Current Insitu Instrumentation and Its Limitations

In-situ Measurements

The airborne platform used for in-situ measurements must be suitable. For example, in some applications a tethered balloon might be adequate and economical, whereas in others a long-range aircraft is essential. For campaigns requiring only a few instruments, a light aircraft or an unmanned airborne vehicle (UAV) might be the best option; in other projects, a large capacity aircraft is necessary to carry all of the technical gear as well as the many operators and scientists involved. Some applications require an aircraft that can fly at high, or low, altitudes.

There is a host of in-situ instruments available to measure the properties of aerosols, in particular cloud condensation nuclei (CCN), ice-forming nuclei

(IN), and cloud properties such as cloud droplets, ice particles, cloud water content, precipitation-sized particles, and special properties related to extinction and scattering of the particles themselves. Often it is necessary to measure gas phase chemical composition and the chemical particle properties to simulate the interaction of clouds with their environment properly. It is not possible to address all of the available instruments in this chapter, nor can we successfully forecast new advances in instrumentation. Thus, the reader will be referred to some general papers, and an outline of some of the available instruments will be provided.

In the 1970s, Particle Measuring System (PMS) and, specifically, Bob Knollenberg effected a new era of cloud in-situ measurements. Knollenberg (1981) describes some of these new probes; for a more recent summary of in-situ measurement techniques, see Baumgardner et al. (2002). Figure 4.1 (see next section) indicates the size ranges of some standard PMS probes for size distribution measurements. Lawson (1998, 2001, 2006) describes new instrumentation that can help image and discriminate small ice particles in the atmosphere, which is a very difficult problem at the moment (see below). Korolev et al. (1998b) introduce a new probe that can simultaneously measure cloud liquid water content (LWC) and cloud total water content (TWC). Rogers et al. (2001) and Rose et al. (2007) discuss some of the problems of measuring IN and CCN.

Table 4.1 shows some cloud microphysical parameters that are measured in all-liquid, all-ice, and mixed-phase clouds, along with their nominal sample volumes and possible accuracy. Many problems exist in creating such a sim-plifi ed overview; however, the intent is to provide useful information to the nonexpert. Hallett (2003) discusses some of the errors brought about by sampling statistics, which are not adequately addressed in Table 4.1. Some in-situ measurement parameters, which are important but not listed, include in-cloud temperature and relative humidity (or supersaturation), as well as updraft and turbulent velocities. Out-of-cloud dewpoint is also an important variable. Temperature in-cloud is an interesting case, because it is often assumed that it can be measured within 1°C, providing the probe does not become wet (Jensen and Raga 1993). Using a fast response probe, however, Haman et al. (2001) demonstrated rapid fluctuations of 1-2°C over distances of one or two centimeters, which could significantly influence in-cloud superaturation. Unfortunately, methods to measure in-cloud supersaturation are limited, an exception being the technique described by Gerber (1991). Korolev and Isaac (2006) show measurements of supersaturation within all-liquid, all-ice, and mixed-phase clouds, where the all-liquid and mixed-phase clouds are close to water saturation. There are several techniques to measure in-cloud turbulence; for a summary of such measurements made in cumulus clouds, see MacPherson and Isaac (1977) and Siebert et al. (2006).

In-situ data must be collected with instruments that perform the tasks required. Isaac et al. (2005) describe instrumentation necessary for obtaining

Table 4.1 In-situ measurements normally made in all-liquid, all-ice, and mixed-phase clouds. Nominal sample volume rate is shown at 100 m s-1; source of the nominal accuracies are indicated in the footnotes. Individual datasets should provide their own numbers. The accuracies assume that probes are properly mounted on the aircraft and calibrated. CCN and IN measurements are not necessarily done in-cloud. SS: supersaturation; N: number concentration; D: particle diameter; MVD: median volume diameter.

Table 4.1 In-situ measurements normally made in all-liquid, all-ice, and mixed-phase clouds. Nominal sample volume rate is shown at 100 m s-1; source of the nominal accuracies are indicated in the footnotes. Individual datasets should provide their own numbers. The accuracies assume that probes are properly mounted on the aircraft and calibrated. CCN and IN measurements are not necessarily done in-cloud. SS: supersaturation; N: number concentration; D: particle diameter; MVD: median volume diameter.


Size Range

Nominal sample (vol. at 100 ms"1)

Possible accuracy


All-liquid Clouds

CCN 1-2'3

0.1-1.3% SS

1 1 mur1


No known calibration standards

LWC 4'5'6

0.01-3 gm-3



Errors higher for low LWC; only accurate for MVD < 40 |im.

Cloud droplets7'8

2-50 jim

30 cm3 s_1

N: 20%, D: 1-2 jim

Accuracies depend on airspeed, size, and concentration

Large droplets9' 10'u

50-500 jim

5-15 1 s-1

N: 25%, D: 10%

Accuracies depend on airspeed and size; 50-100 |im is poorly measured for all phases

Precipitation drops7

>500 jim

200 1 s-1

N: 10%, D: 10%

All-ice Clouds


0 to -40°C, ice saturation to 20% SS water

1 1 min-1


Very difficult measurement

Ice water content13

0.01-2 gm"3



Controversy about measurements by imaging

Small ice particles14

2-100 jim

and scattering probes related to shattering off probe tips.

Parameter Size Range Nominal sample (vol. at p„ss¡h|c. accuracy Comments

Large ice9'10'11

>100 (im

50-200 1 s-1

N: 25%, D: -15%

Particle shape11'15

Requires 5-10 pixels

Automated recognition software is available but not standardized

Mixed-phase Clouds


0.01-3 g m"3



Accuracy depends on ice/liquid fraction.


0.01-3 g m"3



Ice water content13

0.01-2 g m"3



Small size distribution16

2-100 |im


No validated technique for phase segregation

Large sizes9'10' n>16

>100 (im

5-200 1 s-1

N: 25%, D: 15%

Need good liquid/ice discrimination software

Particle shape11'15

Requires 5-10 pixels

Can be accomplished by assuming all circular 2-D images represent liquid drops


•Lance et al. (2006) 2Rose et al. (2008) 3Roberts et al. (2006) 4Biter et al. (1987)

5Korolev et al. (1998a) 6Strapp et al. (2003) 7Knollenberg (1981) 8Baumgardner & Korolev (1997)

"Strapp et al. (2001) 10Gayet et al. (1993) "Korolev et al. (1998b) 12Rogers et al. (2001)

13Korolev et al. (2008) 14Heymsfield (2007) 15Korolev et al. (2000) 16Cober et al. (2001b)

data for aircraft icing certification tests, and much of the information in that paper is relevant here. When selecting and evaluating datasets, it is necessary to know (a) the parameters that are required, (b) the accuracy required for those measurements, (c) the range of conditions over which the measurements must be made (e.g., temperature, altitude, LWC, drop size, length scale), and (d) the types of clouds to be examined (convective, stratiform). In general, the dataset must be sufficiently large to obtain a representative sample. Generally, a few case studies are not good enough to characterize clouds for climate models or to evaluate remote-sensing techniques, although they may be exceptionally useful for examining physical processes.

It should be stressed that it is very difficult to measure IN because they can activate through many different mechanisms (e.g., deposition, contact, immersion, condensation freezing; cf. Kreidenweis et al., this volume). The difficulties are compounded because it is necessary to obtain vertical profiles of IN, thus requiring airborne instrumentation. The uncertainties in measurements of small ice particles are also quite large. Combined, these two problems represent a large uncertainty in characterizing ice in clouds and in understanding ice formation mechanisms in the atmosphere, which are a blend of primary nucleation through IN and secondary processes through ice multiplication. This area definitely requires further instrument development.

It is often assumed that by measuring the particle size spectrum, the liquid, ice, or TWC can be determined by a simple integration of the spectrum. However, significant sizing errors exist for any measurement, and when the diameters within a size bin of a probe are cubed to calculate volume or mass, the resulting error in the integrated liquid or ice water content becomes large (Baumgardner 1983). Thus, it is recommended that a probe specifically designed for direct measurement of liquid, ice, or TWC, such as a hot wire probe or counterflow virtual impactor, be selected. An icing rate indicator is also very useful to determine whether supercooled liquid was encountered during data collection. If the ramp voltage is measured during the tests, then a rough estimate of LWC can be obtained to provide a check with other instrumentation (Mazin et al. 2001; Cober et al. 2001a).

Imaging probes that measure the shape of the particles are very useful and, in some cases, essential. Many clouds contain both ice and liquid particles (mixed-phase clouds). Cober et al. (2001b) and Korolev et al. (2003) have documented that such clouds occur frequently and discuss some of the associated measurement issues. Summarizing supercooled in-cloud measurements, Isaac et al. (2001) report that 25% of maritime and 49% of continental clouds were characterized as mixed phase. If it is not possible to assess whether ice crystals are present, these particles can be misinterpreted as supercooled large drops, thereby greatly biasing liquid cloud median volume diameter estimates. The hot-wire LWC probes also measure a fraction of the ice particles present, thus overestimating the LWC if the ice mass concentration is high.

There is a need to select instruments that have been evaluated and reported in the open literature. Manufacturers are continuously producing better instruments, but they often do not operate to the specifications provided. Users should look for comparison tests of instrumentation performed in icing wind tunnels (e.g., Strapp et al. 2003). These tests can give evaluations of the effectiveness of the probes and their associated accuracies and limitations over a wide range of environmental conditions. All instrumentation have some weaknesses and strengths. It is important to know what these are before selecting sensors or samplers for a particular application. Instruments and their resulting data should be selected for use only if their accuracies have been documented and demonstrated.

Other concerns that must be considered are:

1. Are the probes adequately de-iced for the temperature and liquid water ranges expected? It is common for newly designed probes not to have adequate de-icing heaters.

2. Will imaging probes and droplet spectrometers fog during rapid descents? Fogging can also occur in climbs into temperature inversions. Fogging and icing signatures (e.g., Brenguier et al. 1993) must be known during the data analysis, and corrections need to be applied.

3. Some probes work well at low aircraft speeds (e.g., typical turboprop speeds) but the electronics are not fast enough for proper operation at high speeds (typical jet aircraft speeds).

Remote-sensing Instruments

The launch of CloudSat (Stephens et al. 2002) and CALIPSO (Winker et al. 2007) in 2006 marked the beginning of a new era in cloud and aerosol spaceborne remote sensing with the start of continuous radar and lidar observations. At the same time, new polarimetric and hyperspectral imagers already in orbit or under preparation provided an improved level of insight into clouds, aerosols, and their impact on weather and climate.

Ground-based remote-sensing observations are often understated for their role in global Earth system remote sensing, yet they excel in temporal sampling resolution, accuracy, and continuity, and provide data that cannot be obtained from satellites (e.g., cloud and aerosol surface radiative forcing). The major drawback of ground-based observations is their limited spatial coverage. Thus, the establishment of networks over the last decade represents an important achievement (Kinne et al., this volume).

Aircraft observations are an indispensable tool for observing clouds. Most satellite applications are tested on aircraft before they are deployed in space, and field measurements are conducted to validate satellite measurements and algorithms on a regular basis; intensive, vertically resolved aircraft observations link the globally extensive ground and spaceborne observations. UAVs are becoming an increasingly important tool for atmospheric research.

Most atmospheric remote sensing is based on the measurement of electromagnetic radiation that has retained information about its interaction with atmospheric constituents through scattering, absorption, and emission. An exception is sodar (sonic detection and ranging), which uses reflection of acoustic waves on atmospheric boundaries and is an important tool for monitoring wind speed and atmospheric stability.

The wavelengths used for atmospheric remote sensing lie within three spectral regions ("windows"), which are characterized by a low opacity of various atmospheric gases within the Earth's atmosphere. This is illustrated in Figure 4.1, where the clear-sky atmospheric opacity is plotted as a function of wavelength and frequency. Also plotted are the normalized Planck curves of solar and terrestrial emission. The first range is called "shortwave" or "solar" window. It is located around the maximum of solar irradiance (490 nm wavelength), originating from blackbody emission of the Sun's photosphere with a temperature of about 5800 K. It comprises the near-ultraviolet, visible, and near-infrared wavelengths and is bounded by stratospheric ozone absorption and molecular scattering at the short wavelength end. The second range ("longwave," "thermal infrared," or "terrestrial" window) is centered at a wavelength of 10 pm (corresponding to the peak in Earth's blackbody emission with a

Frequency lOPHz IPHz lOOTHz lOTHz ITHz lOOGHz lOGHz 1GHz lOOMHz lOMHz 1MHz





_Terrestrial emission

_Terrestrial emission

lOnm lOOnm 1pm 1Opm 1OOpm 1mm lOmm lOOmm 1m 1Om lOOm 1km

Wavelength microwaves_radio waves_

microwave . radiometers

¡]d¡rVIS/NIR/IR imagers/sounders:! W Ka'XC S radar remotesensinS

aerosol particles cloud droplets drizzle rain


10nm 1o0nm 1pm 10pm 100pm 1mm 10mm 100mm 1m 10m 100m 1km


Figure 4.1 Top: atmospheric opacity with normalized solar and terrestrial emission. Bottom: spectral ranges of remote-sensing techniques and size ranges for some in-situ instruments.


surface temperature of 290 K). The transition between solar and terrestrial range is often defined at about 4 ^m. The terrestrial range extends to 100 ^m and includes various carbon dioxide and water vapor absorption features. The third range ("radio window," 1 cm to 30 m wavelength) is used for microwave radiometers and radars. Some remote-sensing techniques and their typical wavelength ranges are shown in Figure 4.1.

If the measured radiation originates from a natural source (solar or terrestrial emission), the technique is called passive remote sensing; active remote sensing refers to those methods where the radiation originates from the instrument itself. Some examples of passive remote sensing include visible and infrared imagery as well as microwave radiometers. Active techniques include radars, lidars, and ceilometers. Remote sensing can be further distinguished by viewing geometry (nadir or limb) and by viewing mode (profilers, imagers, and scanners). Imagers cover large areas with sometimes high horizontal resolution, but they contain only limited information about the vertical structure of the atmosphere. In addition, sensitivity to thin cirrus and aerosols is generally low, especially over bright surfaces. Profilers view the atmosphere on slant paths and thus have high vertical, but very limited horizontal, resolution. They are very sensitive to cirrus, aerosols, and gases because of (a) the large photon path length through the atmosphere and (b) the low surface contributions to the signal. However, the large slant path contributes to strong attenuation and thus the signal saturates at relatively low vertical optical thickness. Some techniques combine nadir and limb viewing using multiple detectors or along/cross-track scanning modes. Spaceborne active instruments use nadir-viewing geometry. Ground-based instruments can be zenith viewing (e.g., lidars, sky imagers), scanning (e.g., radars), or tracking (sun photometers).

Spectrally resolved measurements enable the retrieval of vertical atmospheric structure even for nadir-viewing geometries (AIRS, Chahine et al. 2006; AMSU, Mo 1996), and allow the separation of clouds, aerosols, water vapor, and other gases by means of their spectral signature. On the other hand, to determine the energy budget of clouds and aerosols from space, broadband albedo and emission measurements are required for the solar and terrestrial spectral ranges, respectively.

The polarization state of radiation can be exploited for both passive and active remote sensing. It is useful to detect particle shape and orientation as well as to separate the contributions to top-of-atmosphere radiance from atmosphere and surface (Herman et al. 1997).

Table 4.2 lists cloud and aerosol parameters accessible with remote sensing. The accuracy and sensitivity of a technique to a particular parameter depends on the underlying physics. For example, radars have a high sensitivity to precipitation and can determine its geometrical distribution. This is because the radar reflectivity, given by Rayleigh back-scattering of the emitted signal, is proportional to the sixth moment of the drop size distribution, (D6), and is therefore heavily weighted by large drops. Microwave radiometers, in contrast,

Table 4.2 Cloud and aerosol parameters, and a technique for retrieval.

et me o e

Cloud top/base height • lidar/ceilometer, NIR, stereo height

Plume/aerosol height • lidar

Cloud cover • VIS/IR imagers


Vertical structure, thickness • radar

Vertical structure, thickness • lidar

Optical thickness • VIS imagery

Optical thickness, Angstrom parameter • sun photometers, imagers, lidar


Effective radius • VIS/NIR imagery

Single-scattering albedo • scanning sun photometers

Cloud albedo / radiative forcing • spectral/broadband imagers

Aerosol direct effect / radiative forcing • spectral/broadband imagers

Ice/liquid water path • microwave radiometer/imager

Size distribution • sun photometers, polarimetry

s ic si £

Crystal shape • polarimetry

Aerosol type (dust, sea salt, ...) • sun photometers, polarimetry

p rop ic


Precipitation rate; fall speed • microwave radiometer, Doppler radar

Thermodynamic phase • VIS/NIR imagery

Mixture state (internal/external) • currently no remote sensing technique

provide a direct measurement of the cloud brightness temperature and emis-sivity, which can be related to the third moment of the drop size distribution, (D3), and thus to column-integrated water content (liquid water path, LWP) and precipitation rate. However, little information about the spatial structure can be retrieved. Visible or infrared imagery are heavily weighted by the properties near cloud top. This is the method of choice for determining the cloud radiative forcing because optical thickness, r, related to (D2), as well as cloud cover and cloud-top effective drop radius, re, can be retrieved directly. Through the simple relationship, LWP = 2/3 pr x re (where p represents water density), LWP can be inferred as well, assuming that the effective radius at cloud top is representative for the whole cloud, which is often not the case because of the vertical cloud structure. Conversely, a cloud column-averaged effective radius can be obtained when optical thickness retrievals are combined with LWP retrievals from a microwave radiometer.

Cloud drop number concentration is not easily accessible through remote sensing; even if a number of moments such as (D3) and (D2) are known, assumptions about the shape of the drop size distribution are needed to derive its integral, the number concentration. A variety of algorithms exists for retrieving cloud thermodynamic phase (Chylek et al. 2006), but this is particularly difficult when both liquid and ice phase occur. Other parameters that are not well constrained from remote sensing include cloud and aerosol absorption and heating rates. This difficulty is fundamental: spaceborne sensors measure flux on top of the atmosphere; the ground-based counterparts obtain it at the surface; and both are needed simultaneously to determine atmospheric absorption.

Spaceborne aerosol retrievals from imagery suffer from the problem that their radiative signature does not provide sufficient contrast to distinguish it from variability in surface reflectance. In addition, there is currently no method for retrieving aerosol optical thickness in the presence of clouds, because the cloud signal dominates the reflected radiation. Aerosol single-scattering albedo retrievals are even more difficult to obtain, or are possible only for special cases (e.g., from sun glint over water; Kaufman et al. 2002). Ground-based sun photometers are the most reliable source of aerosol optical thickness, but do not work if clouds block the Sun. A promising technique for space- and airborne retrievals is polarimetry. Spaceborne and ground-based lidars provide information about the layering of the aerosols and thin clouds from their back-scatter signal; some systems measure extinction profi les directly, and hence optical thickness (column-integrated extinction). From the linear depolarization ratio, information about the particle shape and thermodynamic phase can be deduced: Whereas spherical particles do not change the polarization of the incident radiation, nonspherical particles (ice crystals and some aerosol types) modify the state of polarization. Some information about aerosol particle size can be retrieved from the Angstrom parameter, which is a measure of the wavelength dependence of the aerosol optical thickness. It could also be used to distinguish various aerosol types from clouds (whose optical thickness is wavelength-independent) in future hyperspectral satellite observations.

In-situ Instrumentation: Problems and Limitations Mounting Problems and Flow around and within Probes

A common error when making in-flight observations is the selection of poor mounting locations for probes. All mounting locations should have an engineering assessment of their suitability. This assessment would preferably be performed with model or wind tunnel simulations. Probes should be in the free stream, outside the boundary layer of the aircraft, and not affected by engines, propeller wash, or other probes. Commonly selected poor probe mounting locations are window mounts, top of the fuselage near the cockpit, and wing tips. Good mounting locations can be found underneath the wing and sometimes the belly of the aircraft. King (1984) shows how 100 pm drops can be affected as they flow over the top of a fuselage (Figure 4.2). In this case, close to the skin, the top of the fuselage becomes a shadow zone for drops of this size. Further away from the skin there is a concentration of particle trajectories.

Figure 4.3 shows the top of the fuselage errors for a mounting location on the Canadian Convair 580 during tests of the Nevzorov probe (Isaac et al. 2006). As

Figure 4.2 Airflow (left to right) over the top of a fuselage showing the streamlines (top) and the trajectories of 100 ^m drops traveling over the top of an F27 fuselage traveling at 90 m s 1 (bottom) as described by King (1984).

Figure 4.2 Airflow (left to right) over the top of a fuselage showing the streamlines (top) and the trajectories of 100 ^m drops traveling over the top of an F27 fuselage traveling at 90 m s 1 (bottom) as described by King (1984).

can be seen, there is a shadow zone and the zone of concentration is also delineated for drops of a specified diameter. For the Convair, the shadow zone maximum occurs at a droplet diameter 160 ^m. Using more sophisticated models, Twohy and Rogers (1993) have shown that much larger particles, with greater inertia, will cross the streamlines and not be shadowed. It should be noted that the behavior of ice crystals is substantially different from that of spherical water drops. King (1985, 1986) discusses this in more detail. Mounting location errors tend to be highest for the intermediate size particle (100-500 ^m).

40 60 80 100 120 Droplet diameter (|jm)

140 160

Figure 4.3 Concentration enhancement calculated based on the King (1984) droplet trajectory and airfl ow model plotted as a function of distance from the fuselage surface of the Canadian Convair 580 aircraft and the droplet diameter (Isaac et al. 2006). Airspeed was assumed to be 100 m s '. Particles greater than about 120 ^m will be in a shadow zone (shaded area). For the Convair, the shadow zone maximum occurs at a droplet diameter 160 ^m. Using more sophisticated models, Twohy and Rogers (1993) have shown that much larger particles, with greater inertia, will cross the streamlines and not be shadowed.

Probes themselves can create flow field problems (e.g., Norment 1988). For example, the Cloud Particle Imager (Lawson et al. 1998) has a long flow tube where particles might be disturbed or broken up before sampling. Korolev and Isaac (2005) describe how optical imaging probes can cause the shattering of particles off the probe inlets (see Figure 4.4). Proper software can eliminate many of these shattering problems but users should be aware that it is necessary to make such corrections (e.g., Field et al. 2006).

Defining In-situ Probe Sample Volumes

For some probes, it is necessary to calculate the sample volumes accurately. This can be very difficult or relatively simple to do. For the PMS OAP probes, one must recognize that the sample volume can depend on particle size and airspeed in a significant manner (Baumgardner and Korolev 1997). Korolev et al. (1998a) describe how corrections might be made for particle size. Probes (e.g., the PMS FSSP) cannot always count fast enough when particles arrive too quickly. The electronics produce a dead time, which needs to be considered when calculating the sample volume (Brenguier et al. 1994).

In-situ Probe Size Ranges

It is very important to measure particles over the complete size range of interest. This is a problem that can be overlooked. Sometimes, for example, PMS OAP 2-DC probes have been used to determine radar reflectivities. However, it

ffi E

Figure 4.4 Number of isolated images per frame for the PMS 2-DP probe and the SPEC HVPS probe (from Korolev and Isaac, 2005). The value of D was calculated as a size that 97% of all the shattering events for a specified number of isolated images per frame are located at D > D____.

Figure 4.4 Number of isolated images per frame for the PMS 2-DP probe and the SPEC HVPS probe (from Korolev and Isaac, 2005). The value of D was calculated as a size that 97% of all the shattering events for a specified number of isolated images per frame are located at D > D____.

is well known that a few large particles can affect the reflectivity because it depends on number concentration multiplied by the diameter to the sixth power. Figure 4.5 shows composite spectra from field projects conducted in southern Ontario, Canada (Isaac et al. 2002). The top, middle, and bottom graphs show the number concentration, mass concentration, and reflectivity concentration, respectively, as a function of TWC. The panels on the left side show the liquid water spectra; those on the right side show the ice water spectra. The basic

Continental CFDE lll and AIRS

Changes in the liquid phase spectra with TWC

Continental CFDE lll and AIRS

Changes in the liquid phase spectra with TWC

Points = 462

.. 0.005 <=

TWC < 0.050


— 0.050 <=

TWC < 0.100


0.100 <=

TWC < 0.150


0.150 <=

TWC < 0.200


— 0.200 <=

TWC < 0.250


— 0.250 <=

TWC < 0.300


0.300 <=

TWC < 0.350


- - 0.350 <=

TWC < 0.900

102 103 Diameter (microns)

102 103 Diameter (microns)

102 103 Diameter (microns)

Figure 4.5 Averaged spectra for CFDE III and AIRS sorted by all-liquid and glaciated 30-s values. The "number" of 30-s spectra used for each TWC are shown.

102 103 Diameter (microns)

Figure 4.5 Averaged spectra for CFDE III and AIRS sorted by all-liquid and glaciated 30-s values. The "number" of 30-s spectra used for each TWC are shown.

probes used and their nominal size ranges were the PMS FSSP standard range (3-45 pm), the PMS FSSP extended range (5-95 pm), the PMS 2-DC (25-800 pm), and the PMS 2-DP (200-6400 pm). For the 2-DC probe, only particles larger than 100 pm were counted, because of uncertainties in counting and sizing particles from 25-100 pm with this probe. This creates a sizing gap between the FSSP and 2-DC probes, which is visible on many of the plots on the left panel. The analysis techniques, including the phase discrimination method, have been described in detail by Cober et al. (2001b).

Continental CFDE lll and AIRS

Changes in the glaciated phase spectra with TWC

Points = 317

- - 0.005 <=

TWC < 0.050


-0.050 <=

TWC < 0.100


-0.100 <=

TWC < 0.150


-0.150 <=

TWC < 0.200


-0.200 <=

TWC < 0.250


-0.250 <=

TWC < 0.900

102 103 Diameter (microns)

102 103 Diameter (microns)

102 103 Diameter (microns)

Figure 4.5 (continued)

Figure 4.5 shows that the reflectivity values continue to increase as one goes to increasingly larger sizes. The roll-off at the largest sizes may be a function of the software, and for the ice particle spectra, this may be an indication that the largest sizes were not measured. It should be emphasized that software techniques allow one to measure beyond the maximum size range of the probe, which is 800 pm and 600 pm in the case of PMS 2-DC and 2-DP probes, respectively (see below). However, to do that, one must assume either spherical or symmetrical particles. Obviously, these assumptions can be wrong in the case of ice particles.

In-situ Probe Analysis Software

The analysis of cloud microphysical data often requires sophisticated software. Performing an automated analysis using software is quite difficult. Kingsmill et al. (2004) describe some of the problems. Isaac et al. (2005) provide an example of how different assumptions in the software can lead to different results, as described below.

For a freezing rain case with data collected from the Canadian Freezing Drizzle Experiment III, Table 4.3 illustrates difficulties that can be encountered through the analysis of PMS 2-D imagery (Knollenberg 1981). The analysis was performed using the Cober et al. (2001b) method and a second software package developed through Environment Canada called "2-D Analyzer," which uses the techniques described by Heymsfield and Parrish (1978). Particle diameter is determined from the 2-D imagery by either measuring the dimension in the X direction (direction of flight) or the Y dimension. Diameter can also be computed from the total particle area assuming that the particle is a sphere. Extended area (EXA) is determined using the geometrical reconstruction, after Heymsfield and Parrish (1978), which basically looks at a portion of the drop that is imaged and estimates the diameter of the whole particle. The center-in (CIN) technique must have the center of the particle imaged. The double end element (DEE) technique restricts the analysis to particles that are completely imaged and do not shadow the end elements of the diode array.

Table 4.3 Comparison of various methods of determining LWC during a freezing rain encounter 5.5 min long during CFDE III. H&P: Heymsfield and Parrish (1978).

Analysis technique

Circular particles

Irregular particles

LWC g m 3 LWC g m 3 (125-6400 pm) (125-2000 pm)






2-D Analyzer






2-D Analyzer






(0.205 with

(0.205 with









2-D Analyzer

For both the EXA and CIN techniques, particle geometry must be assumed to determine the probe sample volume accurately. These techniques work well for circular drops but do not apply for irregularly shaped ice particles. The case described in Table 4.3 contained drops between 125-2000 pm in diameter, and ice crystals up to 5000 pm in diameter. The majority of the hydrometeors observed were circular in shape (i.e., drops), and the drop median volume diameter was between 800-1000 pm. A small portion of the mass was in the small droplet size range that would be measured by the PMS FSSP probe (< 100 pm). The 2-D Analyser software provides similar LWC values between the EXA, CIN, and DEE techniques, which should be the case when the majority of the particles images are circular in shape. However, for the range 125-6400 pm, there is a significant disagreement with the LWC from the Cober software. This results from the fact that the 2-D Analyser is interpreting ice crystals with sizes from 2000-6000 pm as drops and computing their associated LWC. The Cober software more accurately segregates the circles (drops) from the non-circles (ice crystals) and hence avoids this problem. When the two software programs are compared over the range 125-2000 pm, the LWC values agree within 10%. This demonstrates that the application of advanced software analysis techniques to determine LWC can be quite erroneous if used blindly. It also provides an example of why 1-D measurements of hydrometeors should not be used to compute LWC, since 1-D instruments cannot separate drops from ice crystals in a mixed-phase environment.

This analysis shows that substantial differences can be obtained when using different analysis techniques. It emphasizes the point that software should be fully understood and used with caution.

In-situ Probe Calibrations

Where possible, calibrations of all probes should be done before and after a field campaign. Many probes can be calibrated in wind tunnels when they are accessible. Strapp and Schemenauer (1982) provided a good illustration of how problems unknown to their users can be detected with tunnel calibrations. They examined 14 Johnson-Williams (J-W) cloud LWC probes with 23 sensor heads from ten research organizations in the National Research Council of Canada's icing tunnel and found:

.. .six of the 14 systems had at least one sensor head with a nonfunctional shell or strut heater on arrival, presumably unknown to its owner. This defect can cause erroneous data at below freezing and above freezing temperatures as observed during these tests. Three systems displayed a large difference in measurement when changing sensor heads. Three more had at least one probe with a strong airspeed dependence. Based on these problems alone, nine of the 14 systems could provide improper measurements if the most unfortunate set of sensor heads were used (Strapp and Schemenauer 1982, p. 106).

More recent tests using the NASA Icing Research Tunnel have been performed by Strapp et al. (2003). Figure 4.6 shows how the King LWC probe compared with the tunnel reference LWC and how the response of this probe rolls off as the median size of the droplets get larger. Beyond median volume diameters of 40 pm, the error grows to over a 20% underestimation. Strapp et al. (2003) also show how optical array probes can measure different concentrations of large drops within the tunnel conditions (Figure 4.7).

Wind tunnels are expensive to operate and are not always available for instrument calibration. It is easier and cheaper to run instrument and data simulators. For example, this could mean putting glass beads or reticles of a known size through the sample area of an optical probe, preferably at speeds close to those experienced during flight. Dye and Baumgardner (1984) describe the calibration of the PMS FSSP using glass beads. Probes should be calibrated

LWC SWEEPS 12.9-17.6 pm

0 0

Post a comment