By continuing to browse this site you agree to us using cookies as described in About Cookies
Notice: Wiley Online Library will be unavailable on Saturday 7th Oct from 03.00 EDT / 08:00 BST / 12:30 IST / 15.00 SGT to 08.00 EDT / 13.00 BST / 17:30 IST / 20.00 SGT and Sunday 8th Oct from 03.00 EDT / 08:00 BST / 12:30 IST / 15.00 SGT to 06.00 EDT / 11.00 BST / 15:30 IST / 18.00 SGT for essential maintenance. Apologies for the inconvenience.
 Ground-based lidar, radar and microwave radiometer observations at Eureka, Canada, Barrow, Alaska and over the western Arctic Ocean measure physical characteristics and morphology of stratiform clouds. Despite transition of a cold atmosphere (−15 C)through ice supersaturated conditions, ice is not observed until soon after a liquid layer. Several cases illustrating this phenomenon are presented in addition to long-term observations from three measurement sites characterizing cloud phase frequency. This analysis demonstrates that clouds composed entirely of ice occur less frequently than liquid-topped mixed-phase clouds at temperatures warmer than −25 to −30 C. These results indicate ice formation generally occurs in conjunction with liquid at these temperatures, and suggest the importance of liquid-dependent ice nucleation mechanisms.
 Arctic climate is changing at a rate exceeding that of other regions [Symon et al., 2005]. Quantitative predictions of high-latitude changes, and impacts on future climate, rely on sophisticated climate models. Despite expanded efforts, models still struggle to accurately portray cloud processes, resulting in surface and atmospheric radiative budget errors [e.g., Klein et al., 2009]. Of observed high-latitude clouds, mixed-phase stratiform (MPS) layers are among the most radiatively influential [e.g., Shupe et al., 2006].
 Modeling studies [e.g., Harrington and Olsson, 2001] show MPS lifetime depends strongly on ice concentration due to efficiency of the Bergeron-Findeissen process [Pruppacher and Klett, 1997]. Therefore, understanding ice nucleation in these layers is imperative. Heterogeneous ice formation at temperatures (T) > −40 C may occur via several primary modes, namely deposition, condensation, immersion and contact freezing. In deposition freezing, ice-supersaturated conditions result in deposition of water vapor onto ice forming nuclei (IN), forming ice crystals. Condensation freezing is similar, except aerosol particles (hereafter particles) involved are less efficient IN, possibly containing both soluble and insoluble mass. These mixed particles are observed at high-latitudes, often as dust or soot and sulfates or organics [e.g., Bigg and Leck, 2001]. Soluble mass on these particles requires an environment approaching water saturation to undergo sufficient wetting to nucleate ice. Immersion freezing is comparable to condensation freezing, except droplets grow to super-critical sizes (transitioning from haze to cloud droplets) before freezing. Finally, contact nucleation involves collision of existing water droplets with IN.
 Several recent theories on Arctic MPS ice formation have been presented, including immersion nucleation [de Boer et al., 2009b], evaporation ice nucleation [Fridlind et al., 2007], and contact nucleation [Morrison et al., 2005]. Additionally, Prenni et al.  presented in-situ observations indicating immersion or condensation nucleation are important mechanisms within these clouds. Because MPS are observed in T between −40 and 0 C [Shupe et al., 2006; de Boer et al., 2009a], no heterogeneous nucleation mechanism can be ruled out due to conditions alone. However, observations demonstrate patterns in timing of liquid and ice formation, providing clues toward understanding ice nucleation. We present observations demonstrating instances when ice, despite having a lower saturation vapor pressure than liquid, is not formed without presence of liquid. Statistics of observed cloud phase are provided to generalize these findings.
 Observations come from a combination of Arctic surface-based remote sensors. Included are data from a 35 GHz Millimeter Cloud Radar (MMCR) [Moran et al., 1998] and microwave radiometer (MWR) at the US Department of Energy (DOE) Atmospheric Radiation Measurement (ARM) facility in Barrow, Alaska, the Study of Environmental Arctic Change (SEARCH) site in Eureka, Canada and the Beaufort Sea Surface Heat Budget of the Arctic (SHEBA) [Uttal et al., 2002] campaign. Additionally, Wisconsin Arctic High Spectral Resolution Lidar (AHSRL Eureka and some Barrow) [Eloranta, 2005], micropulse lidar (MPL, Barrow) and Depolarization and Backscatter Unattended Lidar (DABUL, SHEBA) [Grund and Sandberg 1996] measurements were used. Observation periods include 12/06/07-10/01/98 (SHEBA all times with active MWR), 02/04-02/06 (Barrow) and 08/05-10/09 (Eureka). Because the MMCR and lidars operate at different wavelengths, their measurements respond differently to various hydrometeors. In MPS, lidars are more sensitive to the large combined cross-sectional area from high concentrations of liquid droplets [McFarquhar et al., 2007], while the MMCR signal is dominated by fewer but larger ice crystals. Lidar depolarization measurements from the AHSRL and DABUL help distinguish between non-depolarizing particles (e.g., spheres) and depolarizing particles. In addition to remotely sensed measurements, T and relative humidity (RH) are from radiosondes (sondes). These sondes may exhibit dry biases at low T [e.g., Miloshevich et al., 2009]. However, given the range of sonde models used no corrections are applied to presented measurements, in part to avoid confusion about their influence on RH profiles. The (possibly dry-biased) profiles presented provide adequate evidence that the atmosphere approached or exceeded ice saturation in clear-sky conditions.
3. Methods and Results
 Qualitative and statistical evaluations of ice presence in MPS clouds are presented. Six cases (02/27/06, 02/26/07 and 10/11/07 – Eureka; 11/02/04 and 11/05/04 – Barrow; and 03/10/98 – SHEBA) are examples of a more broadly observed phenomenon. These cases illustrate MPS at multiple locations and times of year. Provided in Figure 1 are lidar backscatter cross-section (or backscatter for DABUL) (β′), lidar depolarization ratio (δ), radar reflectivity (Zmmcr) and T and RH profiles. T and RH are from sondes launched at the time indicated by orange bars. For 10 March, DABUL was not operating at 00Z, but MMCR reflectivity showed no sign of clouds or precipitation.
 All cases illustrate similar patterns in appearance and evolution of an MPS layer. Low β′ is indicative of clear air. This layer is exposed to cold (−25 C ≤ T ≤ −10 C) and humid (RHice generally ≥ 100%) conditions at the time of sonde measurements. When approaching liquid saturation, aerosol particles swell to form haze, increasing β′ and decreasing δ. Haze is distinguishable from small ice through lidar δ, with observed decreases in δ due to the sphericity of haze droplets. With cloud droplet formation, there is a dramatic increase in β′ and soon after the cloud develops, increased in- and below-cloud Zmmcr and below-cloud δ indicate presence of ice. For cloud droplet formation RHliq must reach 100%, requiring atmospheric cooling and/or moistening from the sonde profiles, both favorable for ice formation and growth. Despite this, no ice is detected prior to the appearance of liquid. Additionally, 11/02/04 illustrates a lack of ice after liquid disappears, despite RHice between 100 and 115%. It should be noted that while among the most sensitive radars, the MMCR cannot detect newly nucleated ice that has not undergone growth. Assuming small spherical ice at concentrations of 1 L−1, the minimum detectable reflectivity (≈−55 dBZ at low altitudes) would occur when crystals grow to roughly 50 μm. Under typical conditions from presented profiles (T = −20 C and RHice = 110%), a capacitance growth model predicts ice with a 5 μm radius to grow to this size in roughly 790 seconds. At RHice closer to liquid saturation (roughly 122% for T = −20 C), this growth would occur even faster. Assuming a 10 cms−1 average fall speed, this crystal would fall 79 meters during that time, keeping it within ice-saturated conditions (ranging between 200 and 1600 m for presented cases).
 To generalize these findings, hydrometeor phase under varying T and RH is analyzed. Remotely sensed data are used to derive cloud phase using a fixed-threshold, multi-sensor classification. This classification and thresholds used are presented by Shupe . Profiles are limited to 20 minutes surrounding sonde launches and conditions with no hydrometeors above four kilometers. Minimum T (Tmin) reported is either the lowest in-cloud T (when cloudy) or lowest T below four kilometers altitude (when clear), while reported RH with respect to ice (RHice) represents the highest value below four kilometers. The “Mixed” category represents the structure of interest (liquid topped MPS), while “MixCol” represents ice-topped clouds with embedded mixed-phase layers.
Figure 2 illustrates normalized probabilities of analyzed phase occurrence with respect to Tmin (Figures 2, top and 2, middle) and RHice (Figure 2, bottom). Both distributions for all classifications (Figure 2, top), and for ice only, liquid only and “mixed” (Figure 2, middle) are shown. Barrow and SHEBA have few cases where Tmin < −40 C below 4 km. At all locations, MPS are the most common cloud type roughly when −10 C < Tmin < −25 C. Not surprisingly, liquid clouds are more common than ice containing ones at Tmin > −10 C, and ice clouds are more common than liquid-containing ones when Tmin < −30 C. Interestingly, for Barrow and SHEBA MPS are more common than ice clouds at all RHice. This is not true at Eureka, due mainly to a large number of ice cloud cases at low Tmin. These statistics indicate that over a range of ice-friendly conditions (−25C < Tmin <−10 C, RHice ≥100%) MPS are dominant.
Figure 3 shows scatter plots of phase in Tmin-RHice space. Liquid and MPS cases roughly fall in line with the Clausius-Clapeyron relation at water saturation (dashed grey line). Observations falling above this line are likely a result of Tmin and RHice definitions used and radiosonde errors. Similarly, presence of ice and MPS at RHice < 100% is likely explained by radiosonde dry-biases discussed previously or radiosonde drift with altitude. All locations show similar distributions of MPS at −25 C < Tmin < −10 C. At warmer Tmin, Barrow appears to have more liquid clouds than SHEBA or Eureka, while Eureka has more ice clouds at colder Tmin. Eureka ice clouds mainly occur when Tmin < −25 C. Eureka has more clear cases, occurring at all T at low RH. With few exceptions, ice is present by itself only in situations where Tmin < −25 C and 100% ≤ RHice < liquid saturation. Under these conditions, little liquid is observed, likely due to consumption of water vapor by ice growth.
4. Discussion and Summary
 These observations demonstrate several interesting phenomena. First, as observed by Kanji and Abbatt , aerosol layers detected at −20 C under ice-supersaturated conditions do not readily form ice via the deposition mode. This may be due to aerosol chemistry, though this can not be proven with the sensors employed. This pattern appears robust over a range of seasons and locations. Appearance of ice once liquid is established suggests liquid-related ice nucleation. While responsible pathways can not be determined from these measurements alone, it supports liquid-dependent mechanisms such as contact nucleation, immersion freezing, or evaporation freezing, and falls in line with previous in-situ analyses [e.g., Prenni et al., 2009; Rangno and Hobbs, 2001] and discussions on super-cooled liquid [e.g., Rauber and Tokay, 1991]. Condensation and deposition freezing do not appear to readily occur prior to cloud droplet nucleation, since haze layers are observed with no evidence of ice formation and growth. While utilized measurements are incapable of proving inactivity of condensation and deposition freezing after liquid forms, it seems unlikely that these mechanisms would be initiated and dominate right at liquid saturation.
 Long-term statistics presented, while not directly providing phase order of appearance, supply evidence that ice clouds occur far less frequently than MPS at T > −25 C. We speculate that ice-first morphology via deposition or condensation freezing regimes would result in increased ice-only observations, due in part to an ice-induced sink of atmospheric water vapor limiting cloud formation. A lack of ice-only cases together with individual examples of morphology suggests that liquid droplets may be an important driver for ice nucleation at T > −25 C. This is consistent with findings of Ansmann et al.  who showed a strong relationship between initiation of ice and presence of liquid in tropical altocumulus, Hobbs and Rangno  demonstrating links between presence of liquid droplets and increased ice amounts from analysis of in-situ measurements, Cooper and Vali  describing limited effectiveness of deposition freezing in continental wave clouds and Ohtake et al.  who assessed ice forming in the lower Arctic atmosphere to mainly result from droplet freezing. The idea of liquid-dependent ice nucleation also supports work by Rangno and Hobbs  who demonstrated a relationship between ice concentration and droplet threshold diameter for Arctic clouds. Generalization of these findings to global scales will require further observational analysis, including a review of satellite data.
 Finally, these observations have implications on simulation of ice nucleation in models of all scales. Many microphysical schemes allow nucleation of ice below water saturation when −25 C ≤ T ≤ 10 C [e.g., Lin et al., 1983; Hong et al., 2004; Morrison and Gettelman, 2008]. Observations presented imply that limiting ice nucleation when RHliq < 100% and −25 C ≤ T ≤ 0 C (or conversely increasing ice nucleation at RHliq = 100%) may be prudent, and support improvement of parameterizations of immersion and contact freezing. Additionally, possible impacts of aerosol chemistry makes improved aerosol modules important contributors to realistic microphysical simulation.
 The authors acknowledge E. Eloranta, J. Hedrick, J. Garcia and I. Razenkov for AHSRL data and R. Moritz for advice on SHEBA sonde measurements, as well as DOE, NOAA and SEARCH teams for MMCR data. Information from Eureka weather station personnel was helpful in dataset compilation. Data were also made available by the US DOE ARM program, and the SHEBA team. Finally, we acknowledge funding from the National Aeronautics and Space Administration (NASA: NNX07AQ81G) and US DOE (U.S. DOE: ER64187-1027586-0011923 and DE-FG02-05ER63965). NCAR is sponsored by the National Science Foundation. LBNL is managed by the University of California under U.S. DOE grant DE-AC02-05CH11231.