Quantification of the Effects of Shattering on Airborne Ice Particle Measurements (original) (raw)
1. Introduction
Small ice particles may play a significant role in radiation transfer and precipitation formation, and their associated parameterizations have been included in many numerical climate and weather prediction models. Debates around the problem as to whether small ice particles are omnipresent in ice clouds extend well over three decades and began when optical particle size spectrometers (Knollenberg 1976) were commonly adopted for airborne cloud particle sampling in the mid-1970s. Early airborne measurements suggested that the number concentration of ice particles in glaciated clouds is dominated by small particles with sizes less than 100 μm (e.g., Heymsfield and Platt 1984). Such observations indicated that the ice crystal number concentration is essentially always dominated by small ice particles, as illustrated conceptually in Fig. 1 (curve 1). Small ice particles in high concentrations have been measured in regions with undersaturated and supersaturated environments with respect to ice. Numerical simulations suggest that small ice particles should quickly grow to larger sizes in a supersaturated environment, or completely evaporate in a subsaturated environment, suppressing the concentration of ice crystals D < 100 μm (Fig. 1, curve 2).
Recently, there has been mounting evidence that a large fraction of these concentrations of small ice particles may be artifacts, the result of larger ice particle impacts on probe tips and inlets followed by breakup into small fragments. After bouncing off the tip or inlet, the shattered fragments may travel into the sample area and cause multiple artificial counts of small ice.
The earliest recognition of the potential significance of particle shattering was brought up by Cooper (1977), who suggested artifact filtering based on the characteristically short interarrival times of such particles produced by breakups.
Gardiner and Hallett (1985) compared the forward scattering spectrometer probe (FSSP) concentration and calculated the water content with that measured by the cloud replicator and a Johnson–Williams hot-wire probe, respectively. They found that in ice clouds the FSSP size distribution appeared to have an unusually flat tail, and the number concentration exceeded that estimated from the replicator by a factor of 2 to 3. The calculated FSSP water content was systematically higher than that measured by the Johnson–Williams probe. No clear explanation was found for the difference in the concentrations measured by FSSP and the replicator. Shattering of ice was considered as one of several possible explanations, and the authors recommended that the FSSP not to be used for measurements in ice clouds.
Gayet et al. (1996) found that FSSP and optical array probe (OAP-2DC) size distributions agreed well in the overlap area in cirrus clouds with small compact-shaped ice particles. However, in the presence of large irregular ice crystals, the FSSP concentration could exceed that of the OAP-2DC by more than an order of magnitude. The authors concluded that the FSSP measurements were unreliable and that FSSP data obtained in ice clouds with large ice particles should not be used for microphysics characterization. However, no explanation of the cause of the elevated FSSP concentration was provided.
Field et al. (2003) used a fast FSSP with an interarrival time option to measure particle spacing in ice clouds. The interarrival time distribution in ice clouds was found to have a bimodal shape with modes at 10−2 and 10−4 s corresponding to approximately 1-m and 1-cm spatial separations. The particles from the long and short interarrival time modes corresponded to estimated concentrations of 0.1–1 and ~100 cm−3, respectively. No conclusions were drawn as to whether the latter localized clusters of high particle concentration were natural or artifacts. Assuming they were artifacts, their interarrival time algorithm suggested average and maximum concentration overestimates of a factor of 2 and 5, respectively.
Korolev and Isaac (2005) presented OAP-2DC, OAP-2DP, and high volume precipitation spectrometer (HVPS) images of fragmented particles as direct evidence of shattering in these probes. The shattered particles were identified from their multiple fragments within single images and by analysis of the interarrival time. The image analysis identified shattering resulting from two causes: (i) mechanical impact of particles with probe surfaces upstream of the sample area, which usually generates a large number of small fragments, and (ii) aerodynamic stresses caused by the flow around the probe housing, which may be sufficient to induce particle breakup. The aerodynamic shattering typically results in fewer numbers of larger fragments than that from mechanical impact. It was found that the number of shattered particles increased with the increasing original particle size. The analysis of the fragmented images suggested that the measured number concentration can be overestimated by 10%–20%.
Field et al. (2006) applied an interarrival time algorithm to filter out shattering artifacts, choosing a threshold interarrival time in the range 10−4 to 10−5 s, depending on the instrument and the aircraft type used for the data collection, to reject the short interarrival mode. It was found that the OAP-2DC and cloud imaging probe (CIP) concentrations were reduced by up to a factor of 4 when the mass-weighted mean size exceeded 3 mm. The ice water content (IWC) estimate was reduced by up to 20%–30%, most notable in cases of narrow size distributions. It was also concluded that the effect of shattering on the measured size distribution reaches up to sizes of several hundred microns.
Heymsfield (2007) compared the extinction coefficient and IWC calculated from FSSP, cloud and aerosol spectrometer (CAS), and OAP-2DC size distributions to those measured by the counterflow virtual impactor (CVI) (Twohy et al. 1997) and the cloud-integrated nephelometer (CIN) (Gerber et al. 2000). The FSSP and CIN extinctions were found to be at least as large as the extinction in that derived for the large particles (>100 μm), and the FSSP-derived IWC was approximately 15% of that derived for large particles measured by OAP-2DC. The conclusion drawn from this analysis was that all of the above probes are measuring a combination of natural small crystals and shattered large ice crystals.
McFarquhar et al. (2007) compared cloud droplet probe (CDP) and CAS size distributions measured in anvil cirrus. The CAS has a tubular inlet design and its measurements are anticipated to be prone to shattering. Contrary to that, the CDP has an open-concept inlet, and therefore its measurements are expected to be less susceptible to the shattering effect as compared to that for the CAS. The concentration measured by CAS was systematically higher than that measured by CDP in the presence of large ice particles (D > 100 μm) measured by 2D probes, and in extreme cases it reached 2 orders of magnitude higher. The enhanced CAS concentration was attributed to shattering on the CAS tubular inlet.
Jensen et al. (2009) compared measurements of a CAS and a two-dimensional stereo probe (2D-S) in cirrus clouds. They found that the CAS small particle concentrations exceeded those of the 2D-S by 1–2 orders of magnitude in the 10–50-μm size range where the instruments overlap. The ratio of CAS and 2D-S concentrations was found to strongly correlate with large crystal mass, suggesting that the discrepancy is caused by the shattering of large crystals on the CAS inlet.
Vidaurre and Hallett (2009) used replicator and cloudscope data to analyze the process of particle fragmentation on impact with solid surfaces. It was shown that the number of ice fragments on replicator images increased with the original particle size and that 100–200-μm plates impacting the Formvar-coated surface may break into hundreds of fragments. They also suggested using the Weber number, given by the ratio of the ice particle kinetic energy to its surface energy, as a criterion for the particle break up. However, no experimental data were presented to support this approach.
Despite the substantial evidence of the significant effect of shattering on ice particle size distribution measurements, the shattering hypothesis was not commonly accepted in the cloud physics community for many years. Many researches argued that shattered particle fragments, after bouncing off a probe upstream surface shed along the surface of the arms or inlets and that they could not travel several centimeters across the airflow at an aircraft speed of ~100 m s−1 to reach the probe’s sample volume.
Direct experimental support for the shattering hypothesis has been provided by the National Aeronautics and Space Administration (NASA) Glenn Research Center (GRC), using high-speed video records of ice particle bouncing conducted at the Cox and Company wind tunnel facility. For the first time, it was visually documented that at aircraft speeds rebounding ice particles can travel several centimeters across the airflow, which is sufficient to reach a typical probe’s sample volume (Korolev et al. 2011, 2013a). These results initiated a series of efforts undertaken by Environment Canada (EC) in collaboration with NASA GRC to optimize the shapes of probe arm tips and to quantify their role in the contamination of ice particle measurements by shattering. At the first stage, the probes’ arm tips were redesigned targeting the following three goals: (i) to mitigate the effect of ice particle shattering on measurements; (ii) to minimize the disturbance of the airflow upstream of the sample volume in order to minimize aerodynamic stresses experienced by particles that may cause changes to their orientation, fragmentation, and spatial distribution when passing through the sample volume; and (iii) to mitigate shedding of liquid water, which may get into the optical windows of the probe. The modification of the probes’ arms was supported by computational fluid dynamics (CFD) analysis, bouncing simulation, and wind tunnel testing with high-speed video recording of the interaction of ice particles with the probes’ housing. The results of this work and the performance of the tips with the optimum shape (K-tips) were described in detail in Korolev et al. (2013a).
The second stage was focused on the in situ characterization of the effect of shattering on ice particle measurements during a dedicated Airborne Icing Instrumentation Evaluation (AIIE) flight campaign. Preliminary results of this work were described in Korolev et al. (2011).
This work presents the detailed analysis of these AIIE airborne results. The paper is organized as follows. Section 2 describes the AIIE flight campaign and strategy for characterization of the effects of shattering. Section 3 presents the results of the effect of shattering on the FSSP measurements. Section 4 discusses the effect of shattering on the OAP-2DC and CIP, along with the effectiveness of the antishattering tips. Section 5 describes the role of different microphysical, environmental, and sampling parameters on the measured shattering effect. Section 6 discusses the existence of small particles and status of the historical data. Section 7 presents the conclusions of this work.
2. AIIE flight campaign
a. Instrumentation
Environment Canada, in collaboration with the National Research Council (NRC) of Canada, conducted the AIIE flight campaign to study the accuracy of the microphysical characterization of ice clouds and to improve understanding of the problem of small ice particles in clouds. The primary objectives of the AIIE campaign were 1) to quantify the effect of shattering on ice measurements; 2) to evaluate the performance of the antishattering tips developed for this study; 3) to evaluate the possibility of reaching closure between direct measurements of IWC, the extinction coefficient, and those deduced from size distribution measurements of particle probes; and 4) to evaluate the performance of different modifications of hot-wire total condensed water sensors. This article will focus on the first three above-mentioned objectives.
A suite of airborne instruments for characterization of cloud microphysics and for measurement of state parameters was installed on the NRC Convair-580 aircraft (Table 1). Particle sizes and concentrations were measured by two Particle Measurement Systems (PMS) FSSPs, two PMS optical array probes (OAP-2DC and OAP-2DP), two Droplet Measuring Technologies (DMT) CIPs, and a SPEC Inc. 2D-S probe. Both FSSPs were set to operate in the nominal size range of 2–47 μm through the entire project. The two CIPs had a 15-μm pixel resolution, whereas the two OAP-2DCs had a 25-μm pixel resolution. The OAP-2DP and 2D-S were operated at 200- and 10-μm pixel resolution, respectively. All particle probes were routinely checked and calibrated during the flight operation.
Table 1.
List of instruments installed on the NRC Convair-580 during the AIIE project.
In mixed-phase clouds, small shattered ice fragments may be confused with cloud droplets and therefore mislead the quantification of the effect of shattering on ice measurements. For this reason, the focus of most of these measurements was ice clouds, and therefore the identification of the phase composition played an important role here. Phase composition was deduced from a composite analysis of particle probes: hot-wire probes and the Rosemount icing detector (RID) as described in Korolev et al. (2003). The Rosemount icing detector was used to detect the presence of liquid water in clouds with amounts exceeding 0.005 g m−3 (e.g., Mazin et al. 2001).
Measurements of particle probes are intensively used in the cloud physics community for calculations of the extinction coefficient (β) and IWC. To determine the role of shattering, the values of β and IWC calculated from particle probes were compared with those measured by the cloud extinction probe (CEP) and Nevzorov total water content (TWC) hot-wire sensor, respectively. The CEP utilizes a direct method of reduction of light transmission to measure the extinction coefficient from first principles (Korolev et al. 2013, manuscript submitted to J. Atmos. Oceanic Technol.). The Nevzorov deep-cone TWC sensor has been modified in order to improve its performance in capturing and evaporating ice particles (Korolev et al. 2013b). Both the CEP probe and Nevzorov deep-cone sensor are not affected by shattering. However, in the absence of mature calibration standards for ice sprays, the absolute accuracy of the Nevzorov TWC measurements in ice clouds remains a subject of further investigation.
The particle 2D probe data are asynchronous and usually have low particle counting statistics at 1-s averaging time. This may create problems during comparisons of the 2D data with short distance–scale 1D measurements (FSSP, Nevzorov, CEP, etc.). To reduce 2D sampling statistics errors, the 2D data were averaged over 5-s intervals. The 1D data were also averaged over 5-s intervals and synchronized with the 2D data. For some special FSSP comparison cases, both 1D and 2D data were averaged over 30 s.
Thirteen research flights were conducted in the vicinity of Ottawa (Canada) during the period of March–April 2009. The duration of each flight ranged from 2 to 3 h.
b. Strategy for quantification of the shattering effect
To characterize the effect of shattering, the following approach has been used:
- The original manufacturers’ tips, inlets, and probes’ arms were replaced with modified ones designed to mitigate shattering (Korolev et al. 2013a).
- Each probe type was simultaneously flown in its modified and standard configuration, side by side (Fig. 2). The quantification of the effect of shattering was based on the comparisons of the measurements of the modified and standard probes. The difference between their measurements was ascribed to shattered artifacts.
Fig. 2.
Cloud particle probes installed on the NRC Convair-580 during the AIIE flight campaign. Pairs of the (a) OAP-2DC and (b) FSSP and CIP, with modified and standard tips or inlets were mounted side by side on the same pylons. This enabled direct comparisons of the measurements of standard and modified probes and quantification of the effect of the ice shattering on the measurements.
Citation: Journal of Atmospheric and Oceanic Technology 30, 11; 10.1175/JTECH-D-13-00115.1
The characterization of shattering was mainly focused on three particle probes: FSSP, OAP-2DC, and CIP. Each particle probe had several probe tip modification options based on CFD analysis and wind tunnel testing, as described in Korolev et al. (2013a). In total, eight sets of OAP-2DC, three sets of CIP, and three sets of FSSP tips and/or arms were tested during the AIIE project. The inlet tube of the modified FSSP was removed, and the original hemispherical tips were replaced with new designs (Fig. 2b). The modification of the CIP tips, in addition to the mitigation of shattering, also targeted reduction of the out-of-focused “donut-looking” images, by reducing the distance between the field apertures from the original 10 cm (Fig. 2b) down to 4, 5, or 6 cm in the modified arms. During the first half of the project, the 2D-S was flown with the standard manufacturer’s tips, and for the second half, the modified tips were installed. No modifications were performed to the 2DP probe. After each flight a different set of modified tips was installed on the OAP-2DC, CIP, and/or FSSP, whereas the probes with standard tip configurations remained unchanged on every flight.
During the first few flights, and at the end of the project, all probes were flown in standard configuration to establish that the electronics and optics of each pair of probes produced equivalent measurements before and after modifications. Thus, after establishing the equivalence of measurements for each pair of probes with the standard configuration, differences observed in between modified and standard probes could be attributed to the modifications and not to poor probe precision. Figure 3 shows the measurements for flight with the standard configuration pairs of OAP-2DCs, CIPs, and FSSPs, revealing nearly equal measurements from each pair. The systematic concentration biases of 10%–20% for pairs of 2DCs and CIPs are most likely related to differences in the response time of the probes’ electronics or differences in timing of the image sampling determining the sample volume. It should be noted that in spite of the fact that the pairs of the probes were designed to be identical, their performance can be nevertheless significantly different (Gayet et al. 1993).
Fig. 3.
Scatter diagrams of data obtained from pairs of identical particle probes with the standard housing configurations: (a) 2DC counts (flight 21 Mar 2009), (b) CIP counts (flight 30 Mar 2009), and (c) FSSP counts (flight 30 Mar 2009).
Citation: Journal of Atmospheric and Oceanic Technology 30, 11; 10.1175/JTECH-D-13-00115.1
The goal of each research flight was to evaluate the performance of each set of modified tips in a variety of cloud conditions with different ice particle habits, sizes, concentrations, and ice water contents. For this reason, the flights were deliberately conducted in deep precipitating glaciated cloud systems associated with frontal clouds. In such clouds the characteristic size of ice particles usually gradually changes from a minimum near the cloud top to a maximum in the precipitation below the cloud base. Cloud sampling was conducted during spiral or porpoise ascent to maximum altitude (7.5 km) and then descent to the minimum permitted altitude (usually 0.3–1.5 km). The temperature of the ice clouds ranged from −35°C to 0°C.
To estimate the effect of airspeed and angle of attack on the performance of the standard and modified tips, a horizontal leg with level, pitch-up, and pitch-down maneuvers was carried out on each flight at a preselected altitude.
c. Basic assumptions
The flow of the particles passing through the sample volume can be considered as consisting of two components. The first component is associated with the particles located in the probe’s forward-extended sample volume. Assuming that the probe’s arms are aligned parallel with the airflow, these particles do not impact the arms and remain intact passing through the probe’s sample area. The second component consists of fragments of shattered particles resulting from impact with the probe’s arm tips or inlets. These particles originate outside of the probe’s sample volume, and they are diverted into the sample volume by the cross-flow velocity component imparted to shattered fragments due to bouncing from the probe’s tip or inlet.
These two particle flows are independent of each other, and therefore the net concentration measured by the probe (
) can be expressed as
where
is the concentration of shattered particles, and
is the actual concentration of ice particles that would be measured by the probe in the absence of shattering. Equation (1) assumes that there is no secondary shattering caused by the interaction between the fragments of shattered particles and intact particles related to
. This assumption is supported by the large spatial separation of natural ice particles passing through the probe’s sample area. For example, for a concentration of 10 L−1, the average spacing of ice particles passing through the OAP-2DC sample area (~50 mm2) is approximately 2 m. This will result in a low occurrence of coincidences of the shattered fragments and intact particles in the sample volume. The effect of the aerodynamic fragmentation of ice particles (Korolev and Isaac 2005) is also neglected in Eq. (1). Aerodynamic fragmentation is most relevant to large fragile aggregates. The analysis of 2D imagery showed that such particles were occasionally present in the AIIE measurements. However, their infrequent occurrence minimized the effect of aerodynamic shattering in this study.
In the following sections, the particle concentration measured by the modified probes (
) will be considered as a proxy of the actual particle concentration (i.e.,
) and the concentration measured by the probe with the standard tips (
) will be considered to be contaminated by shattering (i.e.,
). The difference between the concentrations measured by the standard and modified probes (
) will be considered as a proxy of the concentration of the shattered particles (i.e.,
), and the shattering effect on the measured concentration will be estimated as
As it will be shown below, the modified tips still produce shattered particles. Therefore, the shattering estimates from Eq. (2) represent a lower estimate of the shattering effect.
The above relationships also refer to any additive microphysical parameters like the extinction coefficient (β) or IWC. Accordingly, the effect of shattering on the measurements of β and IWC can be estimated from relationships similar to Eq. (1):
The following two sections describe the results of the estimation of the effect of shattering on the FSSP and 2D imaging probes’ measurements.
3. Effect of ice shattering on FSSP measurements
a. Effect of ice shattering on the number concentration measurement
Korolev et al. (2013a) have discussed mechanisms leading to artifact creation on the FSSP. The standard FSSP housing has a tubular inlet. Shattered particles that may rebound toward the FSSP sample area originate on the flattened section of the leading edge of the inlet tube. When flying at a nonzero yaw or angle of attack, ice particles may also impact the inner walls of the inlet tube and rebound into the sample area. As described earlier, the inlet tube of the modified FSSP was removed and the original hemispherical tips were replaced with new designs (Fig. 2b).
The principle of the particle sizing of the FSSP is based on the measurement of forward scattered light by single particles passing through the sample area. The FSSP size bin thresholds are calibrated for spherical water droplets. Scattering properties of nonspherical ice particles are quite different from that of liquid droplets. In the following discussion, the FSSP data analysis was performed assuming that ice particles have the same scattering properties as liquid droplets. The authors concede that this may result in large errors in particle sizing. This specifically refers to large ice particles outside the nominal FSSP size range. Such particles may be undersized and measured as small ones. Fugal and Shaw (2009) attempted to establish ice size bin thresholds for the FSSP based on theoretical calculations of phase-scattering functions for different ice particles habits. However, pending laboratory validation of these results, FSSP measurements in ice clouds should be used with great caution, including those with the modified tips. The calculations of the FSSP extinction coefficient and IWC presented below were deduced assuming spherical ice particles. This has been done only for the purposes of comparisons of modified and standard FSSPs, and no scientific conclusions about cloud microphysics have been drawn based on these calculations.
It should be noted that the FSSP electronics used in this study did not have an interarrival time option. Therefore, the interarrival time algorithm for filtering shattered artifacts could not be applied to the FSSP data.
Figure 4a shows time series of the particle concentration measured by standard and modified FSSPs during ascent through a sequence of ice and mixed-phase clouds. In mixed-phase and liquid cloud regions, the total particle number concentrations measured by both FSSPs are in close agreement. The presence of supercooled liquid in this cloud is confirmed by the increase of the RID signal (Fig. 4d) due to ice accretion when the LWC is sufficiently high (>0.01 g m−3) and temperature is sufficiently low T < −4°C, appropriate to an airspeed of 100 m s−1 (Mazin et al. 2001). This specifically refers to the time period between 1437:40 and 1439:00 UTC. Liquid clouds between 1435:40 and 1437:30 UTC do not exhibit changes in the RID voltage because the air temperature is too warm (_T_ > −4°C) for ice accretion. Further evidence of the presence of droplets in the cloud regions identified as mixed phase comes from the appearance of high numbers of very small 2D images shown in Fig. 5 (part 1), with the occasional large ice crystal image. Such image records are typical when in cloud droplets of adequate size and/or supercooled large droplets (SLD).
Fig. 4.
Time series of (a) cloud particle concentration measured by the modified and standard FSSPs, (b) extinction coefficient measured by the CEP and calculated from the modified and standard FSSPs, (c) IWC measured by the Nevzorov probe and calculated from the modified and standard FSSPs, and (d) temperature and RID signal. Arrows 1 and 2 at the top of (a) indicate the areas where the CIP particle images in Fig. 5 have been sampled. The measurements were conducted on 8 Apr 2009 on the northwest side of Ottawa in a glaciated frontal cirrostratus–nimbostratus cloud system. An increase of the RID ramp voltage between 1437:40 and 1439:00 UTC indicates the presence of supercooled cloud liquid water. Liquid clouds between 1436:00 and 1439:00 UTC does not cause changes in the RID voltage since their LWC is below the RID sensitivity threshold (0.01 g m−3) and the air temperature is too warm (T > −4°C) (Mazin et al. 2001).
Citation: Journal of Atmospheric and Oceanic Technology 30, 11; 10.1175/JTECH-D-13-00115.1
Fig. 5.
Cloud particle 2D images measured in the areas indicated by arrows “1” (mixed phase) and “2” (ice cloud) in Fig. 4. The images were measured with a modified CIP with 15-μm pixel resolution. The particle images in 1 were sampled at P = 600 mb and T = −8°C, and 2 at P = 530 mb and T = −14°C.
Citation: Journal of Atmospheric and Oceanic Technology 30, 11; 10.1175/JTECH-D-13-00115.1
In glaciated ice clouds outside of these mixed-phase regions, the modified FSSP number concentration was 10−2 cm−3 or lower, whereas the standard FSSP with the inlet tube varied between 1 and 10 cm−3. In ice cloud regions, the absence of liquid droplets is supported by a flat or decreasing RID signal (Fig. 4d) and the CIP particle imagery (Fig. 5, part 2) showing ice particles with no evidence of liquid droplets.
The FSSP total number concentrations in the mixed-phase regions in Fig. 4a are quite low, which is often observed for clouds containing SLD. These concentrations are similar to the background concentrations of ice particles measured elsewhere by the standard FSSP in Fig. 4a. In this case, the standard FSSP is unable to distinguish mixed-phase cloud regions from ice regions based on number concentration due to artifacts in the ice regions. However, the modified FSSP shows only a very small signal in ice clouds and, in this particular case, can clearly identify liquid-containing clouds. The removal of the sample tube thus represents a potentially significant improvement of FSSP measurements in mixed-phase clouds.
Figure 6 shows an example of an exceptionally high concentration of ice particles measured by FSSP with the standard inlet tube. Similar to Fig. 4, concentrations measured by standard and modified FSSPs agree well in mixed-phase and liquid clouds. In regions of ice cloud the particle modified FSSP concentration is again very low at ~10−2 cm−3, but in a region of relatively high IWC at 1808 UTC, the standard FSSP concentration reaches 50 cm−3. Such a high concentration of ice particles may easily be erroneously interpreted as liquid cloud in the absence of other data. However, the flat signal of the Rosemount icing detector (Fig. 6d) indicates the absence of any significant amount of liquid in this cloud. The existence of true concentrations of small ice particles at such high values would have a significant impact on precipitation formation and radiation transfer in ice clouds. The comparisons in Fig. 6a indicate that the high concentration of small ice particles of the unmodified FSSP appears to be almost entirely artifact in this case.
Fig. 6.
As in Fig. 4, but measurements were conducted on 1 Apr 2009, northwest of Ottawa, in ice clouds associated with a frontal system.
Citation: Journal of Atmospheric and Oceanic Technology 30, 11; 10.1175/JTECH-D-13-00115.1
Figure 7 shows comparisons of the average size distributions measured by the standard and modified FSSPs for all-ice (Fig. 7a) and liquid or mixed-phase (Fig. 7b) clouds sampled during the AIIE project. The agreement is quite good in liquid and mixed-phase clouds (Fig. 7b), except for the larger sizes with D > 30 μm. This is consistent with Cober et al. (2001), who showed that ice crystals in mixed-phase clouds start to bias standard FSSP observed droplet spectra for sizes >30 μm, when the ice concentration (>100 μm) exceeds 1 L−1. In contrast to Fig. 7b, in ice clouds the ratio of the measured concentrations varies from 1.5 to 3 orders of magnitude, depending on the particle size bin (Fig. 7a).
Fig. 7.
Size distributions measured by the standard and modified FSSP in (a) ice clouds and (b) liquid and mixed-phase clouds, averaged over the entire AIIE flight dataset.
Citation: Journal of Atmospheric and Oceanic Technology 30, 11; 10.1175/JTECH-D-13-00115.1
Figure 8 presents scatter diagrams of total concentrations measured by both FSSPs in all-ice (Fig. 8a) and liquid or mixed-phase (Fig. 8b) clouds. Since the counting rate of the modified FSSP in ice clouds was very low, usually a few counts per minute, the concentrations in Fig. 8a were averaged over 30 s. The concentrations of liquid and mixed-phase clouds shown in Fig. 8b are 1-s averages. Figure 8a indicates that shattered particles dominate the unmodified probe number concentration in ice clouds. At the same time the correlation between the two probes is quite weak (0.53). However, the two probes agreed very well in liquid or mixed-phase clouds dominated by liquid droplets (Figs. 7b, 8b), supporting that the differences in ice clouds were not due to any fundamental optical and/or electronic response differences between the two probes, but rather were due to the removal of the sample tube and the modification of the tips.
Fig. 8.
(a) Scatter diagram of the cloud particle total number concentration measured by the standard and modified FSSPs in all ice clouds sampled during AIIE and (b) liquid and mixed-phase clouds sampled on 8 Apr 2009. Because of the very low sampling statistics by the modified FSSP, the data in (a) were averaged over 30-s time intervals. The data points in (b) are 1-s averages.
Citation: Journal of Atmospheric and Oceanic Technology 30, 11; 10.1175/JTECH-D-13-00115.1
For all cases observed during the AIIE project, the particle number concentration measured in ice clouds by the standard FSSP exceeded the concentration of the modified FSSP. Typically the 10-s average concentration measured by the modified probe varied from 0.0 to 0.1 cm−3, compared to 0.5 to 10 cm−3 for the unmodified probe.
b. Effect of ice shattering on the extinction coefficient and IWC measurements
Figures 4b,c and 6b,c show the comparisons of the extinction coefficient (
) and IWC (
) calculated from the standard FSSP measurements to those measured by the CEP (
) and Nevzorov deep-cone hot-wire TWC sensor (
). The FSSP extinction coefficient and IWC were calculated by applying the standard FSSP size calibration for water droplets. As mentioned earlier, this may cause large errors in the measurements of ice particles’ sizes and thus IWC and β.
As seen from Figs. 4b and 6b, in some ice cloud regions 
is as high as that measured by the CEP. The scatter diagram of the 
versus 
in Fig. 9a shows a high correlation coefficient (0.94), and on average the FSSP extinction is close to the CEP. This observation is consistent with that obtained by Heymsfield (2007).
Fig. 9.
Scatter diagrams of the (a) extinction coefficient and (b) IWC calculated from the standard FSSP measurements vs that measured by the CEP and Nevzorov deep-cone sensor. The measurements were conducted in ice clouds during the AIIE project.
Citation: Journal of Atmospheric and Oceanic Technology 30, 11; 10.1175/JTECH-D-13-00115.1
Similar to the extinction coefficient, the 
correlates well with 
(Figs. 4c, 6c), and on average the former is approximately 5 times lower than the latter (Fig. 9b).
These favorable correlations of FSSP-derived 
and 
, with the first-principles bulk instruments like the CEP and Nevzorov deep cone, without considering the effect of shattering on the FSSP measurements, could be falsely used as a justification of the validity of the FSSP measurements and eventually result in misleading conclusions about the presence and role of small ice particles in ice clouds. However, most of the standard FSSP counts in ice clouds appear to be shattered fragments, while the extinction coefficient and IWC calculated from the modified FSSP always remains close to zero (Figs. 4b,c and 6b,c).
4. Effect of ice shattering on 2DC and CIP probes measurements
a. Effect of ice shattering on the number concentration measurement
In this section we consider the effect of shattering on the OAP-2DC and CIP particle imaging probes. The sample volume of these probes is located between two arms that protrude forward from the probes’ canister (Fig. 2a). The original probes have arms with hemispherical- (2DC) and saucer-shaped (CIP) tips. For this study, they were replaced with arms with the spear-shaped K-tips designed to deflect particles away from the sample volume (Korolev et al. 2013a). As will be shown, 2DP measurements are less susceptible to shattering, and for this reason the 2DP tips remained unchanged.
Figures 10 and 11 show comparisons of particle images measured in ice clouds by pairs of standard (left) and modified (right) OAP-2DCs and CIPs, respectively. It is evident that the probes with the standard tips show a large number of small particles not observed on probes with modified tips. Since we can conceive of no other mechanism by which these modifications to the probes could have eliminated real particles, it is contended that these additional small particles must be artifacts resulting from shattering.
Fig. 10.
Comparisons of the ice particles images measured by two OAP-2DCs with the (left) standard and (right) modified tips. The images on the left have many more small particles than those on the right. The majority of the small particles on the left results from the probe tip shattering. The pixel resolution of both OAP-2DCs is 25 μm. The image sampling was conducted in a frontal cirrostratus–nimbostratus cloud system in the vicinity of Ottawa, 8 Apr 2009, at P = 690 mb and T = −15°C.
Citation: Journal of Atmospheric and Oceanic Technology 30, 11; 10.1175/JTECH-D-13-00115.1
Fig. 11.
As in Fig. 10, but for measurements by two CIPs. The pixel resolution of both CIPs is 15 μm. The image sampling was conducted in frontal nimbostratus clouds in the vicinity of Ottawa, 1 Apr 2009, at P = 580 mb and T = −13°C.
Citation: Journal of Atmospheric and Oceanic Technology 30, 11; 10.1175/JTECH-D-13-00115.1
Comparisons of the total number concentrations measured by standard and modified OAP-2DCs and CIPs are shown in Figs. 12a and 12b, respectively. No filters for shattering events were applied to the measurements of any of these instruments. The difference in the measured concentrations in Figs. 12a and 12b thus must result from shattering reduction due to the change in the shape of the probes tips. Concentrations measured by the probe with standard tips are in the range of 10’s to 100’s L−1, whereas the modified probe total concentrations are significantly lower. The difference becomes most pronounced in areas of the cloud with the largest particles (Fig. 12c; 1410–1440 UTC).
Fig. 12.
Time series of ice particle concentration measured in ice clouds by (a) two OAP-2DCs with standard and modified arms and tips, (b) two CIPs with the standard and modified and arm tips, (c) maximum size _D_max and median mass size _D_mmd, and (d) temperature and altitude. The measurements were conducted on 8 Apr 2009, northwest of Ottawa, in a glaciated frontal cirrostratus–nimbostratus cloud system.
Citation: Journal of Atmospheric and Oceanic Technology 30, 11; 10.1175/JTECH-D-13-00115.1
Modified and standard probe size distributions are shown in Fig. 13 for a 10-min period from the ice cloud shown in Fig. 12. The measurements were conducted in precipitating bullet rosettes (Fig. 10) formed at altitudes above 7.5 km and temperatures below −40°C. The data yield the important conclusion that although the effect of shattering is difficult to distinguish for particles larger 500 μm, it increases significantly toward smaller sizes, where it can strongly contaminate the spectrum. This behavior was generally observed in a variety of other cloud cases from the AIIE project.
Fig. 13.
Comparisons of size distributions before and after interarrival time artifact filtering corrections as measured by standard and modified (a) OAP-2DCs and (b) CIPs. Size distributions measured by the modified 2D-S and standard OAP-2DP are shown in both (a) and (b). The size spectra were averaged over the time interval 1409:00–1421:00 UTC, shown in Fig. 12 during descent from P = 520 mb and T = −24°C to P = 820 mb and T = −9°C.
Citation: Journal of Atmospheric and Oceanic Technology 30, 11; 10.1175/JTECH-D-13-00115.1
b. Effectiveness of the filtering algorithm
Shattering events were identified and filtered out by identifying fragmented images (Korolev and Isaac 2005) and using the interarrival time algorithm described by Field et al. (2006). The interarrival time algorithm uses the fact that after impact a shattered particle forms a spatial cluster of closely spaced fragments, with an interarrival time much shorter than that between natural particles. If the frequency distributions of the interarrival times associated with the shattered and intact particles are well separated, their interarrival times can then be used for the identification of shattering events. The cutoff time τ* is determined from the interarrival frequency distribution as the minimum between the two modes associated with short (shattered particles) and long (intact particles) interarrival time modes.
Interarrival time frequency distributions for the six imaging probes used in this study, measured in the large particle region of the ice cloud of Fig. 12, are shown in Fig. 14. The distributions associated with short and long interarrival times have a small overlap area, suggesting that they can be used to filter out shattered events. The cutoff time depends on the airspeed, sample area, and particle concentration. For the probes of this study, the cutoff time at 100 m s−1 usually varies in the range 10−4 s < τ* < 10−3 s.
Fig. 14.
Frequency distribution of the interarrival time intervals for (left) the standard and (right) modified versions of (top to bottom) OAP-2DC, CIP, and OAP-2DP and 2D-S measured in the large particle region of Fig. 12: 8 Apr 2009 1414:00–1421:00 UTC. The mode associated with the short interarrival time is assumed to result from shattered particles, whereas the natural particles are assumed to form the longer interarrival mode. The dotted lines show the cutoff time (τ*) used during data processing for filtering the shattered events. The numbers in the top left corners indicate the fraction of counts in the small interarrival mode (with t < τ) to the total number of counts.
Citation: Journal of Atmospheric and Oceanic Technology 30, 11; 10.1175/JTECH-D-13-00115.1
The ratio of the counts in the short interarrival mode (
) to the total counts (
) can be used to characterize the shattering efficiency ɛ = 
. During processing, particles associated with n s are rejected, whereas the rest of the particles associated with the long interarrival mode 
are accepted. It should be noted that ɛ reflects the proportion between accepted and rejected particle counts but not their concentrations. These ratios are included in Fig. 14. The comparisons of the standard and modified OAP-2DCs and CIPs (Figs. 14a–d) indicate that the modified tips shatter less. The diagram in Fig. 14e suggests that the OAP-2DP measurements are less affected by shattering relative to other probes. The likely reason for the OAP-2DP low sensitivity to shattering is the coarse pixel resolution (200 μm). Since most of the shattered particles have sizes smaller than 200 μm, the shattered fragments are not counted by the probe, and therefore they do not affect OAP-2DP measurements. The configuration of the OAP-2DP arms and tips is also less conducive to shattering effects.
Size distributions, corrected for the shattering events using the algorithm above, are also presented in Fig. 13. The concentration of small particles in standard probes corrected (thick blue lines) for shattering events is still higher than the modified probes with no corrections (thin red lines). This suggests that the 2DC and CIP interarrival time algorithms used here for shattering for cases as in Fig. 13 were insufficient to eliminate all the shattering artifacts.
This conclusion is supported by examining standard OAP-2DC and CIP imagery in Fig. 15, where the interarrival time rejected and accepted images are color coded. The artifact images are characteristically elongated and appear as donut looking (i.e., out of focus) (see also Figs. 10a, 11a). As seen from Fig. 15, the algorithm identifies and rejects a large fraction of these characteristically shaped artifact images. However, some fraction of such particle images evades rejection. This can be explained by the fact that in some cases only one fragment from a shatter group will end up intersecting the sample volume, while other group fragments pass outside. The interarrival times of these single fragment particles will be indistinguishable from the natural particles. A conceptual diagram of such events is shown in Fig. 16b.
Fig. 15.
Example of the results of the image rejection–acceptance processing performed for the standard (left) OAP-2DC and (right) CIP. Particles rejected due to interarrival time are highlighted in green; rejected multifragment images in blue; accepted partial images in yellow; and accepted complete images in white. Most particles are rejected due to short interarrival time. However, some small images (presumably resulted from shattering) were accepted. At the same time many images that appear as intact were rejected due to the interarrival time. This OAP-2DC and CIP image set was sampled in the same cloud shown in Fig. 12 at P = 715 mb and T = −12°C.
Citation: Journal of Atmospheric and Oceanic Technology 30, 11; 10.1175/JTECH-D-13-00115.1
Fig. 16.
Conceptual diagram of filtering shattered particles sampled by the probe with the help of the interarrival time algorithm. (a) Ideal case: 1) each shattering events consists of a multiple fragments passed through the sample volume and 2) the shattering events are spatially well separated from the intact particles. (b) Real case: 1) shattered fragments may pass through the sample volume alone and may not be associated with other shattered fragments and 2) intact particles may not be spatially separated with the shattered fragments.
Citation: Journal of Atmospheric and Oceanic Technology 30, 11; 10.1175/JTECH-D-13-00115.1
The fraction of small shattered fragments that evade the filtering algorithms may contribute significantly to the concentration of small particles in the 2D probe measurements. For the coherent illumination used in the 2D probes, the depth-of-field l of the sample volume V depends on particle size as 
(e.g., Korolev et al. 1998). Since the sample area A is inversely proportional to the particle concentration N, the calculated concentration of particles is inversely proportional to the square of their sizes (i.e., 
). For the case of CIP, the result of this dependence is such that one particle of D ~ 15 μm contributes to the concentration the same way as one hundred particles of D ~ 150 μm.
Another shortcoming of the filtering algorithm is the rejection of some large intact particles, some examples of which can be seen in Fig. 15. From a statistical viewpoint, intact large particles can pass through the sample volume at the same time as shattered particles (Fig. 16b), or two intact particles may pass through the sample volume with the interarrival time shorter than τ*. In either case, the intact particles will be rejected by the interarrival time algorithm. The comparisons of uncorrected and corrected size distribution in Fig. 13 suggest that the fraction of rejected intact particles with D > 500 μm is relatively small for the cases when the short and long interarrival modes are well separated. The analysis of cases with large overlaps of the short and long interarrival modes showed that the interarrival time algorithm becomes ineffective and detrimental due to the rejection of a large fraction of intact particles and the acceptance of a significant amount of shattered artifacts.
It appears that for the OAP-2DC and CIP probes, it is unlikely that the number concentration of particles less than 500 μm can be fully corrected using interarrival time algorithms in cases strongly contaminated by shattering (e.g., Fig. 13). Figure 13 indicates that the antishattering tips alone appear to be more effective at reducing shattering than the filtering algorithm alone. But it is also seen from Fig. 13 that the interarrival time algorithm applied to the probes with the modified tips appears to further filter out even more potential shattering events. This leads to two important conclusions. First, the modified tips still shatter ice particles, though at a much smaller rate relative to the original design. Second, the modified antishattering tips should be applied together with a shattering filter processing algorithm for the best effectiveness in minimizing the shattering effect.
The 2D-S with modified tips along with the manufacturer’s filtering algorithm provides the lowest concentrations at small particle sizes, while agreeing well with other probes at larger sizes (Fig. 13). This suggests that it has the best overall efficiency in identifying and filtering out shattered events relative to the other imaging probes considered here. The effectiveness of the shattering algorithms depends on the probe’s pixel resolution and response time of the electronics. The 2D-S has a smaller pixel resolution and shorter response time than the OAP-2DC and CIP, explaining its better performance. With the data available from this study, no conclusions can be drawn as to whether the antishattering algorithm with the standard 2D-S tips is more efficient than the antishattering tips alone in mitigating the shattering effect, as claimed in Lawson (2011).
c. Effect of ice shattering on the extinction coefficient and IWC measurements
Figures 17b and 17c show a time history of the extinction coefficient (
) and IWC (
) measured by the standard and modified OAP-2DC. For comparison purposes, the standard OAP-2DC measurements were left uncorrected (
, 
), whereas the modified 2DC data were corrected for shattering artifacts (
, 
).
The extinction coefficient was calculated from 2D data using the direct area calculation (DAC) (Korolev et al. 2013, manuscript submitted to J. Atmos. Oceanic Technol.) that derives
from the total area coverage of the particle shadowgraphs as
Here, Q ≈ 2 is the extinction efficiency; A j is the particle shadow area; L is distance between the OAP-2DC arms;
is the sample area covered by the probe’s laser beam having width w and the distance between the arms d; and L = U_Δ_t is the distance flown by the airplane moving at speed U during time Δ_t_. One of the advantages of the DAC technique is that it allows a more accurate estimation of the contribution into extinction of the partial images as compared to the estimation of the total particle-projected area
from size-to-area parameterizations. For the partial images, only some fraction of their area is viewed by the probe and therefore their sizes remain unknown.
The IWC was calculated from the 2D data as
where the particle mass
was calculated based on the size-to-mass parameterization
, with the coefficients a = 1.9 and b = 7.38 × 10−11 selected following Brown and Francis (1995). The dependence of the sample area
on particle size
was included, and partial images were treated as per Heymsfield and Parrish (1978) and Korolev and Sussman (2000). Because of the presence of large particles outside the size range of the OAP-2DC, the ice mass calculations from the OAP-2DC data alone may result in underestimation of IWC. For this reason the IWC was also calculated from size distributions combined from 2DC and 2DP data (
) (Fig. 17c).
Fig. 17.
Spatial changes of (a) particle concentration measured by two OAP-2DCs with standard and modified tips, (b) extinction coefficient measured by the CEP and deduced from the measurements of OAP-2DC with modified and standard tips, and (c) IWC measured by the Nevzorov probe and deduced from OAP-2DC with modified and standard tips and composite size distribution 2DC + 2DP with applied corrections. Interarrival time corrections were applied to the OAP-2DC with the modified tips but not to the standard OAP-2DC data. The gray color highlights the cloud region where liquid water was present, and it should be excluded from comparisons. The data were collected during ascent and descent when pressure and temperature changed from P = 420 mb and T = −25°C to P = 710 mb and T = 0°C.
Citation: Journal of Atmospheric and Oceanic Technology 30, 11; 10.1175/JTECH-D-13-00115.1
The values of 
and 
in Figs. 17b and 17c were compared with the extinction coefficient and IWC measured by the CEP and Nevzorov deep-cone sensor, respectively. As seen from these diagrams, 
and 
calculated for standard uncorrected OAP-2DC (blue line) are systematically higher than that measured by the modified and corrected OAP-2DC (red line). The differences vary from 5%–50% depending on the microphysical properties of the cloud.
Figure 18 shows distributions of the extinction coefficient and IWC calculated for the OAP-2DC for the measurements shown in Fig. 17. As seen from Fig. 18, the difference between the standard (blue lines) and modified (red line) OAP-2DC calculated for 
and 
is mainly associated with the particles smaller than 1 mm, and it becomes most pronounced for D < 500 μm. In Fig. 18a, the extinction coefficient in each size bin in Eq. (5) was calculated based on size-to-area parameterization 
, where γ = 0.55 and σ = 1.97 (Mitchell 1996), and the dependence of the sample area 
versus 
was included.
Fig. 18.
Comparisons of (a) extinction and (b) ice mass distributions measured by the standard 2DC, before and after interarrival time corrections and measured by the modified 2DC with applied corrections. The major differences are observed for D < 1 mm. The distribution measured by 2DP is also shown. The first four 2DP size bins are not shown because of large uncertainties concentration and sizing. The measurements were conducted on 4 Apr 2009 (Fig. 17) during descent from P = 460 mb and T = −21°C to P = 520 mb and T = −15°C.
Citation: Journal of Atmospheric and Oceanic Technology 30, 11; 10.1175/JTECH-D-13-00115.1
Figure 19 shows scatter diagrams between 
, 
, and 
, calculated from the standard uncorrected and corrected OAP-2DC versus those calculated from modified corrected OAP-2DC measurements in all AIIE ice clouds. It turns out that the concentration obtained from the uncorrected standard probe 
has weak correlation (0.59) with the concentration measured by the modified and corrected probe 
(Fig. 19a). After applying corrections to the standard probe, the correlation coefficient increases to 0.71. However, the scattering of the points still remains large (Fig. 19d). Basically this suggests that the shattering algorithms cannot identify and filter out all artifacts in the OAP-2DC measurements. This conclusion is consistent with that obtained in previous section about the low efficiency of the interarrival time algorithm.
Fig. 19.
Scatter diagrams of the (a) ice particle number concentration, (b) extinction coefficient, and (c) IWC measured by standard OAP-2DC with no corrections vs those measured by the modified OAP-2DC with applied corrections. (d),(e),(f) As in (a)–(c), but the standard OAP-2DC data were corrected on shattering for the entire AIIE flight dataset.
Citation: Journal of Atmospheric and Oceanic Technology 30, 11; 10.1175/JTECH-D-13-00115.1
As seen from Figs. 19b and 19c, the extinction measured by the OAP-2DC without corrections (
) is overestimated due to shattering on average of 22%, whereas IWC (
) is overestimated by 33%. Approximately the same estimates were provided earlier by Field et al. (2006) and Heymsfield (2007). The estimates herein of the effect of shattering on the extinction and IWC represent a lower estimate, since the modified and corrected 2DC data are still contaminated by the shattering artifacts. After applying corrections to the standard probe, the difference in estimated 
and 
between standard and modified probes was reduced to 3% (Fig. 19e) and 10% (Fig. 19f), respectively.
Estimates of the effect of shattering on the extinction coefficient and IWC calculated from the CIP data are hindered by limitations imposed by the truncation of particle images larger than 128 slices along the flight direction. Such images can be seen in Fig. 11. For this reason the analysis of the CIP extinction coefficient and IWC was omitted here.
5. Parameters affecting shattering
The in situ measurements presented in this study from the AIIE project, along with numerical simulations of bouncing and wind tunnel experiments (Korolev et al. 2013a), indicate that the effect of shattering is complex and depends upon a large number of parameters. These parameters can be split into three categories. The first category is related to the microphysical properties of the cloud particles: size, projected area, mass, habit, rebound coefficient, and surface tension. The second category is related to the environmental conditions: air pressure, temperature, and airspeed. The third category is related to the sampling configuration factors, such as yaw and the angle of attack, probe mounting location on the airplane, housing of the probe, and the shape of the tips and their temperature. Below we briefly describe the effect of each of the above parameters on shattering.
a. Effect of microphysical properties
1) Integral microphysical parameters
The effect of shattering manifests itself in an increase of the measured particle concentration, which, following section 2c, is estimated in Eq. (2) as 
. The value 
will be used here as a quantitative characteristic estimate of shattering on particle concentration. Below we attempt to establish a relationship between 
and various integral properties of cloud microstructure.
For the FSSP measurements 
was calculated as the difference between standard and modified probes; that is, 
. For the OAP-2DC, the value 
was determined as a difference between the measurements of the uncorrected standard and corrected modified probes; that is, 
. For the CIP data, the value 
was calculated the same way as for the OAP-2DC.
The 
is expected to result from the shattering of large particles and therefore would be dependent on their concentration. However, Fig. 20a shows poor correlation between 
and concentration measured by 2DC and 2DP (_N_2D). In fact, 
has its highest correlation coefficients with 
(0.94) and the extinction coefficient 
(0.91) (Figs. 20b,c). As mentioned above, 
and 
were measured by the CEP and Nevzorov probe, which are not sensitive to shattering, and their measurements are based on first principles.
Fig. 20.
(a) Scatter diagrams of 
vs ice particle concentration measured by the OAP-2DC with the modified tips, (b) extinction coefficient measured by CEP, and (c) IWC measured by the Nevzorov deep cone for the entire AIIE flight dataset.
Citation: Journal of Atmospheric and Oceanic Technology 30, 11; 10.1175/JTECH-D-13-00115.1
Figure 21 contains similar relationships for OAP-2DC, again showing high correlation between 
, 
, and 
. The CIP has very similar dependencies as those for OAP-2DC in Fig. 21 (not shown here for the sake of brevity).
Table 2 shows a summary of the correlation coefficients of 
, 
, and 
for different microphysical parameters. The correlation coefficients between mean particle size (_D_mean), mean volume size (_D_mm), maximum particle size (_D_max), and 
for all three probes are all relatively low. In contrast, the corresponding correlation coefficients for the integral size (
), extinction (
), and IWC (
) with 
are quite high.
Table 2.
Correlation coefficients between Δ_N_FSSP, Δ_N_2DC, Δ_N_CIP, and integral microphysical parameters. Here, _N_2D is the particle concentration measured by the modified and corrected 2DC and 2DP; _D_mean, _D_mm, and _D_max are the mean, median mass, and maximum particle sizes, respectively, calculated from composite 2DC + 2DP size distributions; _β_CEP is the extinction coefficient measured by CEP; and IWCNevz is the IWC measured by the Nevzorov probe.
Because of the similarity of the correlation coefficients between the 
–
and 
–
pairs, it is not clear whether the ice particles’ mass or their projected area dominate the effect of shattering. It can be speculated that the ice particle volume, which is representative of the particle mass, controls the number of shattering fragments. In this regard, it should be noted that a high correlation coefficient with extinction naturally results in a high correlation for IWC, since they are in turn usually highly correlated to each other in ice clouds. Previous studies have shown that both the ice particle–projected area (e.g., Mitchell 1996) and its mass (Locatelli and Hobbs 1974; Brown and Francis 1995; Heymsfield et al. 2010; etc.) can be well approximated by the same form of parameterization (i.e., 
), where the coefficient ν for both area and mass parameterizations in most ice clouds is close to 2. This explains the similarity of the distributions 
and 
in Fig. 18 and therefore the high correlation between IWC and the extinction coefficient in ice clouds. For all ice clouds sampled during the AIIE project the correlation coefficient between the 
and 
measurements is 0.96.
Since the FSSP, OAP-2DC, and CIP probes utilize different principles of particle sizing and have different inlet configurations and different sample areas, the similarity of the shattering effect on their measurements suggests that it may be common also to other probes. In other words, as suggested in Figs. 20 and 21 and Table 2, the number of fragments generated by shattered particles may be universally related in a linear manner to ice particles’ mass and projected area.
2) Ice particle habits
It is anticipated that the number of fragments created during shattering, may be also related to the ice particle habit. For example, ice pellets and delicate aggregates of dendrites with the same mass have a greatly different density, and thus they are anticipated to shatter differently on impact with a solid surface and generate a different number of fragments. However, the data collected during the AIIE flight campaign are insufficient to draw any conclusions regarding the dependence of the number of shattered fragments on particle habit.
3) Rebound coefficient
The rebound (restitution) coefficient determines the speed of the particle after impact and therefore the travel distance across the airflow. The rebound coefficient for the artificial ice particles produced in the Cox wind tunnel, which were identified to bounce without shattering, was estimated from the high-speed videos to vary from 0.55 to 0.85 (Korolev et al. 2013a). However, the rebound coefficient for the fragments of naturally grown shattered ice particles remains unknown.
b. Effect of environmental conditions
1) Air pressure and temperature
Air pressure and temperature determine the viscosity of the air, which controls the drag force and eventually affects the particle travel distance across the airflow after rebounding from the probe’s inlet or tips. Numerical simulations of the bouncing of spherical ice particles have shown a strong pressure effect, with increased travel distance across the flow with decreasing pressure (Korolev et al. 2013a). One of the important consequences of this effect is that at high altitude the measurements will be more likely contaminated by rebounding smaller particles or fragments than at lower altitudes (higher pressure) due to the higher likelihood of reaching the probe sample volume. This effect may be particularly important for the FSSP, used by many research groups for measurements of cirrus clouds in the upper troposphere.
2) Airspeed
Airspeed affects shattering in two important ways. First, airspeed is a determining parameter in the kinetic momentum of the rebounding particle, which affects the travel distance the across the airflow. Numerical simulations have shown that the cross and parallel velocity components of the rebounding particle are directly proportional to the initial particle speed. As a result, the rebounding particle arrives at approximately the same location in the sample volume downstream (Korolev et al. 2013a), independent of velocity.
Second, airspeed determines the kinetic energy of the impact of the particle, which is converted into surface energy and ultimately thermal energy through viscous dissipation, inducing ice defect production down at the molecular scale (Vidaurre and Hallett 2009). The kinetic energy controls the number and size of shattering fragments created on impact with the solid surface. As the airspeed increases, shattering fragments are expected to be smaller and more numerous. The effect of the airspeed on shattering can only be discussed here conceptually, and further laboratory and theoretical studies are required for its quantification.
c. Effect of sampling configuration factors
1) Yaw and angle of attack
At nonzero yaw or angle of attack, the inner surface of one of a probe’s arms can be exposed to the airflow. Particles that would normally pass parallel to these arms can now impact, and may shatter and then be redirected into the sample volume, resulting in contamination of measurements by shattering artifacts. Analysis of high-speed videos obtained during the Cox wind tunnel experiments demonstrated shattering at a nonzero angle of attack (Korolev et al. 2013a). The pitch-up and pitch-down maneuvers conducted during the AIIE project also showed an enhanced number of shattering artifacts when the 2DC arms were not aligned with the airflow.
2) Shape of the inlet
This study has shown that the shape of the probe inlet plays one of the most important roles in determining the contamination of measurements by shattered fragments. Open-concept inlets, consisting of protruding forward arms (e.g., optical array probes), have significant advantages over closed-concept inlets typically consisting of tubes (e.g., FSSP, CAS, CPI, and others). Tubular inlets typically have a larger area projected into the airflow relative to forward protruding arms and therefore will shatter more particles at the nonzero angle of attack than open-concept inlets. This conclusion is consistent with the observations of McFarquhar et al. (2007). Numerical simulations have shown that tubular inlets also create more air disturbance in the vicinity of the sample area relative to the open-concept inlets (Korolev et al. 2013a).
3) Temperature of the tips
Tip heating may mitigate the effects shattering. If the temperature of the tips is high enough, the ice particles may melt or stick to the surface without bouncing. Analysis of high-definition videos from the Cox wind tunnel tests indicates that applying more than 100 W to the OAP-2DC tip at 80 m s−1 and P = 1000 mb results in melting and sticking of small ice particles instead of bouncing (Korolev et al. 2013a). Such behavior of ice particles interacting with a heated surface is similar to that used in the hot-wire probes for measurements of condensed water content. Presumably, this heating may be beneficial to tubular inlets as well if appropriately designed.
4) Housing of the particle probe
A probe’s housing affects the airflow around the probe and determines accelerations and decelerations experienced by ice particles approaching its sample volume. Large aggregates of dendrites or needles with weak bonding will generally experience deceleration on approach to the sample volume of a 2D probe, which may induce breakup into smaller fragments (Korolev and Isaac 2005).
5) Mounting location on the airplane
If a particle probe is located in an area with a disturbed airflow (e.g., too close to the fuselage or wing) then particles may experience large stresses and get fragmented before passing though the sample volume. Numerical simulation suggests that depending on the flight altitude and the airspeed, ice particles rebounding from the airframe may contaminate a significant area around the aircraft (up to several tens of centimeters) (e.g., Engblom and Ross 2003). This should be taken into account when choosing locations for cloud microphysical instrumentation on an airplane.
Accounting for all of the above effects in numerical models of shattering is a challenging problem. The effort is hindered by the lack of information on the rebound coefficient of ice, the number and size of shattered fragments, the bounce angle of the shattered fragment, and potentially many other factors. In general, attempts to quantify the effect of shattering using one parameter may be incomplete.
6. Discussion
a. The existence of small ice particles
Figure 13 shows that optical array probes still produce some concentrations at the smallest size, even after implementing shattering mitigating tips and implementing shattering software filters. At the present time, these small particles cannot be definitively stated to be real. Such particles measured by optical array probes may be subject to out-of-focus errors away from the object plane and sizing and counting errors due to the discrete nature of their particle sizing. This results in a potentially large error in depth of field, and thus sample area, particularly for small particles. These problems exist for spherical water drops and are more complicated to quantify for irregular ice particles (Korolev et al. 1998). For the 2DC and 2DP probes, additional errors can be caused by the response time of the electronics and the lost leading image slice. As a consequence, the Environment Canada group has considered concentration measurements for particles less than 4 pixels in size (100 and 800 μm for the 2DC and 2DP respectively) unreliable. Similar errors are possible for the CIP and 2DS probes, where particles 4 pixels in size for the probes used in this study would be 60 and 40 μm in size. All measurements below these ~4 pixel-sized particles must be treated with caution.
Since the FSSP response to ice particles remains largely uncharacterized, the possibility that the FSSP counts ice particles outside its nominal size range cannot be excluded. Figure 22 shows a scatter diagram of the best estimate of 2D concentration (
) versus the concentration measured by the modified FSSP. The figure reveals that 
does not correlate well with 
and generally dominates over _N_2D. This suggests that FSSP measurements cannot be solely explained by the counting of particles outside its size range. It has been shown earlier that in spite of significant mitigation, modified tips produce some shattered particles that are measured, and therefore it is possible that the relatively high concentration of small ice measured by the FSSP is still from shattered fragments. But it also cannot be excluded that some of the FSSP counts are true natural small ice particles. At this point in time, the individual contributions of each of the above three possible explanations for FSSP particle counts cannot be quantified.
Fig. 22.
Scatter diagram of ice particle concentration measured by the modified OAP-2DC with the corrections vs modified FSSP for the entire AIIE flight dataset. Both concentrations are considered to be the best estimates in the size ranges associated with each of these probes.
Citation: Journal of Atmospheric and Oceanic Technology 30, 11; 10.1175/JTECH-D-13-00115.1
Unfortunately, the question of the existence and abundance of small ice particles in ice clouds remains open.
b. Historical data
Large OAP-2DC and FSSP datasets have been collected by the community over the past 30 yr and have been used for the parameterization of cloud microphysics for weather predictions and climate models and for the validation of remote sensing instruments. This study has shown that shattering may significantly contaminate such microphysical measurements in ice clouds. It was also shown that the effectiveness of the antishattering software filtering algorithms applied to OAP-2DC measurements is relatively poor, and many shattering artifacts remain present in the data. This raises important questions: (i) to what extent can the historical data be used for microphysical characterization of ice clouds, and (ii) can the historical data be reanalyzed to filter out the data affected by shattering?
As was shown above, for 2D probes shattering affects the number concentrations primarily for particles smaller than ~500 μm (Fig. 13). Depending on the size distribution, the overestimation of the number concentration at a particular size may reach a factor of 100. Such errors are excessively large and such data should not be used in cloud parameterizations and validations of numerical simulations. The possibility of adequately improving retrieval algorithms and artifact filtering for the D < 500 μm part of the size range is at the moment an open issue.
The larger part of the particle size distributions (D > 500 μm) measured by 2D probes are much less affected by shattering. Therefore, it stands to reason that historical OAP-2DC measurements of concentration can be used for D > 500 μm after applying existing antishattering corrections.
The effect of the shattering on the higher moments of particle size distributions (β, W, etc.) was found for this study to be much less than for the number concentration. This is because the contribution to the higher moments was dominated by particles D > 500 μm, which are less affected by shattering. For the 2D probes, the systematic biases in the AIIE extinction coefficient and IWC measurements were of the order of 20%–30% (section 4c). For radar reflectivity, the systematic biases due to shattering were approximately 10%. Such biases are within the errors related to particle sizing from 2D imagery and the accuracy of the size-to-area and size-to-mass parameterizations. Therefore, the extinction, IWC, and radar reflectivity calculated from historical data may be acceptable after antishattering corrections, depending on the application.
These conclusions refer to the size distributions with 
> 1000 μm; that is, when the large particles that are the presumed sources of significant shattering are present. In this work, we do not have enough information to quantify the effect of shattering for the cases with narrow particle size distributions and 
< 200 μm, which are typical for cirrus clouds.
The results of this study show that the FSSP historical data should generally not be used for the characterization of the microstructure of ice clouds. The majority of the FSSP counts measured in ice clouds appear to be shattering artifacts. At the moment the development of retrieval procedures for the FSSP measurements in ice clouds has not been extensively studied and must address potentially complex problems.
7. Conclusions
This study has quantified the effect of particle shattering on cloud particle probe measurements. The quantification was based on comparisons of two of the same type of instrument flown side by side, one original and one that had been modified with special arms and tips to mitigate shattering. Since both probes had identical optics and electronics, the difference in their measurements was ascribed to the effect of shattering. Such an approach has a distinct advantage over previous studies where the effect of shattering was estimated by comparing particle probes with a different principle of measurement and different probe-housing configurations.
This work was focused on the quantification of the shattering effect for the FSSP, OAP-2DC, and CIP. Since the modified tips are also prone to some shattering, the estimated effect of shattering should be considered as a lower estimate (i.e., the actual effect of shattering on the measurements is expected to be higher than that obtained here).
The results demonstrated that the shattering of ice particles is a serious problem for airborne microphysical characterization of ice clouds. The analysis of a variety of cloud cases from the AIIE flight program yielded the following important conclusions:
- The majority of the counts measured in ice clouds from an FSSP with a standard sample tube are shattered fragment artifacts. Such measurements can result in misleading conclusions and they should be used with great caution.
- The number concentration of particles in part of the size range with diameters less than ~500 μm can be strongly contaminated by shattering artifacts for OAP-2DC and CIP measurements. The overestimation of the number concentration at a particular size may reach a factor of 100. The measured concentration of particles with D > 500 μm is less affected by shattering and can be used after applying shatterer-filtering software corrections.
- The number of shattered fragments measured by the particle probes depends on a large variety of different parameters associated with the microphysical propertied of particles, environmental conditions, and the sampling arrangement (section 5). The bulk extinction coefficient and IWC were found to be linearly related to the number of shattered artifacts measured by all three probes. This finding can be used in future retrieval algorithms for the approximate corrections of the shattering effect.
- Since the extinction coefficient, mass, and radar reflectivity of typical size distributions measured in the overall AIIE dataset are normally dominated by the larger particles, estimates of these parameters from 2D particle images are significantly less affected by shattering than the number concentration. In most cases, the estimated overestimation attributable to shattering artifacts in the extinction and IWC calculations is approximately 20% and 30%, respectively, and for the radar reflectivity ~10%.
- It was demonstrated that the interarrival time algorithm alone is unable to filter out all shattering events from OAP-2DC and CIP measurements. But by modifying the probe tips (K-tips), and by applying interarrival time shatterer-filtering algorithms, the effects of shattering can be significantly reduced. The methods are complementary and should be used together to maximize the mitigation efficacy. For future flight campaigns, it is recommended that research groups adopt modified K-tips and apply such algorithms to reduce the effect of shattering.
It must be cautioned that the entire range of possible natural cloud conditions was not examined during the limited number of the AIIE research flights. Most of the data were obtained in clouds with relatively broad ice particle size distributions with 
> 1 mm, where the large particles thought to be the source of shattered particles are present. In this regard, more measurements should be conducted for the characterization of the shattering effect for cases with narrow size distributions typical for cirrus clouds.
Large sets of OAP-2DC data collected by the community over the past 30 yr have been used for parameterization of cloud microphysics for numerical weather and climate models and for the validation of remote sensing instruments. Many of these datasets are likely to have been contaminated by shattering artifacts and should be reexamined. A series of dedicated flight campaigns to study the effect of shattering and other problems related to the accuracy of ice measurement should be considered as future high priority for the cloud physics community.
In spite of the large reduction in the number concentration distributions of ice particles by modifying probes and applying filtering algorithms, measurements in ice clouds are still dominated by small particles. At this time it cannot be determined whether this is real or a result of remaining shattering events and/or other contaminating factors. Other instrument problems and limitations (depth-of-field definition problems, out-of-focus images, image digitization, etc.) contribute additional uncertainty to the accuracy of small ice particle measurements. Given the fundamental scientific importance of this issue, further efforts must be invested toward the improvement of small ice measurements.
Acknowledgments
This work was funded by Environment Canada, Transport Canada, the Federal Aviation Administration, and NASA. Special thanks to Sara Lance (NOAA), Jorge Delgado (NOAA), and Dave Rogers (NCAR) for loaning their cloud particle probes for the AIIE project. We express our sincere gratitude to the NRC pilots Anthony Brown, John Aitken, and Tim Leslie for their outstanding cooperation during the AIIE flight operations. The efforts of the NRC Project Manager Dave Marcotte and NRC technicians in preparing and organizing the Convair-580 flights, and of the participants from DMT Inc. for their self-funded support during the AIIE project, are greatly appreciated. The support and analysis of 2D-S data by Brad Baker and Paul Lawson from SPEC Inc. are much appreciated. Authors express their gratitude to Darrel Baumgardner and two anonymous reviewers for their thoughtful comments.
REFERENCES
- Brown, P. R. A., and Francis P. N. , 1995: Improved measurements of the ice water content in cirrus using a total-water probe. _J. Atmos. Oceanic Technol., 12, 410–414.
Brown, P. R. A., and FrancisP. N. , 1995: Improved measurements of the ice water content in cirrus using a total-water probe. J. Atmos. Oceanic Technol., 12, 410–414.
)| false - Cober, S. G., Isaac G. A. , Korolev A. V. , and Strapp J. W. , 2001: Assessing cloud-phase conditions. _J. Appl. Meteor., 40, 1967–1983.
Cober, S. G., IsaacG. A. , KorolevA. V. , and StrappJ. W. , 2001: Assessing cloud-phase conditions. J. Appl. Meteor., 40, 1967–1983.
)| false - Cooper, W. A., 1977: Cloud physics investigation by the University of Wyoming in HIPLEX 1977. Bureau of Reclamation Rep. AS 119, 321 pp.
- Engblom, W. A., and Ross M. W. , 2003: Numerical model of airflow induced particle enhancement for instruments carried by the WB-57 aircraft. NASA Aerospace Rep. ATR-2003(5084)-01, 29 pp.
- Field, P. R., Wood R. , Brown P. R. A. , Kaye P. H. , Hirst E. , Greenaway R. , and Smith J. A. , 2003: Ice particle interarrival times measured with a fast FSSP. _J. Atmos. Oceanic Technol., 20, 249–261.
Field, P. R., WoodR. , BrownP. R. A. , KayeP. H. , HirstE. , GreenawayR. , and SmithJ. A. , 2003: Ice particle interarrival times measured with a fast FSSP. J. Atmos. Oceanic Technol., 20, 249–261.
)| false - Field, P. R., Heymsfield A. J. , and Bansemer A. , 2006: Shattering and particle interarrival times measured by optical array probes in ice clouds. _J. Atmos. Oceanic Technol., 23, 1357–1370.
Field, P. R., HeymsfieldA. J. , and BansemerA. , 2006: Shattering and particle interarrival times measured by optical array probes in ice clouds. J. Atmos. Oceanic Technol., 23, 1357–1370.
)| false - Fugal, J. P., and Shaw R. A. , 2009: Cloud particle size distributions measured with an airborne digital in-line holographic instrument. _Atmos. Meas. Tech., 2, 259–271.
Fugal, J. P., and ShawR. A. , 2009: Cloud particle size distributions measured with an airborne digital in-line holographic instrument. Atmos. Meas. Tech., 2, 259–271.
)| false - Gardiner, B. A., and Hallett J. , 1985: Degradation of in-cloud forward scattering spectrometer probe measurements in the presence of ice particles. _J. Atmos. Oceanic Technol., 2, 171–180.
Gardiner, B. A., and HallettJ. , 1985: Degradation of in-cloud forward scattering spectrometer probe measurements in the presence of ice particles. J. Atmos. Oceanic Technol., 2, 171–180.
)| false - Gayet, J.-F., Brown P. R. , and Albers F. , 1993: A comparison on in-cloud measurements obtained with six PMS 2D-C probes. _J. Atmos. Oceanic Technol., 10, 180–194.
Gayet, J.-F., BrownP. R. , and AlbersF. , 1993: A comparison on in-cloud measurements obtained with six PMS 2D-C probes. J. Atmos. Oceanic Technol., 10, 180–194.
)| false - Gayet, J.-F., Febvre G. , and Larsen H. , 1996: The reliability of the PMS FSSP in the presence of small ice crystals. _J. Atmos. Oceanic Technol., 13, 1300–1310.
Gayet, J.-F., FebvreG. , and LarsenH. , 1996: The reliability of the PMS FSSP in the presence of small ice crystals. J. Atmos. Oceanic Technol., 13, 1300–1310.
)| false - Gerber, H., Takano V. , Garrett T. J. , and Hobbs P. V. , 2000: Nephelometer measurements of the asymmetry parameter, volume extinction coefficient, and backscatter ratio in Arctic clouds. _J. Atmos. Sci., 57, 3021–3034.
Gerber, H., TakanoV. , GarrettT. J. , and HobbsP. V. , 2000: Nephelometer measurements of the asymmetry parameter, volume extinction coefficient, and backscatter ratio in Arctic clouds. J. Atmos. Sci., 57, 3021–3034.
)| false - Heymsfield, A. J., 2007: On measurements of small ice particles in clouds. _Geophys. Res. Lett., 34, L23812, .
Heymsfield, A. J., 2007: On measurements of small ice particles in clouds. Geophys. Res. Lett., 34, L23812, doi:10.1029/2007GL030951.
)| false - Heymsfield, A. J., and Parrish J. , 1978: A computational technique for increasing the effective sampling volume of the PMS two-dimensional particle size spectrometer. _J. Appl. Meteor., 17, 1566–1572.
Heymsfield, A. J., and ParrishJ. , 1978: A computational technique for increasing the effective sampling volume of the PMS two-dimensional particle size spectrometer. J. Appl. Meteor., 17, 1566–1572.
)| false - Heymsfield, A. J., and Platt C. M. R. , 1984: A parameterization of the particle size spectrum of ice clouds in terms of the ambient temperature and ice water content. _J. Atmos. Sci., 41, 846–855.
Heymsfield, A. J., and PlattC. M. R. , 1984: A parameterization of the particle size spectrum of ice clouds in terms of the ambient temperature and ice water content. J. Atmos. Sci., 41, 846–855.
)| false - Heymsfield, A. J., Schmitt C. , Bansemer A. , and Twohy C. H. , 2010: Improved representation of ice particle masses based on observations in natural clouds. _J. Atmos. Sci., 67, 3303–3318.
Heymsfield, A. J., SchmittC. , BansemerA. , and TwohyC. H. , 2010: Improved representation of ice particle masses based on observations in natural clouds. J. Atmos. Sci., 67, 3303–3318.
)| false - Jensen, E. J., and Coauthors, 2009: On the importance of small ice crystals in tropical anvil cirrus. _Atmos. Chem. Phys., 9, 5519–5537.
Jensen, E. J., and Coauthors, 2009: On the importance of small ice crystals in tropical anvil cirrus. Atmos. Chem. Phys., 9, 5519–5537.
)| false - Knollenberg, R. G., 1976: Three new instruments for cloud physics measurements: The 2D-spectrometer probe, the forward scattering spectrometer probe and the active scattering spectrometer probe. Preprints, Int. Conf. on Cloud Physics, Boulder, CO, Amer. Meteor. Soc., 554–561.
- Korolev, A. V., and Sussman B. , 2000: A technique for habit classification of cloud particles. _J. Atmos. Oceanic Technol., 17, 1048–1057.
Korolev, A. V., and SussmanB. , 2000: A technique for habit classification of cloud particles. J. Atmos. Oceanic Technol., 17, 1048–1057.
)| false - Korolev, A. V., and Isaac G. A. , 2005: Shattering during sampling by OAPs and HVPS. Part I: Snow particles. _J. Atmos. Oceanic Technol., 22, 528–542.
Korolev, A. V., and IsaacG. A. , 2005: Shattering during sampling by OAPs and HVPS. Part I: Snow particles. J. Atmos. Oceanic Technol., 22, 528–542.
)| false - Korolev, A. V., Strapp J. W. , and Isaac G. A. , 1998: Evaluation of the accuracy of PMS optical array probes. _J. Atmos. Oceanic Technol., 15, 708–720.
Korolev, A. V., StrappJ. W. , and IsaacG. A. , 1998: Evaluation of the accuracy of PMS optical array probes. J. Atmos. Oceanic Technol., 15, 708–720.
)| false - Korolev, A. V., Isaac G. A. , Cober S. , Strapp J. W. , and Hallett J. , 2003: Microphysical characterization of mixed-phase clouds. _Quart. J. Roy. Meteor. Soc., 129, 39–66.
Korolev, A. V., IsaacG. A. , CoberS. , StrappJ. W. , and HallettJ. , 2003: Microphysical characterization of mixed-phase clouds. Quart. J. Roy. Meteor. Soc., 129, 39–66.
)| false - Korolev, A. V., Emery E. F. , Strapp J. W. , Cober S. G. , Isaac G. A. , Wasey M. , and Marcotte D. , 2011: Small ice particles in tropospheric clouds: Fact or artifact? Airborne icing instrumentation evaluation experiment. _Bull. Amer. Meteor. Soc., 92, 967–973.
Korolev, A. V., EmeryE. F. , StrappJ. W. , CoberS. G. , IsaacG. A. , WaseyM. , and MarcotteD. , 2011: Small ice particles in tropospheric clouds: Fact or artifact? Airborne icing instrumentation evaluation experiment. Bull. Amer. Meteor. Soc., 92, 967–973.
)| false - Korolev, A. V., Emery E. F. , and Creelman K. , 2013a: Modification and tests of particle probe tips to mitigate effects of ice shattering. _J. Atmos. Oceanic Technol., 30, 690–708.
Korolev, A. V., EmeryE. F. , and CreelmanK. , 2013a: Modification and tests of particle probe tips to mitigate effects of ice shattering. J. Atmos. Oceanic Technol., 30, 690–708.
)| false - Korolev, A. V., Strapp J. W. , Isaac G. A. , and Emery E. , 2013b: Improved airborne hot-wire measurements of ice water content in clouds. _J. Atmos. Oceanic Technol., **30, 2121–2131.
Korolev, A. V., StrappJ. W. , IsaacG. A. , and EmeryE. , 2013b: Improved airborne hot-wire measurements of ice water content in clouds. J. Atmos. Oceanic Technol., 30, 2121–2131.
)| false - Lawson, R. P., 2011: Effects of ice particles shattering on the 2D-S probe. _Atmos. Meas. Tech., 4, 1361–1381.
Lawson, R. P., 2011: Effects of ice particles shattering on the 2D-S probe. Atmos. Meas. Tech., 4, 1361–1381.
)| false - Locatelli, I. D., and Hobbs P. V. , 1974: Fall speed and masses of solid precipitation particles. _J. Geophys. Res., 79, 2179–2185.
Locatelli, I. D., and HobbsP. V. , 1974: Fall speed and masses of solid precipitation particles. J. Geophys. Res., 79, 2179–2185.
)| false - Mazin, I. P., Korolev A. V. , Heymsfield A. , Isaac G. A. , and Cober S. G. , 2001: Thermodynamics of icing cylinder for measurements of liquid water content in supercooled clouds. _J. Atmos. Oceanic Technol., 18, 543–558.
Mazin, I. P., KorolevA. V. , HeymsfieldA. , IsaacG. A. , and CoberS. G. , 2001: Thermodynamics of icing cylinder for measurements of liquid water content in supercooled clouds. J. Atmos. Oceanic Technol., 18, 543–558.
)| false - McFarquhar, G. M., Um J. , Freer M. , Baumgardner D. , Kok G. L. , and Mace G. , 2007: Importance of small ice crystals to cirrus properties: Observations from the Tropical Warm Pool International Cloud Experiment (TWP-ICE). Geophys. Res. Lett.,34, L13803, .
- Mitchell, D., 1996: Use of mass- and area-dimensional laws for determining precipitation particle terminal velocity. _J. Atmos. Sci., 53, 1710–1723.
Mitchell, D., 1996: Use of mass- and area-dimensional laws for determining precipitation particle terminal velocity. J. Atmos. Sci., 53, 1710–1723.
)| false - Twohy, C. H., Schanot A. J. , and Cooper W. A. , 1997: Measurement of condensed water content in liquid and ice clouds using an airborne counterflow virtual impactor. _J. Atmos. Oceanic Technol., 14, 197–202.
Twohy, C. H., SchanotA. J. , and CooperW. A. , 1997: Measurement of condensed water content in liquid and ice clouds using an airborne counterflow virtual impactor. J. Atmos. Oceanic Technol., 14, 197–202.
)| false - Vidaurre, G., and Hallett J. , 2009: Particle impact and breakup in aircraft measurement. _J. Atmos. Oceanic Technol., 26, 972–983.
Vidaurre, G., and HallettJ. , 2009: Particle impact and breakup in aircraft measurement. J. Atmos. Oceanic Technol., 26, 972–983.
)| false