A continuous flow diffusion chamber (CFDC) was used to measure ice formation by cloud particle residuals during the Cirrus Regional Study of Tropical Anvils and Cirrus Layers-Florida Area Cirrus Experiment. These measurements were directed toward determining the relative contributions of homogeneous nucleation, heterogeneous nucleation, and secondary ice formation processes to the concentrations of ice crystals in anvil cirrus formed from convection. The CFDC sampled residual particles remaining after evaporation of cloud particles initially collected by a counterflow virtual impactor. This allowed, for the first time, determination of the ice nucleation ability of particles that included the presumed nuclei for cloud-ice formation. The approach proved successful for estimating concentrations of heterogeneous ice nuclei (IN) transported into anvil clouds, but experimental issues limited measurements of homogeneous freezing and, consequently, in determining the role of secondary ice formation. Results suggest agreement within a factor of 2–3 between CFDC heterogeneous IN concentrations and anvil ice crystal concentrations in the size range above ∼30 μm. IN concentrations also correlated with ice concentrations inferred from measurements by the FSSP (Forward Scattering Spectrometer Probe). However, measured IN concentrations were nearly two orders of magnitude lower than FSSP concentrations. This difference may have resulted from homogeneous freezing, secondary ice formation, or other unidentified ice formation processes that were not fully captured by the CFDC. The data suggest that heterogeneous nucleation played a smaller role than homogeneous nucleation in determining anvil ice crystal concentrations, except during periods of strong desert dust ingestion by cumuli. Nevertheless, heterogeneous nucleation may provide the source for larger ice crystals present in anvil regions.
If you can't find a tool you're looking for, please click the link at the top of the page to "Go to old article view". Alternatively, view our Knowledge Base articles for additional help. Your feedback is important to us, so please let us know if you have comments or ideas for improvement.
 In July 2002, a continuous flow ice thermal gradient diffusion chamber (CFDC) was employed to measure the ice nucleating ability of aerosol particles during the Cirrus Regional Study of Tropical Anvils and Cirrus Layers-Florida Area Cirrus Experiment (CRYSTAL-FACE). This measurement campaign investigated the formation processes and physical properties of low latitude cirrus clouds, specifically cirrus anvils formed from deep convection [Jensen et al., 2004]. The CFDC measurements were intended to provide insight into the relative importance of different ice formation processes in cumuli and the cirrus anvils they produce. Cirrus play a dual role in the Earth’s radiation budget, scattering incoming solar radiation while simultaneously trapping infrared radiation from the surface and lower atmosphere. The net effect of cirrus on surface temperature depends on the physical properties of the clouds, which in turn can be affected by the mechanism by which they form [Haag and Karcher, 2004]. Tropical cirrus can also affect climate through their role in the water vapor budget; specifically, the amount of water transported to the upper troposphere is driven by deep convection, but limited by precipitation processes. Many studies of cirrus formed in situ, by various scales of uplift from mesoscale to synoptic scale, have been reported in the literature [DeMott et al., 1998; Heymsfield and Miloshevich, 1993; Jensen et al., 2005, 2001; Sassen and Dodd, 1988]. These studies have explored the competing roles of heterogeneous versus homogeneous freezing [DeMott et al., 1997; Gierens, 2003] and the interplay with cloud dynamics that results when nucleation competes with the diffusional growth of ice at lower temperatures [Heymsfield et al., 2005]. Fewer studies have considered cirrus anvil formation [Garrett et al., 2005; Jensen and Ackerman, 2006], the focus of this study. In this case, cloud properties are influenced by the full range of primary ice formation processes, including heterogeneous nucleation in the source cumulus cloud, homogeneous freezing of liquid cloud droplets as they are cooled to temperatures below about −38°C, and homogeneous freezing of haze particles at even lower temperatures. Also, potentially active are known secondary ice formation processes, including ice splinter formation during the riming of ice particles at temperatures between −3 and −8°C [Hallett and Mossop, 1974], mechanical fracturing of ice crystals during evaporation [Dong et al., 1994], and generation of ice fragments by crystal-crystal collisions [Vardiman, 1978], as well as unidentified primary and secondary ice formation processes.
 Ice nucleation can occur at any temperature below 0°C by heterogeneous nucleation on some aerosol insoluble component [Vali, 1985]. The particles on which ice formation occurs in this case are known as ice nuclei (IN). Heterogeneous IN are presumed to be vital to the initiation of ice in mixed-phase clouds and may affect cirrus cloud formation as well. Four heterogeneous ice nucleation processes have been identified: deposition, in which water vapor adsorbs as ice to the IN surface typically below water saturation; condensation-freezing, in which ice forms as supercooled liquid water condenses on a cloud condensation nucleus (CCN); immersion, in which a water droplet (containing previously activated CCN) freezes as it cools; and contact freezing, in which a supercooled droplet freezes when an IN comes in contact with its surface. Heterogeneous IN can initiate the first ice formation, act to broaden ice crystal size distributions, and lower maximum concentrations of ice particles in clouds [DeMott et al., 1998]. At colder temperatures, mostly below −38°C, homogeneous freezing of ice in activated liquid droplets [Heymsfield and Miloshevich, 1993; Sassen and Dodd, 1988] or haze particles [Karcher and Lohmann, 2002; Koop et al., 2000] becomes important. The precise temperature at which homogeneous freezing occurs is dependent on droplet size and composition. Both measurements [Garrett et al., 2003; Twohy and Poellot, 2005] and modeling [Heymsfield et al., 2005; Phillips et al., 2005] studies show strong evidence for homogeneous freezing at low temperatures during CRYSTAL-FACE.
 Several authors have used CRYSTAL-FACE data together with numerical model simulations to investigate the role of different ice formation mechanisms in determining anvil cirrus properties [Fridlind et al., 2004; Phillips et al., 2005; van den Heever et al., 2005], specifically considering aerosol particle inputs. van den Heever  found that variations in heterogeneous IN, CCN, and giant CCN concentrations can have significant effects on both dynamical and microphysical processes of convective storms that form over Florida. Specifically, cloud-resolving model simulations showed that increased aerosol concentrations at levels up to 4 km lead to stronger and more numerous updrafts, while increased IN concentrations produce ice at warmer temperatures, produce deeper anvils and accelerate precipitation processes. Cloud model simulations by Fridlind et al.  and Phillips et al.  emphasized the role of secondary activation by aerosols entrained into cumuli from the middle troposphere on impacting a predominant homogeneous freezing process and thereby determining the ice crystal distributions in some strong convective clouds over the Florida area. These results were supported by detailed analyses of cloud microphysical data [Heymsfield et al., 2005], but this latter study also pointed out that the more dominant homogeneous freezing process in initial cloud turrets impacts ice formation processes at later times or in subsequent cloud turrets.
 Physical studies of cloud particle residues [Cziczo et al., 2004; Twohy and Poellot, 2005] have provided further constraints on understanding potential contributions of heterogeneous and homogeneous ice formation processes in anvil cirrus formation during CRYSTAL-FACE. Using counterflow virtual impactors (CVI) to sample anvil cloud particles and analyze their residual nuclei, both studies confirmed the presence of soluble particles as residual nuclei, especially sea-salt particles of likely marine boundary layer origin, suggesting the ubiquitous role of homogeneous freezing as an ice formation mechanism. Cziczo et al.  also inferred that the largest of these soluble particles in the accumulation mode preferentially participated in ice formation. Twohy and Poellot  analyzed residual particle compositions to smaller sizes using transmission electron microscopy and found that while particles larger than a few tenths of a micron were often equally soluble and insoluble, in agreement with Cziczo et al. , smaller aerosols were often predominately soluble and dominated inferred anvil ice formation. Still, there were occasions when residual particle compositions (mineral dusts, metals, black carbon) suggested that heterogeneous ice nucleation at least equally participated in anvil ice formation. This was particularly the case when high-mass loadings of Saharan dust were present at lower altitudes and apparently ingested into clouds.
 Absent in previous analyses from CRYSTAL-FACE has been direct constraint of the role of heterogeneous ice nucleation on anvil ice formation on the basis of the ice activation properties of ambient aerosol particles. Providing this evidence is the aim of this study based on processing of cloud residual nuclei from the same CVI inlet used by Twohy and Poellot .
2. Experiments and Methodology
 Ice nuclei concentrations were measured using the CSU continuous flow ice thermal gradient diffusion chamber. The instrument was installed on the University of North Dakota Citation II aircraft, which profiled aerosol and cloud characteristics up to 13 km above mean sea level (MSL). The CFDC permits observation of ice formation on a continuous stream of aerosols at controlled temperatures and humidities [Rogers, 1988; Rogers et al., 2001b]. This technique has been used previously in the field to study both heterogeneous [Rogers et al., 2001a, 1998] and homogeneous ice nucleation [DeMott et al., 2003a]. The CFDC is sensitive to all primary nucleation modes, except contact freezing, since the residence time is fairly short. The processing section of the CFDC consists of an annular gap between two vertically oriented ice-coated cylinders. Dry, particle-free sheath air (70–90% of total) constrains the aerosol to a narrow lamina in the annular region. Processing temperature and relative humidity are determined by the temperatures of the ice-covered walls and the location of the lamina [Rogers, 1988]. Particles which form ice grow preferentially, due to the high supersaturations experienced by ice crystals compared to liquid particles. The size differential between ice crystals and aerosols at the outlet of the CFDC, as measured by an optical particle counter (OPC), is the basis for detecting ice formation. An inlet impactor upstream of the CFDC ensures that aerosol particles larger than ∼1.5 μm (aerodynamic diameter) are removed prior to entering the instrument [Rogers et al., 2001b], so that large aerosol particles are not erroneously identified as ice. The bottom third of the chamber has no ice on the warm wall and serves as an evaporation region for liquid particles. This region allows for operation of the CFDC above water saturation, in that it ensures that activated water droplets evaporate prior to reaching the OPC and so are not mistaken for ice. An inertial impactor immediately downstream of the CFDC is used to capture ice crystals on electron microscope (EM) grids, allowing for identification of the chemical composition of the particles on which ice forms [Kreidenweis et al., 1998]; EM analyses are limited to particle diameters above about 50 nm. IN concentrations are reported as 30 s running means and concentrations are given at the temperature and pressure at which the data were collected; ∼500 cm3 of air is sampled per 30 s.
 Aerosol particle sampling during CRYSTAL-FACE was done by two methods. An ambient air inlet was used in some cases, as described by DeMott et al. [DeMott et al., 2003b]. For the measurements reported in this paper, the CFDC was interfaced with a CVI [Twohy et al., 1997]. We elaborate now on this procedure and describe how the method can be applied for studies of ice formation processes.
 In the CVI, cloud particles greater than a “cut size” (a 50% cut size of 4 μm aerodynamic diameter for typical Citation flight conditions [Anderson et al., 1993; Noone et al., 1988]) are separated from the interstitial aerosol and water vapor. This separation is accomplished using a counterflow stream of gas that is supplied through a porous tube inside the CVI tip, which only larger (cloud) particles are able to penetrate. Droplets or crystals are evaporated within the inlet at a temperature of about 50°C, and the water vapor and nonvolatile residual nuclei remaining after droplet evaporation are measured downstream of the inlet. Condensed water content (CWC) is determined by measuring the resulting water vapor using a Lyman-α hygrometer. During CRYSTAL-FACE, some of the residual aerosol particles were collected by inertial impaction onto electron microscope grids for chemical composition analyses [Twohy and Poellot, 2005] and a fraction of the residual particles (∼10% of the flow) were sent to the CFDC. Both the CVI and the CFDC are operated near ambient pressure (stagnation pressure). While transiting between the instruments, the residual aerosol particles remain at cabin temperature and pass through two diffusion driers to reduce the water vapor concentration prior to cooling in the CFDC. The separation process in the CVI also enhances the number concentration of the cloud particles by about a factor of 25, because droplets or crystals in a large sampling volume are impacted into a relatively small sample stream. This enhancement is helpful in detecting IN, which typically are found in low abundance. Using these two instruments in series allows for measurement of ice nucleation by the residual particles as a function of temperature and water saturation. The potential for breakup of ice crystals in the CVI and the related impact of residual particle concentrations on CFDC measurements has not been fully explored. However, Twohy and Poellot  found no evidence for enhanced concentrations of residual particles larger than 0.1 μm diameter relative to ice crystal number concentrations measured by an FSSP-100 and 2D-C in CRYSTAL-FACE. Similarly, IN concentrations measured by the CFDC behind the CVI were less than or approximately equal to ice crystal concentrations from the particle probes. Thus, while crystal breakup may occur when large crystals are present [Korolev and Isaac, 2005], it does not seem to produce substantial enhancements in residual nuclei as measured by the CFDC.
 The measurement strategy for cosampling CVI residual nuclei with the CFDC was aimed at identifying the different processes involved in cumuli and cirrus anvil ice formation. This strategy is depicted in Figure 1. The intent is to alternately (for example, on successive cloud passes) expose residual particles from anvil clouds to conditions for which the maximum contribution of heterogeneous freezing is expected to have occurred (warmer than about −38°C) and for which homogeneous freezing is expected to dominate (colder than about −38°C and above water saturation). Processing residual cloud particles over a range of temperatures and humidities can provide additional insight into freezing mechanisms, as shown in the inset of Figure 1. The arrows indicate how temperature and humidity in the CFDC might be varied to examine heterogeneous ice nucleation by deposition and condensation freezing and homogeneous freezing of either cloud droplets or haze particles. The sum of concentrations of ice crystals activated on the residual particles under these different processing regimes then serve to define primary heterogeneous nucleation and homogeneous nucleation contributions to anvil ice concentrations. These measurements can then be compared to ice particle concentrations inferred from the cloud particle measurements, any differences being due to the action of known secondary ice formation processes or unknown ice formation processes of primary or secondary type. The primary cloud particle instruments used on the Citation aircraft for this study included a Particle Measuring System Optical Array Probe (OAP-2DC) and a Forward Scattering Spectrometer Probe (FSSP-100). The 2DC images particles over a size range of ∼30 μm to just less than 1000 μm. The FSSP is designed for measuring cloud droplets in the range of 3.5–58 μm. Its response to small ice crystals is not well determined, but FSSP concentrations will be used to estimate small ice crystal concentrations in glaciated cloud regions in this study. It should be noted that the FSSP may be prone to overestimating particle concentrations in the presence of ice [Field et al., 2003; Gardiner and Hallett, 1985] due to ice breakup of large particles in the inlet [Heymsfield et al., 2004]. The 2DC has also been shown to overestimate particle concentrations, particularly for cases in which large particles are present [Field et al., 2006]. Measurements of liquid water content (LWC) from a King probe and liquid water presence from the Rosemount icing detector are also discussed. The Rosemount probe measurement provides an output voltage, which can be calibrated for LWC using the method of Mazin et al. . The LWC detection threshold obtained by this method is very low, about 0.01 g m−3 at the Citation’s airspeed [Heymsfield et al., 2005]. The sensitivity of the King Probe is on the order of 0.02 g m−3 [Korolev et al., 1998] but the instrument also responds to ice particles. An additional check on cloud ice concentrations was available over selected integrated periods from a Cloud Particle Imager (CPI) [Lawson et al., 2001]. Data reported from the CPI cover a size range of 20–1500 μm.
 While instrumental issues prevented the perfect implementation of the proposed strategy during CRYSTAL-FACE, we introduce the methodology due to its utility for future studies. The greatest obstacle to successfully implementing the strategy arose due to aircraft cabin heating resulting in temperatures as high as ∼50°C near the CFDC. These high temperatures, while mitigated somewhat during the course of the study, affected the performance of the instrument’s refrigeration compressors, which in turn made cooling to homogeneous freezing temperatures quite difficult and secondarily resulted in a temperature gradient along the cold inner wall of the CFDC at all temperatures. These temperature gradients, which ranged from 5 to 13.5°C (average gradient = 8.7°C, σ = 2.1°C), affect the ability of the instrument to achieve the desired steady state conditions for sufficient time for particles to nucleate and grow ice crystals to detectable size in the CFDC. Laboratory experiments were used as a basis for defining an “effective” cold wall temperature under the influence of a temperature gradient. These experiments involved comparison of homogeneous freezing conditions of monodisperse ammonium sulfate particles, both with and without a temperature gradient on the cold wall [Richardson et al., 2007]. The “effective” cold wall temperature was defined by the experiments without a temperature gradient and was correlated with high confidence to the average cold wall temperature with a gradient in place. Although laboratory experiments were only able to reproduce a 4°C gradient, this relationship was used as a first order correction for recalculating the processing temperatures and humidities in the CFDC during CRYSTAL-FACE. The cold wall temperature correction ranged from 1.0–1.7°C warmer than the original average wall temperature in the −33 to −56°C temperature range. On the basis of this correction, aerosol-processing conditions assuring homogeneous freezing were achieved during only a few select periods throughout the project. Consequently, we were unable to do a comprehensive evaluation of the contribution of homogeneous freezing of cloud droplets to the anvil ice crystal populations throughout the study, as outlined in Figure 1. Furthermore, since the correction applied is based on laboratory studies conducted with a smaller gradient than existed during CRYSTAL-FACE, the reported conditions for the CFDC are thought to represent the lower and upper limits for temperature and humidity, respectively, and the conditions described below may underestimate temperature and overestimate humidity.
3. Results and Discussion
 CFDC data were collected on 18, 19, 21, 23, 25, 28, and 29 July 2002. We focus on the data for which the CFDC was integrated with the CVI, thereby omitting all measurements from 18 and 28 July. We start by assessing data collected in proximity to the homogeneous freezing regime for cloud drops and then consider measurements of heterogeneous IN made at lower temperatures typical of cirrus. In both cases, comparison is made to cloud properties and to residual nuclei chemistry to give inference to ice formation processes. We summarize with composite analyses of the data set on residual nuclei concentrations, size, and chemistry.
3.1. Measurements in the Transitional Regime Between Heterogeneous Freezing and Homogeneous Freezing of Liquid Drops
 While homogeneous freezing can occur on any soluble particle, heterogeneous IN are typically found in much lower concentrations. Previous CFDC measurements of ambient particles have shown greatly enhanced ice formation for homogeneous freezing conditions (Tcfdc < −38°C) relative to IN concentrations measured for heterogeneous nucleation conditions (Tcfdc > −35°C) [DeMott et al., 2003a]. For this study, enhanced CFDC concentrations were expected for in-cloud measurements in which soluble nuclei were present and the CFDC was colder than −38°C and at or above water saturation. Figure 2 shows CFDC measurements of cloud particle residuals as a function of processing temperature and separated according to processing humidity. As noted above, conditions that are supersaturated versus subsaturated differentiate homogeneous freezing of cloud droplets versus haze particles. Data collected on 29 July are omitted, as this flight was heavily influenced by dust [DeMott et al., 2003b; Sassen et al., 2003], and enhanced IN concentrations were observed at all humidities (see later in this paper). As can be seen in the figure, conditions for which temperatures were colder than −38°C and above water saturation were rarely achieved (<3% of the measurements). Furthermore, the few data points meeting these criteria occurred during a time when the CFDC had a temperature gradient of 10°C, and so the humidity correction applied likely overestimates processing supersaturation; the humidity actually attained in the instrument may, in fact, have been subsaturated. However, particles processed above water saturation at temperatures between −35 and −38°C, where ice formation in the CFDC transitions from being dominated by heterogeneous nucleation to being dominated by homogeneous nucleation, show modest enhancements in concentrations, indicating the onset of homogeneous freezing conditions. These data represent the total number of observations for which homogeneous freezing of cloud droplets occurred in the CFDC. Clearly there are not enough measurements to unambiguously quantify the total contributions of homogeneous freezing of cloud droplets to ice particle number concentrations in anvil clouds in even a single cloud case.
 CFDC measurements collected in the transition temperature range (−35 to −38°C) can be compared to measurements made by the other cloud probes. We focus first on data collected on 19 July 2002. Data on this day are responsible for the highest CFDC concentrations shown in Figure 2, when processing conditions nudged into the homogeneous regime. Figure 3a shows CFDC processing temperature, CFDC water supersaturation, ambient temperature, and altitude as a function of time (s after midnight UTC) for data collected in an anvil on 19 July 2002. The sounding includes a stepwise descent followed by a continuous spiral ascent, covering an altitude range of approximately 8 to 11 km. From a static standpoint, the ambient temperature conditions encompass both heterogeneous and homogeneous freezing regimes. While the ice crystals sampled by the CVI and CFDC likely did not form at the altitude (temperature) at which they were observed, and vertical transports complicate attempts to link ambient temperature to freezing processes, we expect that homogeneous freezing, if active within cloud updrafts, would contribute more significantly to ice crystal concentrations detrained at higher altitudes (colder temperatures). Supporting evidence for this expectation has been presented by Twohy and Poellot  who found the fraction of soluble material in cloud particle residual nuclei from the CVI to increase with altitude in the overall CRYSTAL-FACE data set. Figure 3b shows measurements of ambient condensation nuclei (CN), concentrations of cloud particle residual nuclei which induce ice nucleation in the CFDC and CWC from the CVI. During the descent (77,500–79,050 s), CFDC processing supersaturation with respect to water (Ssw) varies between −5% (95% RH) to nearly 7%, while CFDC processing temperature remains warmer than −36°C. For these conditions in the CFDC, heterogeneous nucleation is expected to dominate. IN concentrations vary, but remain near 0.01 cm−3, comparable to previous measurements under similar conditions [DeMott et al., 2003a]. At ∼79,500 s, processing temperatures drop below about −36.2°C while processing humidity remains above water saturation, and CFDC concentrations increase by an order of magnitude, to ∼0.1 cm−3. These observations indicate either a change in the aerosol sampled or a change in nucleation mechanism. During this time, CN and CWC measurements remain roughly constant, and so we expect that the change in measured IN concentrations resulted from a change in freezing mechanism in the CFDC. Thus, for these conditions, −36.2°C and ∼1% water supersaturation, homogeneous freezing was detected for the first time in the CFDC from cloud particle residuals from an aircraft platform.
 We now consider these measurements in the context of the other cloud probes. Figure 3c shows data from the FSSP and 2DC. Throughout the descent (77,500–79,000 s), when the CFDC was limited to measuring heterogeneous nucleation, ambient temperatures remained warmer than −35°C, and so freezing in the atmosphere was also expected to have been due, at least in part, to heterogeneous nucleation. During this time, measured IN concentrations were comparable in number to ice particle concentrations measured by the 2DC, suggesting that much of the ice measured by the 2DC resulted from heterogeneous nucleation. In contrast, FSSP concentrations presumed to be ice based on measured LWC below the sensitivity limit of the King and Rosemount liquid water probes, were nearly two orders of magnitude greater than IN concentrations. These particles are thought to be ice crystals which have formed from homogeneous nucleation at some colder temperature or by some secondary process. In either case, measured IN number concentrations cannot explain the high number concentrations of the FSSP. During the initial part of the ascent, 79,500–79,700 s (approximately 8000–9000 m), ambient temperatures are still warmer than about −30°C and the CFDC has cooled to below −36.2°C. At this processing temperature, both heterogeneous nucleation and some homogeneous nucleation are expected in the CFDC. Specifically, particles of sizes sufficient to activate as droplets at ∼101% RHw (pure solutes larger than ∼50 nm) would be expected to freeze homogeneously in a time- and droplet volume-dependent manner. Given the relatively slow volumetric homogeneous freezing rate of pure water at this temperature (108 cm−3 s−1 maximum value [DeMott and Rogers, 1990]), only a fraction of these particles will have sufficient time to activate in the CFDC. Using the FSSP concentrations to constrain the number of particles sampled by the CFDC, assuming that all of these particles are potential nuclei for homogeneous freezing, and making reasonable assumptions on the activated droplet diameter (variable, but 3 μm is representative based on calculations), we estimate a homogeneous freezing contribution of about 30 per liter in the CFDC at −36.2°C when FSSP concentrations were 2 cm−3. The average measured CFDC ice concentrations were about 80 per liter, consistent with contributions occurring from both heterogeneous and homogeneous nucleation during this flight period.
 From approximately 79,700–80,000 s (9000–10,500 m), ambient temperatures decrease from −29 to −40°C. At the same time, cloud particle measurements from the 2DC and FSSP and CWC from the CVI increase dramatically, possibly from an increased contribution from homogeneous nucleation in the cloud. On the basis of a measured FSSP concentration of 15 cm−3, we expect homogeneous freezing nuclei concentrations of 225 per liter in the CFDC for the processing conditions, while measured concentrations were ∼150 per liter. If soluble nuclei smaller than ∼50 nm contribute to ice in the cloud, the CFDC would likely not capture freezing of these particles, due to the high supersaturation needed to activate these small particles as droplets in the chamber. Finally, as the aircraft continues to climb to above 10,700 m (80,000–80,500 s), a drastically different behavior is observed. In this higher altitude cloud layer with ambient temperatures below −40°C, the other probes continue to measure ice while concentrations measured by the CFDC are generally less than 0.001 cm−3, with slightly higher concentrations observed near the top of the cloud layer. It should be noted that above ∼10,600 m the aircraft spiral encompasses a much larger area. It appears that very few of the aerosol particles responsible for the clouds above 10,700 m serve as effective nuclei for freezing in the CFDC at −36.2°C. Either the residual nuclei were too small to initiate homogeneous ice formation at the processing conditions set in the CFDC or there was an absence of heterogeneous IN at these high altitudes.
 In contrast, data collected on 23 July are for CFDC processing temperatures equal to or warmer than −36.2°C, with CFDC processing humidity below water saturation. These data are comparable to data from the descent on 19 July in that ice nucleation measured by the CFDC instrument is expected to be heterogeneous. Ice nucleation in this case is by deposition nucleation. Figure 4 shows data from a descent sounding through anvil cirrus on 23 July. Throughout the time period shown, ambient temperatures were warmer than −35°C. Some liquid water was detected by the King Probe in the clouds (<0.03 g m−3), particularly at ambient temperatures warmer than −25°C. However, during this time condensed water content from the CVI, which collects both liquid and ice particles, indicated peak values >0.56 g m−3, suggesting that much of the cloud was ice, and not liquid. Ice in this anvil was likely produced over a range of temperatures and potentially by all primary and secondary processes. The FSSP data suggest that numerous small ice particles, or potentially supercooled water, were present for a temperature regime where heterogeneous ice nucleation would otherwise have been expected to dominate. Heterogeneous IN concentrations measured at below water saturation in the CFDC approximate ice formation observed by the 2DC reasonably well. However, times are apparent when IN concentrations are much lower than 2DC concentrations. The particles which nucleate ice in these layers may not nucleate via deposition nucleation, suggestive of the heterogeneity of the type of IN in the atmosphere. It may also be the case that ice detected in these regions resulted from settling of cloud particles from colder layers where homogeneous freezing or unidentified secondary processes occur.
 The chemical composition of the sampled residual nuclei was characterized using transmission electron microscopy (TEM) with energy dispersive X-ray spectroscopy (EDS). TEM analysis was carried out on unprocessed CVI residuals, which represent all cloud particle residuals larger than 0.1-μm diameter, and on cloud particle residuals further processed in the CFDC. TEM data from the CFDC thus represent a subset of the cloud particles analyzed in the CVI: namely, those cloud particle residuals which induced ice nucleation at the temperature and humidity conditions of the CFDC. The data are categorized as number fraction of the particles which contain the following components: salts, including sodium and potassium salts; metal oxides/dust, which includes metal, metal oxide, and crustal dust particles; carbonaceous particles, including soot and organic species; sulfates; and mixed particles, which were crustal dust, oxide, or carbon particles that also included sulfate or salts. Particles which did not fit neatly into any particular category are listed as other. Results from TEM analysis of unprocessed CVI residuals are given in Table 1. These particles were collected prior to the descent on 23 July (temperatures colder than −38°C) and show compositions more representative of soluble material, indicative of homogeneous freezing processes. Also in Table 1 are results from TEM analysis of the residual nuclei which activated ice formation in the CFDC during the descent, for processing conditions aimed at heterogeneous nucleation. Note the marked contrast in residual nuclei composition collected in the CVI compared to those processed further in the CFDC. Enhanced metal oxides/dust in the CFDC processed particles is consistent with heterogeneous nucleation [Hung et al., 2003].
Table 1. Chemical Composition of Particles Sampled on 23 July, as Determined From TEM Analysisa
Metal Oxides/Dust, %
Dust Mix, %
Residual nuclei sampled by the CVI are from the second stage of the impactor, corresponding to particles 0.1 μm < Dp < 0.4 μm [Twohy and Poellot, 2005]. Sampling time for these particles: 79590-80180, and ambient temperature during the particle collection was −56°C. Residual nuclei which induced ice nucleation in the CFDC are for processing temperatures >−36.2°C. Sampling times are 80460-82800, and ambient temperature ranges from −56 to −5°C.
3.2. Residuals Processed at Anvil Temperatures
 CFDC processing of anvil residual nuclei at lower temperatures characteristic of anvil cirrus and in situ cirrus was performed on 2 days during CRYSTAL-FACE. These cases offer insight into the maximum concentrations of heterogeneous IN activated in clouds and transported to upper tropospheric regions. Homogeneous freezing may also occur on haze particles below water saturation at colder temperatures and it is feasible that this process continues to generate ice crystals in anvil regions. Two case studies are described, one a maritime cumulus anvil case and one in which strong Saharan dust influences were likely due to the presence of dust layers at lower altitudes.
3.2.1. Maritime Cumulus Case
 Particles were sampled from an anvil of a maritime cumulus cloud located to the east of the Florida peninsula on 25 July 2002. Residual nuclei were processed to examine ice nucleation activity at near −50°C and at RHw between 91 to 98% (see Figure 5a). As shown in the inset of Figure 1, these conditions will not capture homogeneous freezing of cloud droplets, but rather are focused on measuring homogeneous nucleation of haze particles. According to DeMott et al. [2003b], these processing conditions transected, but were mostly marginally below, the RH conditions required for detecting homogeneous freezing of natural particles in their haze states. Ambient temperatures ranged from −51 to −19°C during the descent, thus encompassing temperature regimes for which homogeneous and heterogeneous nucleation were expected to dominate. The Rosemount icing probe gave no indication of liquid water during this descent. Figure 5b shows measurements for the CFDC and condensed water content from the CVI, while Figure 5c shows measured ice concentrations from the FSSP and 2DC probes. The CFDC approximates the measured 2DC number concentrations reasonably well, particularly at lower altitudes. The 2DC measures larger ice particles, which may be biased toward initial ice formation occurring as the result of heterogeneous nucleation. In contrast, measured CFDC concentrations are nearly two orders of magnitude lower than those measured by the FSSP throughout the descent. Relative to the 2DC, the FSSP measures smaller ice particles, and may be biased toward ice which forms later in the cloud, presumably at the lower temperatures needed for homogeneous nucleation of cloud droplets. Secondary ice formation does not likely play a major role at these colder temperatures, or at least no such mechanisms have been identified. TEM data on the residual nuclei from CFDC ice crystals, collected throughout the descent, indicate that 43% of the particles were either metal oxides/dust or mixed particles, indicative of heterogeneous IN, while nearly 29% of the particles were carbonaceous, which may be involved in either heterogeneous or homogeneous ice nucleation [Prenni et al., 2001]. Given the good agreement between the 2DC and the CFDC, we conclude that much of the ice measured by the 2DC resulted from heterogeneous nucleation. In contrast, 14% of the nuclei which induced ice formation in the CFDC were salts and 7% were sulfates, suggesting that freezing of haze droplets below water saturation contributed at least 20% of the ice measured in the CFDC, while homogenous nucleation of cloud droplets is likely responsible for a greater fraction of the ice in the anvil [Phillips et al., 2005]. These data suggest that both homogeneous and heterogeneous nucleation participated in ice formation in the cloud. Furthermore, because they activated below water saturation in the CFDC, these nuclei likely activated early in the updraft, formed some of the largest ice in the clouds, and may have gone on to influence subsequent ice formation in the upper troposphere after the anvil evaporated. Indeed, Jensen et al.  have shown that the optical depth and radiative properties of subvisible cirrus in the upper troposphere may be very sensitive to IN concentrations.
3.2.2. Saharan Dust Case
 On 29 July 2002, the CFDC sampled residual particles from a cumulus anvil on mostly level flight legs between 8.5 and 10 km MSL. Ambient temperatures ranged from −25 to −37°C. Residual particles were processed at temperatures from about −38 to −49°C, but processing relative humidity exceeded water saturation only briefly. For these processing conditions, contributions from homogeneous nucleation are expected for freezing of haze particles in the CFDC, assuming that the cloud particle residuals contained soluble nuclei. However, as noted above, some of the highest concentrations of heterogeneous IN measured from ambient air during CRYSTAL-FACE were observed on this date and during sampling in Saharan aerosol layers from 2 to 6 km MSL [DeMott et al., 2003b]. Similarly, some of the highest 2DC concentrations measured in clouds were also observed on 29 July. We believe that these observations are due to the ingestion of Saharan dust by clouds and we provide additional evidence for this fact based on the CFDC measurements of anvil particle residuals.
Figure 6 shows CFDC and cloud measurements during anvil transects on 29 July. Striking is the fact that peak 2DC concentrations exceed 1 to 2 cm−3. While this may relate in part to the strength of convection and high condensed water contents on this day, it is equally striking that these high 2DC concentration occur along with CFDC IN concentrations as high as 0.7 cm−3. The enhanced concentrations in the instruments, which coincide with the presence of Saharan dust, are suggestive of the role heterogeneous IN may play in altering anvil cloud properties [van den Heever et al., 2005]. No FSSP data were available for this flight segment due to earlier probe failure. FSSP measurements slightly preceding the time of CFDC measurements from a cloud at approximately the same altitude and with peak CVI water contents of 1 g m−3 contained peak particle concentrations between 10 and 20 cm−3, or about a factor of ten higher than the 2DC particle concentrations. Liquid water measured in these clouds was at or below the detection limit of the Rosemount probe. Data collected from the SPEC Cloud Particle Imager (CPI) are also presented here. As can be seen in the figure, CFDC measurements are generally consistent with both the 2DC and CPI measurements. Like the 2DC, the CPI emphasizes somewhat larger ice particles compared to the FSSP.
 TEM analyses of the residual nuclei collected on 29 July and processed in the CFDC indicate that metal oxides/dust and carbonaceous material each make up about a quarter of the particles collected. This is consistent with previous analysis of aerosol particles from this dust episode, which showed enhanced proportions of mineral dust [Kojima et al., 2005]. However, more than a third of the particles are categorized as containing primary elements indicative of salts. The varied compositions suggest that, despite the presence of high concentrations of heterogeneous IN, homogeneous freezing was also important for cloud ice formation on this day, again suggesting that more than one mechanism was responsible for ice in the clouds.
3.3. Composite Analyses of CFDC Ice Crystal Nuclei From Anvil Cirrus Residuals
 The temperature and humidity conditions necessary for rapid homogeneous freezing of droplets for detection in the CFDC were rarely achieved. Even for cases for which homogeneous freezing of haze droplets occurred, heterogeneous nucleation was also observed. When comparing CFDC measurements to other probes, then, we are primarily comparing measurements of heterogeneous IN, with some contribution from homogeneous freezing, to the total ice number concentration. Figure 7 compares all ice nucleation data from the CFDC to ambient ice number concentrations from the FSSP and 2DC. The CFDC measurements are correlated with both of the cloud probes. Not surprisingly, given that the CFDC samples CVI residuals, the CFDC also correlates well with CWC from the CVI (not shown; r2 = 0.68). The CFDC and 2DC agree quite well, with the 2DC measurements greater than the CFDC measurements by a factor of 2–3. Given that the CFDC predominantly measured heterogeneous nucleation, this suggests that much of the larger ice measured by the 2DC likely resulted from heterogeneous nucleation. The factor of two difference may result from the fact that we eliminate all particles larger than 1.5 μm prior to the particles entering the CFDC (particles which may serve as effective IN), transmission efficiency of residual particles reaching the CFDC, ice formation from homogeneous nucleation in the atmosphere which is not fully captured by the CFDC, and unknown secondary ice formation processes. It recently has been shown that the 2DC can overestimate particle concentrations by up to a factor of four in situations where large particles are present [Field et al., 2006], which also may account for at least part of this difference. FSSP measurements are two orders of magnitude greater than the CFDC measurements. On the basis of the assumptions that the FSSP accurately measures small ice particles from all processes and that the CFDC primarily measured heterogeneous nucleation, these data suggest that only about 1% of the ice present in the anvils results from heterogeneous nucleation. However, we again note that the FSSP may overestimate particle concentrations in the presence of ice [Field et al., 2003; Gardiner and Hallett, 1985; Heymsfield et al., 2004] and likely does not accurately reflect the ice number concentrations present in the anvil. Furthermore, the residual nuclei concentrations above 100 nm determined from electron microscope grids in the CVI prior to CFDC processing fell well below those measured by the FSSP [Twohy and Poellot, 2005]. The extent to which the FSSP overestimates ice crystal concentrations will directly impact this assessment of the fraction of particles which are nucleated by heterogeneous nucleation. Nevertheless, given that the measured FSSP concentrations are two orders of magnitude greater than those of the CFDC and less than one order of magnitude greater than the concentrations of CVI residual particles above 0.1 μm [Twohy and Poellot, 2005], we expect that homogeneous nucleation plays an important role in anvil ice formation.
Figure 8 shows CFDC concentrations as a function of ice supersaturation (Ssi) for data collected throughout the entire study. The data encompass a broad range of temperature (−28 to −50°C) and ice supersaturation (7 to 66%). Also shown in Figure 8 is the parameterization of Meyers et al. , used in many cloud-resolving models for IN concentrations. This parameterization is based on a selection of IN data collected at the surface in midlatitudes, and can only be strictly applied to warmer temperatures (−7 to −20°C) and Ssi ranging from 2 to 25%; an extrapolation of the fit to higher Ssi is shown as a dashed line in the figure. In general, the number concentrations measured during CRYSTAL-FACE (at lower temperatures) are capped by the Meyers formulation. This is despite the fact that the concentrations measured during CRYSTAL-FACE were at least partially influenced by homogeneous nucleation and that on at least one day, 29 July, there were high concentrations of Saharan dust [Sassen et al., 2003]. As such, new parameterizations are needed for quantifying IN concentrations for conditions other than those described in Meyers.
Figure 9 plots particle size of the residual nuclei which were processed in the CFDC and analyzed using TEM (96 total). Because many of the particles were aspherical, data are categorized as the square root of length × width, and are binned in 0.1-μm increments. The figure shows the entire data set, as well as the contributions from each of the categories, with sulfates and salts listed together as “soluble”. The smallest size bin, 0.1 μm, is dominated by soluble particles, while contributions from the metal oxides/dust and carbonaceous categories increase at larger sizes (mode, 0.4 μm), with most of the residual nuclei being smaller than 1.2 μm. It should be reiterated that particles larger than ∼1.5 μm are removed prior to entering the CFDC, and size-dependent particle losses have not been characterized for this experimental configuration.
 The Colorado State University CFDC was used in conjunction with a CVI to measure the ice nucleating characteristics of cloud particle residuals during CRYSTAL-FACE. The measurements were intended to directly measure the relative roles of various ice formation mechanisms in the formation of anvil cirrus. However, measurement of homogeneous ice nucleation was only accomplished on select days, rather than throughout the campaign. The primary causes for the measurement shortcomings have been identified and will be mitigated, and we hope to use this experimental method to quantitatively explore freezing mechanisms in future work.
 The data presented here demonstrate that cloud particle residuals from the CVI can be sampled using the CFDC to quantitatively measure heterogeneous IN concentrations. These measurements suggest reasonable agreement between the CFDC and the 2DC-cloud probe, within a factor of 2–3. Limited measurements from the CPI show equally good agreement. Measured IN concentrations found within clouds ranged from less than 0.001 to nearly 1 cm−3, suggesting the sampling strategy can be used for a broad range of expected atmospheric IN concentrations. For all days considered, heterogeneous IN were observed in the cloud particle residuals. Heterogeneous nucleation also may have participated in ice formation in the mixed phase regions of convective cells, initiating precipitation and reducing the water mass which can reach the anvil to form ice [Phillips et al., 2005]. Indeed, modeling studies suggest that high concentrations of IN cause precipitation to occur earlier in a convective cloud and greatly affect anvil development and structure [van den Heever et al., 2005]. Measurements are needed throughout the cloud to evaluate this issue further.
 Homogeneous freezing was also observed, but measurements were not quantitative for this campaign. As such, while we have constrained contributions from heterogeneous nucleation in this study, the relative importance of homogeneous freezing must be inferred from our measurements in the context of measurements of the cloud probes. These inferences necessarily ignore ice formation from secondary processes, which may play a role in ice formation in anvil clouds, as well as any potential overestimates of ice concentrations by the FSSP. The measured number concentrations of particles which initiated ice formation in the CFDC were two orders of magnitude lower than the cloud-ice number concentrations measured with the FSSP. From this, along with results from other studies [Heymsfield et al., 2005; Phillips et al., 2005; Twohy and Poellot, 2005], we infer that homogeneous nucleation plays a major role in anvil ice formation. Measurement strategies in future campaigns will build on the efforts reported here to directly measure the competition between heterogeneous and homogeneous freezing, and to assess the role of secondary ice formation in anvil cirrus.
 This work was supported by the NASA Code Y Radiation Sciences Program, by NASA grant NAG5-11476, through funding to MRP from NASA grant NAG5-11509, and through funding to AJH from NASA through grant NNH04AA821, Hal Maring, program manager.