Investigations of Mesoscopic Complexity of Small Ice Crystals in Midlatitude Cirrus

Abstract

During the Mid-Latitude CIRRUS campaign small (<50 μm) ice crystals were characterized for their mesoscopic complexity using the Small Ice Detector Mk. 3. It was found that the majority (76%) of all measured small ice crystals were complex. Although some differences were found between the distributions of the complexity parameters between different cloud systems, no real correlation between the mesoscopic complexity and cloud formation mechanisms (liquid origin or in situ) was found. Some decrease in mesoscopic complexity was detected during those missions affected by Saharan dust outbreaks. The link between the crystal growth conditions and the observed mesoscopic complexity was investigated in a case study of an in situ cirrus with the help of trajectory analysis. A relatively weak correlation of r = 0.33 was found between the mesoscopic complexity and the maximum supersaturation with respect to ice along the ice particle trajectory.

Key Points

• Majority of midlatitude small ice crystals contain mesoscopic complexity
• This complexity is found for high-altitude clouds irrespective of their formation mechanism
• The mesoscopic complexity is reduced in some clouds with a high number of heterogeneous ice nuclei

Plain Language Summary

Micrometer-scale features in atmospheric ice particles can significantly change the way these particles interact with sunlight. However, due to the small size of such features, they are difficult to be detected using traditional methods. Using light diffraction, it is possible to detect features in the order of wavelength. Here an instrument measuring diffraction patterns of individual atmospheric ice crystals was used to measure mesoscopic features in midlatitude cirrus clouds and to categorize the measured ice particles as complex or pristine. It was found that the measured ice clouds—independent of their formation mechanism—were mainly composed of complex ice crystals. The presented results support previous remote sensing observations.

1 Introduction

Ice crystal mesoscopic complexity can significantly affect the ice particle single scattering properties (e.g., Baum et al., 2010; Järvinen et al., 2016; Schnaiter et al., 2016; Ulanowski et al., 2006; Yang et al., 2013) and, consequently, the cloud radiative effect. However, our current understanding of ice crystal mesoscopic complexity is mostly limited to laboratory and remote sensing observations. Laboratory studies with environmental scanning electron microscopy have shown that mesoscopic features, such as ice crystal surface roughness, can be found in a wide range of environmental conditions (Butterfield et al., 2017; Magee et al., 2014; Neshyba et al., 2013), and the interpretation of global polarimetric satellite observations have given evidence that the majority (70%) of atmospheric ice crystals have some degree of mesoscopic complexity (Cole et al., 2014). Detailed information both on the spatial scale of ice crystal complexity and on its magnitude are important, since incorrect treatment of the degree of ice crystal complexity in general circulation models can cause large errors in the cloud radiative effect (Yi et al., 2013).

Here the term ice crystal mesoscopic complexity is used to describe all possible deviations from the pristine hexagonal crystal structure in a single ice particle (e.g., surface roughness, hollowness, air inclusions, and polycrystalinity) that cause speckles in the coherent light scattering. When using visible light, these features are in the wavelength scale from 100 nm to several microns. Measuring submicron complexity of atmospheric small ( < 50 μm) ice crystals using coherent light scattering became possible after the development of the Small Ice Detector Mk. 3 (SID-3; Kaye et al., 2008; Vochezer et al., 2016). Ulanowski et al. (2014) were the first to operate SID-3 in midlatitude cirrus and mixed-phase clouds. The authors compared in situ complexity with complexity measured on a set of test particles (ice analogues and mineral dust particles) and found that typical in situ complexity corresponds to that found on the more complex (more surface roughness) subset of the test particles. The authors used the combined roughness definition by (Lu et al., 2006) as a measure of the mesoscopic complexity. Later, Schnaiter et al. (2016) calibrated the SID-3 method in cloud chamber studies for ice particles grown at different environmental conditions (with respect to temperature and ice saturation ratio) and found that there is a threshold complexity parameter, after which the ice particle angular light scattering function remained unchanged, independent of a further increase in the complexity parameter. The authors recommended to use the coefficient k_e from the speckle analysis (Lu et al., 2006) as a complexity parameter. Schmitt et al. (2016) applied this analysis to SID-3 data collected during a campaign over the continental United States and found that the majority (77–86%) of the measured small ice particles has a complexity parameter larger than the defined threshold.

Despite the first in situ evidence of ice crystal mesoscopic complexity, a large uncertainty both of the origin and in the variability of mesoscopic complexity of atmospheric ice clouds still exists. Schnaiter et al. (2016) were able to show that the degree of mesoscopic complexity is directly dependent on the crystal growth speed that is driven by the available water vapor during the growth process. However, Ulanowski et al. (2014) found no correlation between crystal complexity and other thermodynamic or microphysical properties found in situ. In this paper, a new approach is used in an attempt to explain the observed ice crystal mesoscopic complexity in midlatitude high-altitude clouds. Instead of relating the in situ complexity parameter to simultaneously measured thermodynamic properties, a trajectory analysis is used to track the history of air parcels measured along the flight trajectory (Wernli et al., 2016). Such a Lagrangian analysis allows to determine the formation mechanism of the ice cloud particles (liquid origin or in situ) and the maximum ice saturation ratio along the trajectory. Liquid-origin ice clouds are formed via further lifting and freezing of previous liquid cloud particles, whereas in situ ice clouds form directly from the gas phase (Krämer et al., 2016; Luebke et al., 2016; Wernli et al., 2016). The main objectives of this study are (i) to quantify the mesoscopic complexity of ice crystals observed in different types of ice clouds during the Mid-Latitude CIRRUS (ML-CIRRUS) aircraft campaign, (ii) to investigate whether ice crystal complexity differs between different cloud types, and (iii) to analyze whether the measured in situ mesoscopic complexity correlates with maximum ice supersaturation identified along backward trajectories.

2 In Situ Measurements

In situ measurements in midlatitude high-altitude cirrus clouds were performed by the German High Altitude and Long Range Research Aircraft (HALO) during the ML-CIRRUS campaign (Voigt et al., 2017) between 26 March and 15 April 2014. Based from Oberpfaffenhofen (Germany), research flights targeted natural and aviation-induced cirrus over Europe. A special focus was given to cirrus clouds in the outflows of warm conveyer belts (WCBs). WCBs are moist, coherent airstreams in extratropical cyclones, which ascend within 2 days from the (oceanic) boundary layer to the upper troposphere while moving poleward by typically more than 2,000 km. Their ascent is accompanied by intense condensation, cloud formation, and precipitation (e.g., Madonna et al., 2014). From altogether 13 research missions, 6 missions targeting natural cirrus and 1 mission targeting aviation-induced cirrus were selected for this analysis. A summary of the selected missions can be found in Table 1.

Table 1. Summary of the Cloud Types and Air Mass Origins of the Cloud Systems Shown in Figure 2

ML-CIRRUS	Date	Time			Air Mass
Mission	(DD/MM)	(UTC)	Median ke	Cloud type	origina	Remarks
5	27/03	12:25–13:10	5.15	Liquid origin	Continental
5	27/03	14:12–15:15	4.92	WCB, liquid origin, in situ	Continental
6	29/03	15:08–15:56	4.74	Liquid origin, in situ	Continental	Saharan dust outbreak (SDO)
6	29/03	16:54–18:32	4.99	Liquid origin, in situ	Continental	SDO
7	01/04	9:53–10:26	4.71	In situ		halo observation, SDO
8	03/04	15:45–16:19	5.10	Liquid origin, in situ	Continental	SDO
9	04/04	9:06–10:59	4.65	Liquid origin, in situ	Continental	SDO
9	04/04	14:46–15:49	5.04	In situ
11	07/04	8:07–11:04	4.97	In situ		Aviation influenced
13	11/04	9:19–11:37 and	4.88	WCB, liquid origin	Maritime
		14:23–15:39b

• Note. WCB = warm conveyer belt.
• ^a Air mass origin is only defined for liquid-origin clouds that have experienced an ascend. The origin refers to the geographical location, where the ascent has started.
• ^b Same system was sampled twice.

During ML-CIRRUS ice crystal submicron complexity of small ice crystals was measured with the SID-3. The SID-3 probe has an open geometry to minimize artifacts due to ice particle shattering on the probe housing (Cotton et al., 2010; Korolev et al., 2011; McFarquhar et al., 2007), and, for the HALO aircraft, the SID-3 instrument head was further modified so that the blunt instrument front (Cotton et al., 2010) was replaced by a conical one.

In SID-3 two trigger detectors symmetrically positioned at 50° relative to the forward light scattering direction detect individual cloud particles passing a 532-nm 30-mW laser beam. From these particles, light scattered into the annulus-shaped main detector aperture, covering an angular range from 7° to 23° around the laser propagation direction, is recorded by an intensified CCD camera (ICCD, Photek Ltd, UK). In this way, high-resolution two-dimensional scattering patterns of individual ice crystals are acquired at a maximum repetition rate of 30 Hz. The brightness of the ICCD camera can be adjusted to avoid saturation of the image pixels. To be consistent with the laboratory studies by Schnaiter et al. (2016), camera gain settings between 175 and 195 were chosen. The upper limit of the gain settings is higher than those used in the laboratory experiments (180) to allow a good ICCD illumination also during aircraft operation where particle residence times in laser are much shorter than in the laboratory.

For each imaged particle also the particle time-of-flight (TOF) is recorded, which is the measurement of particle residence time in the sensitive volume of SID-3. With the laser profile width of 160 μm a typical residence time of 0.8 μs is calculated assuming an airspeed of 200 m/s. The 48-MHz electronics of SID-3 measures this residence time as a TOF of 38 clock cycles. Sometimes TOFs exceeding 350 cycles are recorded that can be related to a shattering events (Schmitt et al., 2016), where the fragments of a shattered crystal penetrate the sampling volume at intervals too short for the electronics to be separated resulting in a long-lasting single trigger pulse. Typically, longer (>350) TOFs are seen in a separate mode from otherwise lognormally distributed shorter (<350) TOFs, which supports the conclusion that longer TOFs belong to the shattering mode. All the imaged particles with TOFs exceeding 350 (a total of 0.8% of all scattering patterns) were classified as shattered particles and were removed from the analysis.

The 2-D scattering patterns were then analyzed for their mesoscopic complexity using the procedure described in Schnaiter et al. (2016). Only scattering patterns within a narrow mean brightness (gray level) range between 10 and 50 were included in the analysis to avoid image brightness biases in the analysis. This excluded 61–76% of the images, majority having mean image brightnesses larger than 50. The result of this analysis is an optical complexity parameter, k_e, that can have values approximately between 4 and 6 depending on the degree of the actual surface roughness. Previous studies have shown that there is a correlation between the optical complexity parameter k_e and the physical surface distortion (σ) in the range from σ = 0.1 to 0.5 (Schnaiter et al., 2016). A threshold complexity parameter of 4.6 can be applied to distinguish between complex and pristine ice particles.

3 A Trajectory-Based Analysis of Ice Particle History

Backward trajectories have been calculated after the ML-CIRRUS campaign from the aircraft positions every 10 s along all flights (Figure 1 shows a few examples). Trajectories were computed 10 days backward, using wind fields from operational analyses from the European Centre for Medium-range Weather Forecasts using the Lagrangian analysis tool LAGRANTO (Sprenger & Wernli, 2015). The algorithm introduced by Wernli et al. (2016) was then applied to all backward trajectories starting from an ice cloud in order to distinguish between the in situ and liquid-origin formation mechanisms. A subcategory of liquid-origin cirrus is classified as WCB, if the backward trajectory shows an ascent of more than 600 hPa within 2 days (which is typical for a WCB). Along all trajectories, hourly values of temperature and humidity have been interpolated, which allow investigating the supersaturation history of cirrus air parcels. The maximum supersaturation will be used below to analyze its potential effect on the mesoscopic complexity of ice crystals.

Figure 1

Examples of 10-day backward trajectories from selected 12-min flight sequences from missions 7 (M7), 9 (M9) and 13 (M13), respectively. For every 12-min sequence, a set of 72 trajectories has been calculated (every 10 s along the flight path). The time indicated in each panel refers to the central time of the 12-min sequence.

4 Case Studies of Ice Crystal Mesoscopic Complexity During ML-CIRRUS

Altogether 10 individual cloud systems from the seven selected ML-CIRRUS missions were selected for analysis of the mesoscopic complexity. For each system, the Lagrangian formation type (in situ, liquid-origin, WCB) and geographical origin (continental or maritime) was analyzed using backward trajectory analysis. The summary of this classification can be found in Table 1. In 6 out of 10 cases, the sampled cloud systems were a mixture of different cloud types, typically consisting of in situ formed tops with liquid-origin cloud beneath. The air masses were mostly found the be of continental origin arriving from North Africa (Figure 1).

Statistical analysis of the k_e values for each cloud system is shown in Figure 2. The majority (76%) of all the measured ice particles were found to be complex (having k_e values above the threshold of 4.6). However, some differences in the k_e value statistics between the studied cases can be detected. The variation in the median complexity parameter in individual cases cannot be explained only with cloud type (liquid origin or in situ) or air mass origin (continental or maritime). Overall, it can be concluded that a somewhat lower median complexity parameter (4.87) is measured during missions affected by Saharan dust outbreaks compared to the median complexity parameter (4.96) of the missions unaffected by dust. This so-called aerosol effect is based on the fact that higher ice nuclei particle concentrations lead to a lower in-cloud ice supersaturation in cirrus clouds and vice versa (Spichtinger & Gierens, 2009). Schnaiter et al. (2016) showed that ice crystals grown in lower ice supersaturation become more pristine than those grown at higher supersaturations.

Figure 2

Statistical analysis of complexity parameter k_e for selected cloud systems (Table 1). The median k_e are shown as red horizontal line, and the blue box illustrates the interquartile range. The black bars show the 5% and 95 % confidence range. A threshold complexity parameter of 4.6 (red horizontal line) can be used to separate between pristine and complex crystals. The missions affected by Saharan dust outbreaks are highlighted with a gray box.

Additional evidence of the reduction of ice particle complexity during Saharan dust affected flights was seen during mission 7, when a ground-based 22° halo observation was made in the cirrus field that the HALO aircraft was simultaneously sampling. A 22° halo is an optical phenomenon that is related to the presence of pristine ice crystals, although recent studies have shown that also moderately roughened columns or severely roughened columns with transprismatic symmetry on prismatic facets can cause a visible halo phenomena (Forster, 2017; Neshyba et al., 2013). The SID-3 measurements from that time period showed that only 61% of the measured ice crystals were complex. A manual inspection of the individual images showed that between 17 and 27 % of the individual 2-D scattering patterns showed a halo feature although only 4% to 8 % of the scattering patterns showed a clear pristine hexagonal structure (Forster, 2017). Therefore, the in situ observations support the hypotheses that either only a small fraction (around 10 %) of pristine crystals are needed for a 22° halo phenomenon to be visible (Forster et al., 2017) or that halo-producing ice particles can be moderately complex.

5 Dependence of Submicron Complexity on Air Parcel Origin

Recent studies have indicated that microphysical properties differ for in situ and liquid-origin clouds (Krämer et al., 2016; Luebke et al., 2016). Liquid-origin clouds are associated with higher ice water contents, ice crystal concentrations, and ice crystal size compared to in situ cirrus (Luebke et al., 2016). Here the effect of air parcel origin to the mesoscopic complexity of small ice crystals is investigated. Each of the ice particles measured during the missions listed in Table 1 was classified as liquid origin, liquid-origin WCB, or as in situ based on the backward trajectory analysis. A statistical analysis of the k_e values for all the measured small ice particles based on their origin are shown in Figure 3. Compared to Figure 2, less difference is seen in the median k_e value between the different cloud types. The in situ ice particles have only slightly higher median k_e value of 4.98 compared to the k_e values of the liquid origin (4.88) and liquid-origin WCB ice particles (4.90).

Figure 3

Statistical analysis of the complexity parameter, k_e, for ice particles classified as liquid-origin warm conveyer belt (WCB; red), liquid-origin (blue), or as in situ origin (yellow). The boxes illustrate the interquartile range and the horizontal line inside the box the median k_e value. The black bars show the 5% and 95 % confidence range. A threshold complexity parameter of 4.6 (red horizontal line) can be used to separate between pristine and complex crystals.

It is somewhat surprising that no clear difference in the mesoscopic complexity of liquid-origin and in situ formed ice particles can be observed. It could have been assumed that ice formation and growth in mixed-phase environment would support the formation of crystal deformations. First, the ice crystals grow at near water-saturated conditions, where the growth can be rapid, promoting the formation of crystal complexity (Schnaiter et al., 2016). Second, if remaining liquid droplets reach the homogeneous freezing level in the updrafts, these droplets freeze homogeneously. Laboratory studies have shown that homogeneously frozen droplets exhibit one of the largest measured k_e values (k_e>6) during the initial growth (Järvinen et al., 2016). Vapor-grown in situ ice crystals, on the other hand, exhibit more moderate complexity values (Schnaiter et al., 2016). The observed results indicate that processes taking place after the initial growth (e.g., sublimation and regrowth cycles) can further affect the properties of small ice crystals so that the complexity signature of the initial formation and growth is lost.

6 Comparison to Cloud Chamber Experiments

Cloud chamber experiments have shown that for ice particles formed under in situ cirrus conditions, their mesoscopic complexity depends on the crystal growth conditions (Schnaiter et al., 2016). A linear dependency can be observed between the complexity parameter and the ice saturation ratio that drives the crystal growth for ice crystals that were heterogeneously nucleated at 223 K (Figure 4). The mesoscopic complexity of homogeneously nucleated ice crystals was found to be enhanced compared to those formed in heterogeneous nucleation. However, it is not known if a similar relationship between the thermodynamic conditions and ice crystal complexity can be established in natural ice clouds.

Figure 4

Comparison of the complexity parameter k_e of laboratory grown ice crystals (diamond) and ice crystals measured during sampling of an in situ formed cirrus (red dots). The laboratory-generated ice crystals were nucleated either heterogeneously on soot aerosol (light gray symbols) or homogeneously (dark gray symbols) at a temperature of 223 K and after a short sublimation period, grown at a stable ice supersaturation shown at the x axis. The laboratory measurement are previously published in Figure 4 in Schnaiter et al. (2016). For the in situ measured ice crystals, their complexity parameter is related to the maximum ice supersaturation found along the air parcel trajectory. The dashed line represents a linear fit to the field measurements. The correlation coefficient between the maximum ice saturation ratio on the field observations of the complexity parameter is 0.33. ML-CIRRUS = Mid-Latitude CIRRUS.

Ulanowski et al. (2014) found no correlation between the complexity parameter and simultaneously measured in situ relative humidity with respect to ice, indicating that the ice particle history has a more significant role on their microphysical properties than the prevailing thermodynamic conditions at the time of crystal measurement. Here we investigate if a correlation can be found between past thermodynamical conditions and the observed crystal complexity in a simple case study of an in situ cirrus. For the comparison, an in situ cirrus sampled on the coast of Portugal (mission 9) was selected. According to the trajectory analysis, all the ice particles measured in the in situ cirrus formed at a similar temperature of 229 ± 5 K (mean and standard deviation) and no large variations in the maximum vertical velocity was observed (standard deviation 1 ms⁻¹).

The ice particle complexity was linked to the maximum ice saturation ratio observed along the trajectory with the assumption that the ice crystal complexity is mainly defined by the period of the maximum growth. The temperature at the time of maximum ice supersaturation was found to be 217 ± 7 K (mean and standard deviation), which is comparable to the temperature range of 221–215 K where laboratory-produced ice crystals were grown (Schnaiter et al., 2016).

Figure 4 shows that the field measurements of the complexity parameter are located between the laboratory simulations for heterogeneous and homogeneous freezing. The probed cirrus cloud consists of ice particles that were likely formed both heterogeneously and homogeneously, although based on the observed complexity parameters, homogeneous formation pathways might have been more dominant. The sampled air masses had maritime origin, which would support a low number of ice nuclei particles and, thus, homogeneous nucleation pathways. A weak positive correlation of r = 0.33 is found between the maximum ice saturation ratio and the measured complexity parameter, suggesting that the ice crystal complexity cannot be explained only with the maximum ice saturation ratio. As already discussed above, the thermodynamic history of a single ice particle can be diverse, which leads to a complicated relationship between the thermodynamical values along the ice particle trajectory and the particle's morphology.

7 Conclusions

In this study we have shown that the majority of the small ice crystals measured in midlatitude high-altitude clouds contain a significant degree of mesoscopic complexity. Some differences were found in the measured complexity parameters between different sampled clouds, but these differences could not be explained either with cloud air mass origin (marine or continental) or with cloud type (in situ or liquid-origin cirrus). Only missions affected by Saharan dust outbreaks were observed to have lower complexity values compared to missions unaffected. An in situ cirrus was chosen as a case study to investigate if a relationship between the ice parcel history and the measured complexity parameter can be found. In contrast to previous laboratory studies, only a weak correlation was found between the maximum ice saturation ratio along the trajectory and the ice crystal complexity. There are two possible explanations for this weak correlation: (i) the maximum supersaturation along the trajectory is not a good proxy for ice crystal complexity but sublimation and regrowth cycles make this dependency more complicated. Furthermore, (ii) the European Centre for Medium-range Weather Forecasts grid box might be too coarse so that the grid average ice supersaturation ratio might not be representative for the supersaturation ratio experienced by the individual ice crystals, which would explain the large variation in the complexity values seen in Figure 4.

Although this study was not able to point out the origin of ice crystal complexity in natural ice clouds, it provides in situ proof of the importance of mesoscopic complexity in natural ice clouds. Previous studies have retrieved the degree of ice crystal mesoscopic complexity using polarimetric satellite observations (Cole et al., 2014; Hioki et al., 2016; van Diedenhoven et al., 2014). Cole et al. (2014) showed that almost 70% of the global observations can be explained with surface roughness, which is similar to the observation presented here that 76% of the measured ice crystals contain mesoscopic complexity. Therefore, our results support the conclusions of previous polarimetric satellite observations.

Acknowledgments

The SID-3 analysis results and the ECMWF data products are available from the HALO database (https://halo-db.pa.op.dlr.de/mission/2). We gratefully acknowledge all the participants of ML-CRRUS campaign for their efforts, in particular the technical grew of HALO. This work has received funding from the Helmholtz Research Program Atmosphere and Climate and from the German Research Foundation (DFG grants SCHN 1140/1-1, SCHN 1140/1-2, and SCHN 1140/3-1) within the DFG priority program 1294 (HALO).