration accelerates condensational growth, the onset of droplet collection, and hence SLD production. The GA are responsible for some of the SLD (as seen in the decreases in SLD when GA are eliminated) primarily by growing by condensation to sizes large enough to collect substantial amounts of smaller droplets. In the shallower clouds with higher droplet concentrations, where SLD formed only on GA (Fig. 10b), the modeled SLD is insensitive to all of the parameters tested except eliminating the GA. In these shallow clouds, even halving the droplet concentration in the model did not allow enough condensational growth to initiate collection among the droplets formed on the CCN smaller than the GA.

Fig. 10.

Sensitivity of SLD produced in the model calculations to doubling the droplet concentration (dbl_N), halving the droplet concentration (half_N), turning off collection calculations (no_coll), halving the updraft speed (half_w), making all GA nonhygroscopic (nonhygro_GA), and eliminating all GA (no_GA) for cases initialized with more GA. Some results are not plotted for the “half_N” run in (a) because SLD amounts exceeded model upper limit; results are not plotted for “no_GA” runs in (b) because no SLD were produced.

Fig. 10.

Sensitivity of SLD produced in the model calculations to doubling the droplet concentration (dbl_N), halving the droplet concentration (half_N), turning off collection calculations (no_coll), halving the updraft speed (half_w), making all GA nonhygroscopic (nonhygro_GA), and eliminating all GA (no_GA) for cases initialized with more GA. Some results are not plotted for the “half_N” run in (a) because SLD amounts exceeded model upper limit; results are not plotted for “no_GA” runs in (b) because no SLD were produced.

In all of the cases, the amount of SLD produced in the model was insensitive to a few parameters. Halving the updraft speed resulted in fewer droplets being activated as a result of the lower maximum supersaturation but did not have a major effect upon SLD production. The model lacks droplet sedimentation, but the strength of the updrafts used (and observed) here would still limit such effects in reality. The hygroscopicity of the GA also had no effect upon the amount of SLD produced in the model, suggesting that the different compositions of the GA on 25 November and 4 December (Table 2) were not major factors in inhibiting SLD production on those days.

4. Conclusions and discussion

The results of the observational analysis and the numerical modeling presented here lead to the following four conclusions:

  • 1) Supercooled stratocumulus clouds from six different days during the AIRS II field campaign contained SLD ranging from 0.0004 to 30 L−1, and occurred in environments with temperatures entirely less than 0°C, eliminating the possibility that they were formed by melting of frozen hydrometeors.

  • 2) The dominant mechanism of SLD formation in these clouds appears to have been an efficient supercooled warm-rain process, that is, condensation on droplets formed on CCN of less than 1 μm in radius, followed by collision and coalescence among those droplets. This mechanism was more effective in deeper clouds with lower droplet number concentrations associated with fewer ambient CCN, and it was less effective in shallower clouds with higher droplet concentrations resulting from more ambient CCN. Thus, in the clouds studied here, SLD are anticorrelated with droplet number concentration (and thus also CCN) as has been observed with drizzle in warm stratiform clouds (Yum and Hudson 2002; Twohy et al. 2005), although the cloud depth/liquid water content appears to modulate this result in this study.

  • 3) Although GA observed in the clear air during these six flights indicated that they were present in large enough numbers to have been the source of the SLD, the microphysical calculations showed that GA were a secondary mechanism of SLD formation, only becoming the dominant mechanism in shallow clouds with higher droplet number concentrations where the warm-rain process originating on CCN smaller than the GA was much less effective. In these cases, however, the observed and calculated SLD were much less (by several orders of magnitude) than that in the clouds with lower droplet number concentrations where the warm-rain process was more efficient. The solubility of the GA appeared to have little influence on SLD formation despite some suggestion of an influence from the observational analysis.

  • 4) The greatest amount of SLD produced in the six cases studied here resulted from one of the deepest stratocumulus (resulting in a higher liquid water content) with the warmest cloud base but with a moderately high droplet number concentration. These factors worked together to favor an efficient warm-rain process despite the higher droplet number concentration in the cloud.

These results are consistent with previous studies by Song and Marwitz (1989), Rasmussen et al. (1995), Murakami et al. (1992), Bernstein et al. (2004), and Ikeda al. (2007), all of which found that SLD were forming in clouds with lower drop concentrations, consistent with an efficient warm-rain process. Other potential effects on SLD formation, such as entrainment and mixing induced by wind shear near cloud top, or isobaric mixing, have not been addressed in this study. They may be important in other cases but appear not to be required to explain the SLD observed in these AIRS II cases.

The results presented here are also consistent with previous numerical modeling results that suggest that the warm-rain process occurring in clouds forming in environments that contain fewer CCN (often maritime air) is not substantially enhanced by the presence of GA but that for clouds forming in environments that contain more CCN (as in continental or urban air) GA can be more effective for producing drizzle or raindrops. These past modeling studies have been performed for a variety of clouds, ranging from small warm cumulus clouds (Ochs and Semonin 1979; Johnson 1982; Blyth et al. 2003) to warm stratocumulus clouds (Feingold et al. 1999) to stably stratified stratus (Geresdi and Rasmussen 2005). The current study lends strong observational support for these modeling results in stratocumulus clouds; other recent studies of observations collected in small warm cumulus clouds have reported similar conclusions (Göke et al. 2007; Hudson and Mishra 2007).

At this time it is unclear how much the results presented here can aid the operational community in forecasting SLD, because number concentrations of CCN, GA, and cloud droplets are not routinely observed. Running high-resolution weather models at regional scales, as is capable with the Weather and Research Forecasting (WRF) model, might demonstrate whether air trajectories into the forecast region may be bringing air of a more maritime or urban nature. This qualitative information, combined with an estimate of cloud depth/liquid water content, might aid in evaluating the more extreme cases but may not be accurate enough for more subtle balances between CCN/droplet number concentrations and cloud depth/liquid water content. Future work such as WRF modeling of the AIRS II cases to help to predict CCN concentrations might help to evaluate the efficacy of this approach.

A final comment seems appropriate regarding possible anthropogenic effects on SLD production based on analysis of the AIRS II data. The chemical analysis of the GA collected during AIRS II showed that sometimes a large fraction was primarily composed of sodium chloride and other salts—surprising because the AIRS II area was located so far from any bodies of saltwater. Ground-based aerosol studies in this region by Biegalski et al. (1998) and Polissar et al. (2001) identified salts in their data as well, and Biegalski et al. noted higher concentrations of the salts during winter months. They cite possible sources as transport from the oceans or road salt spread on hazardous roads. These two sources are not easily differentiated, because road salt is often mined from salt beds left behind from ancient dried seas. Back trajectories run using the Hybrid Single-Particle Lagrangian Integrated Trajectory (HYSPLIT) model (Draxler and Hess 1998) with the National Centers for Environmental Prediction–NCAR reanalysis gridded meteorological dataset (Kalnay et al. 1996; Kistler et al. 2001) for the dates listed in Table 2 indicated that these giant salt particles would have had to travel from 2 to 4 days from an ocean to the observation region. There was little difference in trajectory origin for the 1 and 3 December flights, during which salt particles were observed, and the 4 December flight, during which salt was not observed, and, to the extent that these trajectories can be trusted, this suggests a local source region for the giant salt particles. Although the effect of GA producing SLD appears to be secondary in this study, the model calculations suggest that more GA (especially ultragiant aerosol) could produce more SLD, especially in more polluted air masses, and thus this possible anthropogenic source of SLD deserves further attention.

Acknowledgments

The authors acknowledge all of the participants of AIRS II who made this study possible, including other principal investigators (J. Hallett, P. DeMott, and M. Bailey) and students who helped to collect data with the C-130 aircraft, the staff at the Research Air Facility at NCAR and specifically the C-130 crew, George Isaac and the Meteorological Service of Canada staff (including Monika Bailey, who handled numerous data requests), and the NASA Glenn Aircraft Icing Group for the use of their facilities in support of the aircraft. Jorgen Jensen provided giant nuclei samples for the electron microscopy analysis. The numerical model used in this study was supplied by Al Cooper of NCAR, and software for analysis of the aircraft data was provided by the Research Aviation Facility at NCAR. Trajectory analysis using the HYSPLIT model was performed by J. Stachnik. Useful feedback on this study at various stages was provided by Ben Bernstein, Marcia Politovich, Roy Rasmussen, Greg Thompson, and Frank McDonough, all from the Research Applications Laboratory of the NCAR. Two anonymous reviewers supplied comments that substantially improved the study. This study was funded by the National Science Foundation, under Awards ATM-0312439, ATM-0312036, and ATM-0313899. NCAR is sponsored by the National Science Foundation.

REFERENCES

REFERENCES
Baumgardner
,
D.
,
1983
:
An analysis and comparison of five water droplet measuring instruments.
J. Climate Appl. Meteor.
,
22
,
891
910
.
Baumgardner
,
D.
, and
A.
Rodi
,
1989
:
Laboratory and wind tunnel evaluations of the Rosemount Icing Detector.
J. Atmos. Oceanic Technol.
,
6
,
971
979
.
Beard
,
K. V.
, and
H. T.
Ochs
III
,
1993
:
Warm-rain initiation: An overview of microphysical mechanisms.
J. Appl. Meteor.
,
32
,
608
625
.
Bernstein
,
B. C.
,
F.
McDonough
,
C. A.
Wolff
,
M. K.
Politovich
,
R. M.
Rasmussen
, and
S. G.
Cober
,
2004
:
Diagnosis of supercooled large drop conditions using cloud water content and drop concentration. Preprints, 11th Conf. on Aviation, Range and Aerospace Meteorology, Hyannis, MA, Amer. Meteor. Soc., P8.6. [Available online at http://ams.confex.com/ams/pdfpapers/81619.pdf.]
.
Bernstein
,
B. C.
,
F.
McDonough
,
M. K.
Politovich
, and
B. G.
Brown
,
2005
:
Current icing potential: Algorithm description and comparison with aircraft observations.
J. Appl. Meteor.
,
44
,
969
986
.
Bernstein
,
B. C.
,
C. A.
Wolff
, and
F.
McDonough
,
2007
:
An inferred climatology of icing conditions aloft, including supercooled large drops. Part I: Canada and the continental United States.
J. Appl. Meteor. Climatol.
,
46
,
1857
1878
.
Biegalski
,
S. R.
,
S.
Landsberger
, and
R. M.
Hoff
,
1998
:
Source–receptor modeling using trace metals in aerosols collected at three rural Canadian Great Lakes sampling stations.
J. Air Waste Manage. Assoc.
,
48
,
227
237
.
Biter
,
C. J.
,
J. E.
Dye
,
D.
Huffman
, and
W. D.
King
,
1987
:
The drop-size response of the CSIRO liquid water probe.
J. Atmos. Oceanic Technol.
,
4
,
359
367
.
Blyth
,
A. M.
,
S. G.
Lasher-Trapp
,
W. A.
Cooper
,
C. A.
Knight
, and
J.
Latham
,
2003
:
The role of giant and ultragiant nuclei in the formation of early radar echoes in warm cumulus clouds.
J. Atmos. Sci.
,
60
,
2557
2572
.
Cantrell
,
W.
, and
A.
Heymsfield
,
2005
:
Production of ice in tropospheric clouds: A review.
Bull. Amer. Meteor. Soc.
,
86
,
1743
1746
.
Cober
,
S. G.
,
G. A.
Isaac
, and
A. V.
Korolev
,
2001a
:
Assessing the Rosemount Icing Detector with in situ measurements.
J. Atmos. Oceanic Technol.
,
18
,
515
528
.
Cober
,
S. G.
,
G. A.
Isaac
, and
J. W.
Strapp
,
2001b
:
Characterizations of aircraft icing environments that include supercooled large drops.
J. Appl. Meteor.
,
40
,
1984
2002
.
Colón-Robles
,
M.
,
R. M.
Rauber
, and
J. B.
Jensen
,
2006
:
Influence of low-level wind speed on droplet spectra near cloud base in trade wind cumulus.
Geophys. Res. Lett.
,
33
.
L20814, doi:10.1029/2006GL027487
.
Cooper
,
W. A.
,
R. T.
Bruintjes
, and
G. K.
Mather
,
1997
:
Calculations pertaining to hygroscopic seeding with flares.
J. Appl. Meteor.
,
36
,
1449
1469
.
Draxler
,
R. R.
, and
G. D.
Hess
,
1998
:
An overview of the HYSPLIT_4 modeling system for trajectories, dispersion, and deposition.
Aust. Meteor. Mag.
,
47
,
295
308
.
Feingold
,
G.
,
M.
Kreidenweis
, and
Y.
Zhang
,
1998
:
Stratocumulus processing of gases and cloud condensation nuclei. 1. Trajectory ensemble model.
J. Geophys. Res.
,
103
,
19527
19542
.
Feingold
,
G.
,
W. R.
Cotton
,
S. M.
Kreidenweis
, and
J. T.
Davis
,
1999
:
The impact of giant cloud condensation nuclei on drizzle formation in stratocumulus: Implications for cloud radiative properties.
J. Atmos. Sci.
,
56
,
4100
4117
.
Gardiner
,
B. A.
, and
J.
Hallett
,
1985
:
Degradation of in-cloud forward scattering spectrometer probe measurements in the presence of ice particles.
J. Atmos. Oceanic Technol.
,
2
,
171
180
.
Geresdi
,
I.
, and
R.
Rasmussen
,
2005
:
Freezing drizzle formation in stably stratified layer clouds. Part II: The role of giant nuclei and aerosol particle size distribution and solubility.
J. Atmos. Sci.
,
62
,
2037
2057
.
Göke
,
S.
,
H. T.
Ochs
III
, and
R. M.
Rauber
,
2007
:
Radar analysis of precipitation initiation in maritime versus continental clouds near the Florida coast: Inferences concerning the role of CCN and giant nuclei.
J. Atmos. Sci.
,
64
,
3695
3707
.
Hudson
,
J. G.
,
1989
:
An instantaneous CCN spectrometer.
J. Atmos. Oceanic Technol.
,
6
,
1055
1065
.
Hudson
,
J. G.
, and
S. S.
Yum
,
2001
:
Maritime-continental drizzle contrasts in small cumuli.
J. Atmos. Sci.
,
58
,
915
926
.
Hudson
,
J. G.
, and
S.
Mishra
,
2007
:
Relationships between CCN and cloud microphysics variations in clean maritime air.
Geophys. Res. Lett.
,
34
.
L16804, doi:10.1029/2007GL030044
.
Huffman
,
G. J.
, and
G. A.
Norman
Jr.
,
1988
:
The supercooled warm rain process and the specification of freezing precipitation.
Mon. Wea. Rev.
,
116
,
2172
2182
.
Ikeda
,
K.
,
R. M.
Rasmussen
,
W. D.
Hall
, and
G.
Thompson
,
2007
:
Observations of freezing drizzle in extratropical cyclonic storms during IMPROVE-2.
J. Atmos. Sci.
,
64
,
3016
3043
.
Isaac
,
G. A.
, and
Coauthors
,
2005
:
First results from the Alliance Icing Research Study II. Preprints.
AIAA 43rd Aerospace Science Meeting and Exhibit, Reno, NV, American Institute of Aeronautics and Astronautics, 1–18
.
Johnson
,
D. B.
,
1982
:
The role of giant and ultragiant aerosol particles in warm rain initiation.
J. Atmos. Sci.
,
39
,
448
460
.
Kalnay
,
E.
, and
Coauthors
,
1996
:
The NCEP/NCAR 40-Year Reanalysis Project.
Bull. Amer. Meteor. Soc.
,
77
,
437
471
.
King
,
W. D.
,
D. A.
Parkin
, and
R. J.
Handsworth
,
1978
:
A hot-wire liquid water device having fully calculable response characteristics.
J. Appl. Meteor.
,
17
,
1809
1813
.
Kistler
,
R.
, and
Coauthors
,
2001
:
The NCEP–NCAR 50-Year Reanalysis: Monthly means CD-ROM and documentation.
Bull. Amer. Meteor. Soc.
,
82
,
247
267
.
Knollenberg
,
R. G.
,
1981
:
Techniques for probing cloud microstructure.
Clouds, Their Formation, Optical Properties, and Effects, P. V Hobbs and A. Deepak, Eds., Academic Press, 15–92
.
Korolev
,
A. V.
, and
G. A.
Isaac
,
2000
:
Drop growth due to high supersaturation caused by isobaric mixing.
J. Atmos. Sci.
,
57
,
1675
1685
.
Korolev
,
A. V.
,
J. W.
Strapp
, and
G. A.
Isaac
,
1998
:
Evaluation of the accuracy of PMS optical array probes.
J. Atmos. Oceanic Technol.
,
15
,
708
720
.
Kovetz
,
A.
, and
B.
Olund
,
1969
:
The effect of coalescence and condensation on rain formation in a cloud of finite vertical extent.
J. Atmos. Sci.
,
26
,
1060
1065
.
Lasher-Trapp
,
S.
, and
J. P.
Stachnik
,
2007
:
Giant and ultragiant aerosol particle variability over the eastern Great Lakes region.
J. Appl. Meteor. Climatol.
,
46
,
651
659
.
Lasher-Trapp
,
S.
,
C. A.
Knight
, and
J. M.
Straka
,
2001
:
Early radar echoes from ultragiant aerosol in a cumulus congestus: Modeling and observations.
J. Atmos. Sci.
,
58
,
3545
3562
.
Lawson
,
R. P.
, and
R. H.
Cormack
,
1995
:
Theoretical design and preliminary tests of two new particle spectrometers for cloud microphysics research.
Atmos. Res.
,
35
,
315
348
.
Marwitz
,
J. D.
,
M.
Politovich
,
B.
Bernstein
,
F.
Ralph
,
P.
Neiman
,
R.
Ashenden
, and
J.
Bresch
,
1997
:
Meteorological conditions associated with the ATR72 aircraft accident near Roselawn, Indiana, on 31 October 1994.
Bull. Amer. Meteor. Soc.
,
78
,
41
52
.
Murakami
,
M.
,
Y.
Yamada
,
T.
Matsuo
,
H.
Mizuno
, and
K.
Morikawa
,
1992
:
Microphysical structures of warm-frontal clouds. The 20 June 1987 case study.
J. Meteor. Soc. Japan
,
70
,
877
895
.
Ochs
III,
H. T.
, and
R. G.
Semonin
,
1979
:
Sensitivity of a cloud microphysical model to an urban environment.
J. Appl. Meteor.
,
18
,
1118
1129
.
Pobanz
,
B. M.
,
J. D.
Marwitz
, and
M. K.
Politovich
,
1994
:
Conditions associated with large-drop regions.
J. Appl. Meteor.
,
33
,
1366
1372
.
Polissar
,
A. V.
,
P. K.
Hopke
, and
R. L.
Poirot
,
2001
:
Atmospheric aerosol over Vermont: Chemical composition and sources.
Environ. Sci. Technol.
,
35
,
4604
4621
.
Politovich
,
M. K.
,
1989
:
Aircraft icing caused by large supercooled droplets.
J. Appl. Meteor.
,
28
,
856
868
.
Pruppacher
,
H. R.
, and
J. D.
Klett
,
1997
:
Microphysics of Clouds and Precipitation.
Kluwer Academic, 954 pp
.
Rasmussen
,
R. M.
,
B. C.
Bernstein
,
M.
Murakami
,
G.
Stossmeister
,
J.
Reisner
, and
B.
Stankov
,
1995
:
The 1990 Valentine’s Day Arctic outbreak. Part I: Mesoscale and microscale structure and evolution of a Colorado Front Range shallow upslope cloud.
J. Appl. Meteor.
,
34
,
1481
1511
.
Sand
,
W. R.
,
W. A.
Cooper
,
M. K.
Politovich
, and
D. L.
Veal
,
1984
:
Icing conditions encountered by a research aircraft.
J. Climate Appl. Meteor.
,
23
,
1427
1440
.
Song
,
N.
, and
J. D.
Marwitz
,
1989
:
A numerical study of the warm rain process in orographic clouds.
J. Atmos. Sci.
,
46
,
3479
3486
.
Stevens
,
B.
,
G.
Feingold
,
W. R.
Cotton
, and
R. L.
Walko
,
1996
:
Elements of the microphysical structure of numerically simulated nonprecipitating stratocumulus.
J. Atmos. Sci.
,
53
,
980
1006
.
Strapp
,
J. W.
,
R. A.
Stuart
, and
G. A.
Isaac
,
1996
:
A Canadian climatology of freezing precipitation, and a detailed study using data from St. John’s, Newfoundland. Proc. Int. Conf. on Aircraft Inflight Icing, Springfield, VA, Federal Aviation Administration, 45–56
.
Twohy
,
C. H.
, and
M. R.
Poellot
,
2005
:
Chemical characteristics of ice residual nuclei in anvil cirrus clouds: Evidence for homogenous and heterogeneous ice formation.
Atmos. Chem. Phys.
,
5
,
2289
2297
.
Twohy
,
C. H.
,
M. D.
Petters
,
J. R.
Snider
,
B.
Stevens
,
W.
Tahnk
,
M.
Wetzel
,
L.
Russell
, and
F.
Burnet
,
2005
:
Evaluation of the aerosol indirect effect in marine stratocumulus clouds: Droplet number, size, liquid water path and radiative impact.
J. Geophys. Res.
,
110
.
D08203, doi:10.1029/2004JD005116
.
Yum
,
S. S.
, and
J. G.
Hudson
,
2002
:
Maritime/continental microphysical contrasts in stratus.
Tellus
,
54B
,
61
73
.

APPENDIX

Quantification of SLD

For each section of cloud where icing was detected by the Rosemount probe, the SLD were quantified. Shadow images larger than 100 μm from the 2DC optical array probe were analyzed manually to determine the number of liquid drops or ice particles. Images that were round with no rough edges were counted as liquid drops; all other images were considered to be ice. Resulting ratios of liquid-to-total particle number were then computed for each time interval. When large numbers of images occurred for a single cloud penetration (over 4500 images per minute over a period of 50 min were collected within one cloud), 1-min samples were analyzed and their ratios were time-weighted by the duration of the periods with similar icing rates. The 2DC images that were smaller than 100 μm in diameter were not included in the particle counts; therefore the SLD concentration in the size range of 50–100 μm was estimated by multiplying the 260X probe total concentration in that size range by the same liquid-to-total particle percentage derived from the 2DC data in that cloud. By adding this number concentration to the 2DC-derived SLD estimate for particles greater than 100 μm, a conservative estimate of SLD was produced. An upper bound on the possible range of SLD number concentrations was derived similarly, but assuming that all 50–100-μm particles observed by the 260X probe were SLD. Although possible error results both from the manual analysis of the 2DC images and from the assumptions behind the 1D probe data, the representation of the SLD concentrations as a range of values is designed to contain those errors and uncertainties.

Footnotes

Corresponding author address: Sonia Lasher-Trapp, Dept. of Earth and Atmospheric Sciences, Purdue University, 550 Stadium Mall Dr., West Lafayette, IN 47907. Email: slasher@purdue.edu

1

Manufactured by Droplet Measuring Technologies, Boulder, CO. A review of the probe operation principles and characteristics can be found in Baumgardner et al. (1983).

2

Manufactured by Stratton Park Engineering Co., Boulder, CO.

3

Profiles shown here may be slightly different than those presented in Lasher-Trapp and Stachnik (2007), because in that study, no restriction was made that the aerosol sampling and cloud sampling occur in the same air mass. The six flights shown here were often in the vicinity of a cold or warm front, and so an attempt was made to limit the clear-air samples to times during which the sampled aerosols were in the same air mass as the supercooled clouds. This procedure consisted of a review of the flight tracks and the synoptic weather charts at different pressure levels, corresponding to the different flight altitudes of the aerosol particle and cloud measurements.

4

The 2DC specifications are that it measures particles as small as 25 μm, but the number concentrations are thought to be less reliable for sizes below 100 μm (Korolev et al. 1998).

5

The averaged clear-air FSSP data do yield a size distribution at each sampling altitude, but, because the largest GA/ultra-GA are so rare and the sampling volume of the FSSP is so small for such particles, the slope out to larger sizes cannot be determined with confidence.

6

For the two days with multiple clouds containing SLD listed in Table 1, the first cloud listed on 25 November and the second cloud listed on 3 December were modeled.