On measurements of small ice particles in clouds



[1] For many years it has been recognized that some aircraft cloud microphysical measurements may be contaminated by ice hydrometeors shattering on probe inlets. Small ice particle concentrations measured by the forward scattering spectrometer probe (FSSP) have been commonly accepted but are especially liable to overestimates. This study investigates this result for two additional airborne microphysical sensors, the cloud and aerosol spectrometer (CAS) and the cloud integrating nephelometer (CIN). The results indicate that conclusions from previous studies of the radiative effects of small ice particles need to be reevaluated. A model has been developed describing probe responses to different combinations of ice water content (IWC) and large ice particle concentrations that may be useful in identifying areas where contributions from small particles are significant.

1. Introduction

[2] Early comparisons of satellite-retrieved, cirrus cloud properties suggest that in-situ microphysical probes inadequately measured sub-50 μm diameter ice particles [Wielicki et al., 1990]. Impactor-type probes were then deployed to measure ice particles below 50 μm. Such probes were limited: the measurements are saturated in all but low IWC regions; large ice shatters upon impaction and data processing is tedious. Digital imaging probes, notably the Cloud Particle Imager (CPI) [Baker and Lawson, 2006] have been developed that measure from 10 or 20 μm and above. The resulting images are highly resolved but the sample volume is currently uncertain and the size range it covers may be insufficient to detect all small ice particles. Since the early 1990s the FSSP has been increasingly used to estimate small ice PSDs. Evidence indicates, however, that ice particles several hundred microns or larger, shatter on the FSSP's forward surfaces, producing artifacts and rendering the small ice data uncertain [e.g., Field et al., 2003]. Even the 2DC measurements can be contaminated by fragments of large ice produced by shattering [Korolev and Isaac, 2005].

[3] Bulk ice cloud properties, including the IWC and the volume extinction coefficient (σ) have been derived using a combination of PSDs from the FSSP and 2D imaging probes. These estimations are now being measured by new instruments: the counterflow virtual impactor (CVI) [Twohy et al., 1997] and other probes measuring total condensed water, TWC = IWC + liquid water content (LWC), and the CIN [Gerber et al., 2000] that measures σ. Evaluation of these direct measurements still requires accurate PSDs for small particles, providing data when the detection thresholds of the CVI, CIN, and related instruments are inadequate and for identifying new ice formation zones.

2. Methods

[4] The data shown here are from the Cirrus Regional Study of Tropical Anvils and Cirrus Layers Florida Area Cirrus Experiment (CRYSTAL- FACE, hereinafter C-F) in southern Florida during July 2002. The University of North Dakota Citation and the NASA WB57 research aircraft sampled cloud layers from 0 to −70°. The aircraft carried FSSPs, 2D imaging probes measuring PSDs for ice >50 μm and particle images, and CINs. NASA's WB57 also carried a CAS that sampled small ice particles from about 2 to 50 μm diameter [Baumgardner et al., 2002] and the Citation carried a CVI. Small 2DC particles may be a combination of real and shattered large ice particles; most of the latter are removed objectively according to the method given by Field et al. [2006]. Even so, the 2DC measurements may be unreliable for D < 100 μm [Strapp et al., 2001]. The IWCs in small, FSSP particles are derived assuming solid ice spheres; these may be an overestimate by up to 25% [see Ryan et al., 1976]. Inherent uncertainties for basic FSSP size calibrations are ≈30% and for sample volumes ≈15% [Baumgardner, 1983] and taken together may limit the uncertainty in the small particle IWC to no better than a factor of 2. Dispersion in the IWC distribution is produced by additional FSSP and CAS measurement errors. Irregularly -shaped small particles further add to sizing uncertainty. Large, 2D particles have densities as given by Heymsfield et al. [2004]. Estimates of the accuracy are given below. The particle cross-sectional areas per unit volume of air, (Aσ/2), uses the FSSP PSD, assuming spheres and areas derived from the 2D images of the particles, with inherent uncertainties of a factor of 2.

[5] Examination of the C-F data and from three additional field campaigns indicates that in regions containing only ice, there are linear relationships between the IWC from large particles from the imaging probes, or of both large and small particles from the CVI and that derived from the FSSP or from the CAS. Partitioning of the data according to the median volume diameter D0, a characteristic size of the PSD given by D0 = 3.67/λ where λ is an exponential fit to the 2D PSD, did not change the linear dependencies. Linear relationships were also found between A (σ/2) from the imaging probes and from the FSSP and the CIN. Conceptually, this indicate that as large particles shatter on the inlets, a constant fraction of the remnants are swept into the probe's sampling area. This result further suggests that the energy of impact of the ice crystals on the FSSP's leading edge is large enough that the initial characteristics of the fragmenting ice crystal are secondary to the final configuration of small ice crystal fragments.

[6] In only-ice regions the following empirical relationships are found, where C is an empirically derived, shattering coefficient; the basic assumption being that large particles shatter and increase the mass or area of ice in the probe measurements.

equation image

where IWCsmall is what we are trying to determine, C2D,FSSP is the shattering coefficient based upon measurements of the particles above about 50 μm in maximum diameter interpreted to be ice with the 2D probe and shattering on the FSSP;

equation image

where the term in parenthesis is another estimate of the IWC in large ice particles. Equations (1) and (2) can also be written in terms of area A:

equation image
equation image

[7] The shattering coefficients in equations (1)()(3) were found by fitting curves to scatter plots of IWCFSSP or AFSSP against IWC or A determined from the 2D probes and TWC from the CVI. In equation (4), CA2D,CIN is found by subtracting A2D from ACIN. Because the latter is about ten times the former, errors in A2D produce minimal error in CA2D,CIN. The shattering coefficients are found by comparing time series of scaled FSSP or CIN data with the 2D or CVI probe data from all flights during a project. The coefficients are found to have a small dependence on airspeed, and obviously will change with probe calibration and mounting location. For example, CA2D,FSSP is 2.75 times larger for the Citation than for the WB57.

3. Results

[8] This section illustrates equations (1)(4) in ideal cases and with actual data. Because of its diversity, the C-F Citation flight on 9 July is used as a test case. There were two cloud layers: the lower one at −3 to −15° contained both ice and supercooled water (SLW) in various proportions, detected by the Rosemount icing detector (RICE) and the King hot wire probe. Both of these probes reported no liquid in the upper layer at −30 to −55°.

[9] In the lower layer the King probe sampled SLW with no 2D particles 68% of the time and 2D particles with no SLW 18% of the time. The mixed-phase regions contained drizzle drops and needle crystals and their aggregates. In these regions SLW dominated the TWC; the median ratio of the 2D IWC to CVI TWC was only 0.05. In the upper layer the mean ratio of calculated IWC from the 2D to the CVI IWC was 1.02 ± 0.29 and if the IWC from the FSSP, assuming natural crystals, is added, the ratio is 1.19 ± 0.34. The uncertainty in the FSSP IWC of about 2 is much larger than that for the 2D probe IWC which is constrained by the CVI IWC.

[10] Evaluation of the time series of FSSP and CIN data from the upper layer gives C2D,FSSP = 0.15, CCVI,FSSP = 0.15, CA2D,FSSP = 1.1 and CA2D,CIN = 7 averaged for all C-F days, 9 for the SLW case on 9 July. In Figure 1a the solid line shows no small natural ice in the cloud; IWCsmall = 0 g m−3, so that all of the IWC from the FSSP (equation (1)) comes from shattering. The dashed line represents equation 1 with IWCsmall = 0.02 g m−3 chosen arbitrarily, with IWCsmall dominating at low values from the 2D IWC and shattering at high ones.

Figure 1.

IWC or LWC in FSSP (small) and 2D probe (large) sizes. Equations relating the two shown. Green points are of constant ratios. (a) Synthetic data. (b) 9 July, Citation data from Crystal-FACE at T < 0°C. Red points from lower layer with SLW. Measurement uncertainties given in Section 2 cause dispersion in the IWC estimates.

[11] In the absence of shattering, the IWCsmall in equation (1) may be a function of the 2D IWC because of real physical processes. This has been reported by McFarquhar and Heymsfield [1997] (hereinafter referred to as MH97). The authors measured ice PSD in tropical outflow cirrus using 2D probe data for large particles and, when shattering was not an issue, for small particles with simultaneous collections from an impactor probe. The PSD was given in terms of temperature and total IWC. Their parameterization has been used in the 9 July case for ice-only regions. Here the inputs are IWCs from the CVI and temperature and outputs of the IWC in FSSP and 2D probe sizes (blue points). Notable here is the trend from low to high proportions of IWC in small particles with decreasing IWC, the dotted straight lines showing constant proportions.

[12] In the mixed-phase regions in this case, TWC is dominated by small liquid droplets, making it possible to substitute LWC for IWC in the first term of equation (1). The dashed line in Figure 1a is produced by adding a constant LWCsmall = 0.02 g m−3.

[13] Figure 1b shows the data from 9 July in which the black points are from measurements with no SLW and the red ones from the lower layer with SLW. The black points fall nicely along the line for C2D,FSSP = 0.15. There is no large deviation from the curve expected for appreciable IWCsmall. However, a natural process cannot be ruled out. The red points in Figure 1b show where the King probe indicated LWC > 0.05 g m−3 or the RICE signal [Heymsfield et al., 2005] indicated this amount and they lie along a roughly constant value for the FSSP IWC, independent of that from the 2D IWC. Both are markedly different from the MH97 parameterization in Figure 1a.

[14] The correlation of the FSSP IWC with that from the CVI TWC may differ from the 2D IWC because the CVI includes condensate from small and large sizes. Figure 2a represents equation (2) and is entirely comparable to Figure 1a. In equation (2), in regions dominated by SLW, TWC replaces equation (1)'s first term and the 2D IWC the second. Figure 2b shows the 9 July data on the same plot with colored points being from clouds with measurable liquid and are close to the 1:1 line. The black points from ice only, fall along the shattering line. The same coefficient, C, suggests that the responses from the 2D and CVI probes are similar in such regions.

Figure 2.

As in Figure 1 but for CVI probe.

[15] In Figure 3a assumptions from Figure 1a are used to simulate the relationship between A's from 2D and FSSP data. The MH97 parameterization predicts that as A decreases, that portion from small particles will increase. Adding to A a constant, arbitrary amount from small particles, produces a curve along the ordinate that approaches its values asymptotically.

Figure 3.

As in Figure 1 but with volume area coefficient = σ/2.

[16] Observations on 9 July reinforce the finding that the A from natural, small ice particles is added to by a constant fraction of large ice particles fractured on the FSSP inlet. Observations of A from the 2D and the FSSP in ice-only regions fall along the shattering curve, suggesting that in the cold cloud layer, natural, small ice particles contributed very little. In the SLW region the data are consistent with shattering in addition to relatively large amounts of A in small particles, here droplets.

[17] The ideal and observed relationships for A for the CIN and the FSSP are given in Figure 4. The relevant equation, combining equations (3) and (4), is given in Figure 4a. Because both the CVI and the CIN cover the whole size range, the general tendency should be similar to those for the CVI/FSSP combinations. With only small particles and no shattering the relation should be on the 1:1 line. Green points in Figure 4b are from locations with SLW only. Values for A from the CIN are parallel to, but larger, than those for the FSSP. Although some of the difference may be accounted for by icing conditions having affected some of the CIN data, the large red points are in SLW regions before CIN icing could have posed significant measurement errors. The red points are from mixed-phase regions. Because values for A in large particles are small, with a median value of only 0.0001 m−1, the points fall on top of those for liquid-only regions. For only-ice regions, without any small natural particles, and ignoring any probe measurement uncertainties, the CIN-FSSP relationship would be similar to that for the 2D probe in Figure 3b. If all of the FSSP particles were natural, the CIN value would increase to A (2D) + A · CA2D,FSSP with the CIN-FSSP on a curve shifted to the right by a factor of about 2 from that given in Figure 3b. Even if A (2D) is taken to be uncertain by a factor of 2, this suggests that CIN measurements are significantly affected by shattering.

Figure 4.

As in Figure 1 but for CIN probe.

[18] In cold ice clouds in C-F, the IWC and A from the NASA-WB57 penetrations were also significantly affected by shattering. A somewhat different shattering response would be expected because the WB57 samples at a true airspeed of 200 m/s and the Citation at about 120 m/s. Data from a WB57 flight through tops of vigorous convection at −55 to −60° are shown in Figure 5. Since the shattering coefficients are much larger for A than for IWC, only A is given. The A's from the FSSP, the CIN, and the CAS all show different, but linear, dependence on A from the 2D probe, also suggesting important, but quantitatively different shattering effects for these probes. Below A of about 0.0001 m−1, the 2D data probably reflect uncertainties attributable to the small sample volume and mismatches in sampling times of 1 Hz for FSSP and CIN and 0.2 Hz for the 2D. The sharp increase in all three numbers of about 0.005 m−1 on the 2D probe in convection is particularly interesting. From the FSSP and the CAS, the factor of five higher number of small particles relative to the numbers in the shattering regions are presumably the result of the homogeneous nucleation also noted by Heymsfield et al. [2005].

Figure 5.

Data from NASA WB57, C-F, 29 July.

4. Summary and Conclusions

[19] Data from the CRYSTAL-FACE program and from other field programs not shown here suggest that the FSSP, the CAS, a related spectrometer probe, and the Cloud Integrating Nephelometer commonly sample a combination of small, real ice particles and shattered, large ice particles. Linear empirically-derived relationships with coefficients accounting for large particle shattering have been developed, which for the C-F clouds are valid for IWC > 0.001 g/m3 and A > 0.0001 m−1 or σ > 0.1 km−1. This should provide a better understanding of how real ice particles and artifacts may be combined in the measurements from these probes although natural processes could account for the observations. Examples are given from cases with supercooled liquid water and from an anvil associated with a vigorous convective system. A more inclusive study also indicates that linear relationships between the IWC or A in large particles and those detected by the FSSP and the CAS are the normal situation. Shattering alone adds about 15% to the large IWC from the FSSP and because the large IWC is typically much in excess of the small one, it is only in relatively extreme situations that the IWC in small particles can be separated from that in large particles. In extinction the problem is even worse because the shattering signal for the FSSP and the CIN is as least as large as the extinction in large particles. Earlier studies using data from these probes to derive IWC or σ and for characterizing ice formation processes should be reevaluated, with appropriate shattering coefficients derived empirically. For these data sets it may be possible to develop algorithms distinguishing real small ice particles from shattered ones. New instruments are now becoming available by which we may improve our knowledge of how small particles contribute to IWC and σ.


[20] The author wishes to thank Charlie and Nancy Knight, Paul Field, Aaron Bansemer, Carl Schmitt and Hugh Morrison for valuable discussions and Cynthia Twohy and Darrel Baumgardner for use of data from their probes. The encouragement of the CloudSat Project Office and support from NASA are appreciated.