Origins of Ice Crystals in Clouds

The recent studies by Hobbs and Rangno [1990] and Rangno and Hobbs [1991] give further evidence for the rapid production of ice crystals near the tops of shallow, maritime cumulus clouds. They show that ice crystal concentrations can increase from 1 l to 100 l in approximately 10 minutes for cloud top temperatures warmer than -12C. They also observed that liquid and frozen drizzle droplets are present below this level just prior to the rapid formation of high concentrations of ice crystals. They show that the ice crystal concentrations build up too fast to be explained by the Hallett-Mossop [1974] rime-splintering mechanism, and suggest that these high ice crystal concentrations might form in localized pockets of high supersaturation with respect to water during the onset of coalescence growth. As mentioned above, the recent results of Fukuta [1993] showed that supersaturations up to 10% may be expected near cloud base due to the slower diffusional growth of droplets as predicted from the diffusion-kinetic droplet growth theory. If local updrafts are created near cloud top, then pockets of high supersaturation may be a possibility. The recent assessment of ice initiation in warm-based convective clouds by Beard [1992], however, suggests that the conditions cited by Hobbs and Rangno [1990] for the production of high supersaturation (the onset of coalescence growth) also requires the presence of very high reflectivity (50 dBZ, where dBZ is the radar reflectivity factor in decibel units), which was not always observed in the results of Hobbs and Rangno, and Rangno and Hobbs. In addition, Baker [1991] showed that high supersaturations associated with the freezing of drops was unlikely to contribute significantly to ice formation due to the relatively small region around the drop compared to the cloud region with high ice crystal concentration. Beard [1992] also suggested that evaporative cooling, thermophoretic shock (particle motion induced by strong temperature gradients), and electric charge mechanisms would only increase the ice concentrations moderately, and were not sufficient to explain the rapid production of ice in maritime clouds where the Hallett-Mossop process was not acting. One possible mechanism that could produce high ice crystal concentrations is the ``evaporation ice nuclei'' mechanism suggested by Rosinski [1991] and highlighted by Beard in his review. Rosinski gives evidence from filter measurements that evaporated droplets occasionally left aerosol particle residues that acted as sorption ice nuclei at temperatures as high as -5C and water vapor saturation over ice of 0.2%. Since cloud droplet concentrations are typically 10-100 cm, high concentrations of ice nuclei could potentially be produced by such a mechanism. Beard [1992] for instance, suggests that only the residue of 1 in 200 evaporated cloud droplets would need to serve as ice nuclei in order to explain the rapid ice production observed near the tops of maritime cumulus clouds by Hobbs and Rangno, and Rangno and Hobbs. Rosinski [1991] also suggested a positive feedback mechanism for the production of ``evaporation ice nuclei.'' As ice crystals start to form on newly produced evaporation ice nuclei, cloud droplets close to the growing ice crystal will evaporate, leading to the production of more evaporation ice nuclei. Calculations are needed to determine the time scale of this feedback mechansim, however. The above discussion shows that more work is needed in the area of evaporation ice nuclei and its potential role in the rapid formation of ice crystal concentrations in shallow, maritime clouds. Further studies are also required on the role of local regions of high supersaturations. Clearly, much work is still needed in order to understand the formation of ice in clouds.

The formation of ice in continental cumulus clouds has been discussed by Braham [1968] and others and shown to follow the so called ``warm-rain freezing mechanism'' whereby raindrops produced by the coalescence process freeze near -5 to -10C and start to rime, leading to rapid ice production of secondary ice by the Hallett-Mossop rime-splintering mechanism. The mechanism for causing the raindrops to freeze at such a warm temperature, however, is not clear due to the lack of apparent ice nuclei at such warm temperatures. Beard's [1992] assessment, however, showed that the direct impaction of giant particles onto the raindrops may provide a reasonable mechansim for the freezing of these raindrops. He points out that the recent extensive Soviet measurements of ice nuclei sampled using an aircraft and segregated by size using a cascade impactor show that giant nuclei (>1m, 10-6m) are the predominant ice nuclei at temperatures warmer than -12C. Beard [1992] showed that typical concentrations of giant nuclei captured by drizzle drops through direct impaction are able to explain the observations of initial freezing of drizzle droplets in convective clouds leading to the initiation of secondary ice production by the Hallett-Mossop process.

The initiation of ice in wintertime clouds has been modeled by Meyers et al. [1992] using the Colorado State University Regional Atmospheric Modeling System (CSU RAMS) model with a new ice nucleation parameterization based on continuous flow diffusion chamber results for deposition and condensation freezing (Rogers [1993]). The chamber results show an exponential variation of ice nuclei concentration with ice supersaturation, reasonably independent of temperature between -7 and -20 C. In addition, a revised contact freezing nucleation parameterization is included in the model. Simulation of an orographic precipitation event over the Sierra Nevadas indicated much better agreement with observations than previously. The major improvement was due to the reduction of ice nucleation by contact freezing by as much as three orders of magnitude.

Heymsfield and Miloshevich [1993] studied ice initiation in cold, lenticular wave clouds over the Rocky Mountains, and showed that homogeneous nucleation is responsible for ice production in these clouds at temperatures colder than -33C. They observed relatively few ice crystals at temperatures warmer than -33C, suggesting that relatively few ice nuclei are present to initiate heterogeneous freezing processes at these warmer temperatures. They also observed liquid water drops down to -40.7C. Jensen et al. [1992] estimated from lidar measurements that volcanic aerosols may increase the ice concentration in cirrus clouds by as much as a factor of 5 at temperatures below -50C. Knollenberg et al. [1993] observed extremely high ice crystal concentrations in tropical anvils using a two-dimensional Grey optical array imaging probe. The measurements were made at altitudes ranging from 13 to 18 km at temperatures from -60 to -90C. Number concentrations of ice crystals greater than 10 cm were measured with sizes less than 100 m. They suggest that homogeneous freezing of ice involving HSO nuclei could possibly explain the high number concentrations observed.

The above studies have all used aircraft to sample the in-situ ice crystal concentrations. As discussed by Rangno and Hobbs [1983], the passage of an aircraft through a cloud may result in the production of ice crystals as well. Recent studies by Woodley et al. [1991], Kelly and Vali [1991], Sassen [1991], and Foster and Hallett [1993] have shown that conditions requiring large thrust from the propellors (for example climbing, icing, or flying at very slow speeds with flaps down) are most likely to produce ice particles and should be avoided in all cloud passes when re-penetration is intended. These studies all suggest that the ice formation process is primarily through the homogeneous freezing of cloud droplets in the region of strong adiabatic expansion and cooling near the propellor tips of the aircraft. A secondary mechanism was enhanced heterogeneous nucleation when the cooling in the propellor tip vortex falls short of that necessary for homogeneous nucleation.