During the Queensland Cloud Seeding Research Program, the “CP2” polarimetric radar parameter differential radar reflectivity Zdr was used to examine the raindrop size evolution in both maritime and continental clouds. The focus of this paper is to examine the natural variability of the drop size distribution. The primary finding is that there are two basic raindrop size evolutions, one associated with continental air masses characterized by relatively high aerosol concentrations and long air trajectories over land and the other associated with maritime air masses with lower aerosol concentrations. The size evolution difference is during the growth stage of the radar echoes. The differential radar reflectivity in the growing continental clouds is dominated by large raindrops, whereas in the maritime clouds differential reflectivity is dominated by small raindrops and drizzle. The drop size evolution in many of the maritime air masses was very similar to those observed in the maritime air of the Caribbean Sea observed with the NCAR S-band polarimetric radar (S-Pol) during the Rain in Cumulus over the Ocean (RICO) experiment. Because the tops of the Queensland continental clouds ascended almost 2 times as fast as the maritime ones in their growth stage, both dynamical and aerosol factors may be important for the systematic difference in drop size evolution. Recommendations are advanced for future field programs to understand better the causes for the observed variability in drop size evolution. Also, considering the natural variability in drop size evolution, comments are provided on conducting and evaluating cloud seeding experiments.
The Queensland Cloud Seeding Research Program (QCSRP) provided for the first time the opportunity to investigate the use of polarimetric radar to observe changes in raindrop size distributions that might be the result of cloud seeding.1 The details of the QCSRP are discussed in Tessendorf et al. (2010). The Australian Bureau of Meteorology recently obtained the “CP2” dual-wavelength (S and X band), dual-polarization radar from the National Center for Atmospheric Research (NCAR). Characteristics of the radar are presented in Table 1 and more extensively in Keenan et al. (2007).
Polarimetric radar measures the horizontal and vertical radar reflectivity factors (Zh and Zυ),2 which offers the opportunity to monitor aspects of the raindrop size distribution utilizing the differential reflectivity Zdr, which is the ratio between Zh and Zυ.3 Differential reflectivity Zdr provides a measure of the flattening of the raindrops and therefore an estimate of raindrop size because the flatness and thus Zdr increase with the size of the raindrops. For example, a monodispersed size distribution of raindrops that are 3.6 mm in diameter has a Zdr of 2 dB, and 5.8-mm-diameter drops correspond to 4 dB (Knight et al. 2002). A number of studies have examined the use of Zdr to estimate raindrop size distributions (Bringi et al. 1986; Wakimoto and Bringi 1988; Brandes et al. 2004). This technique was used by Knight et al. (2008) to study the evolution of raindrop size distributions in trade wind cumulus in the northeastern Caribbean Sea during the Rain in Cumulus over the Ocean (RICO) experiment (Rauber et al. 2007).
This paper does not investigate effects seeding may have had on the raindrop size distribution; rather, it investigates the time evolution of Zdr in convective clouds in Queensland and the variability in this evolution from cloud to cloud on the same day and among days. Cursory examination of the seeded and unseeded cases indicated that the variability in Zdr was large and that understanding this variability was necessary before examining any effects on raindrop size distribution that could be attributed to cloud seeding. Because of the close proximity of the experimental area to the Pacific Ocean, convective storms occurred in both maritime-influenced and continentally influenced air masses. It has long been recognized that the airmass type affects the aerosol and cloud condensation nuclei (CCN) concentrations (e.g., Twomey 1977) and thus probably warm rain formation and drop size distributions (e.g., Hudson and Yum 2001).
Radar data-quality issues are very important when utilizing Zdr data; Zdr can be compromised by a variety of causes, including ground-clutter return, different radar beam patterns, a slight offset in the radar range between the horizontal and vertical channels in regions of strong Z gradients in range from the radar, radar return from insects, and radar return from Bragg scattering. Possible contamination by insects will be discussed in section 3. The data used in this study are discussed in section 2, and the analysis procedures are discussed in section 3. Section 4a compares the evolution of Z and Zdr from the small, maritime clouds in RICO with those from the QCSRP. Sections 4b and 4c examine the variability in the time evolution of Z versus Zdr among storms and among days in QCSRP. Section 5 addresses the interpretation of the results, and section 6 is a summary that includes recommendations for a future field campaign.
Figure 1 shows the location of the CP2 radar relative to the surrounding terrain. It also shows the initial locations of the 45 storm cells used for the analysis in section 4b and presented in Table 2. The radar was located atop a hill, which resulted in considerable antenna sidelobe ground clutter from lower-lying areas; in addition, there was some low-level beam blocking from higher hills to the south. The ground clutter was mitigated in real time by using the NCAR fuzzy-logic algorithm called the clutter mitigation decision filter (Hubbert et al. 2009). Insects produce a strong Zdr signal; thus, contamination of radar return from clouds and precipitation by insects is a possibility. Insects have Zdr values that range mostly between 2 and 10 dB (Mueller and Larkin 1985; Achtemeier 1991; Zrnic and Ryzhkov 1998) and have reflectivities of typically less than 10 dBZ but can range to at least 30 dBZ in convergence lines (Wilson et al. 1994; Russell and Wilson 1997). This topic will be discussed further in section 3.
An azimuth sector varying in width between 60° and 120° was typically chosen for radar scanning where convective clouds were occurring. The procedure was to scan the radar in plan position indicator mode through as many elevation angles as could be acquired in about 3 min, with scans extending from a lowest elevation angle of about 0.5° to an elevation angle that topped the clouds. This configuration typically provided a vertical resolution of <1 km throughout the cloud depth.
A research aircraft was used for measuring trace gases, aerosols, and cloud microphysical properties. Aerosol concentrations were obtained at cloud base for many of the seeded clouds and a few of the unseeded clouds analyzed in this study. The Passive Cavity Aerosol Spectrometer Probe (PCASP4) was used to measure aerosols in the size range of 0.1–3.0 μm, which are common sizes for particles that serve as CCN. The PCASP concentrations varied from clean (e.g., 100 cm−3) to more polluted (e.g., 1500 cm−3), representing a range of airmass conditions (Tessendorf et al. 2011). The lowest values were found in the more maritime-influenced air masses, typically with easterly low-level winds coming from over the ocean.
3. Analysis procedure
The QCSRP convective clouds selected for analysis formed alone and not as part of a cluster. They initially appeared on radar as single cells. Two types of comparison analyses were conducted, each with a slightly different philosophy. First, comparisons were made of the time evolution of Z versus Zdr between RICO and QCSRP clouds (see section 4a). Clouds selected for analysis in the RICO study (Knight et al. 2008) became visible on radar and developed to at least 20 dBZ (usually >30 dBZ) between about 15 and 60 km as seen with the S-band dual-polarization Doppler radar (S-Pol). Isolated cells were preferred, but distinctly individual cells within clusters were also used, and there appeared to be no difference between the two types of cell selections. In both data samples, the reflectivities had a threshold applied at 0 dBZ.5 A total of 190 clouds were selected for RICO, and 30 were selected for Queensland.
The second analysis method compared the time evolution of Z versus Zdr among nearby storms and among days in Queensland. This comparison was initially made without consideration to the type of air mass (see sections 4b and 4c). The 45 QCSRP cells6 examined in sections 4b and 4c were a mixture of seeded cells (16 cases7) and other single cells (29 cases) for which the radar scanning captured at least the growth stage. Once a cell was selected, a polygon was drawn about it at each radar-scanned elevation angle throughout its evolution8 (Fig. 2). The AZ and AZdr values were computed within the polygon [see Eqs. (1) and (2) below]. As in Knight et al. (2008), A is used to signify that Z and Zdr are single numbers calculated over the area of the cell. That is, Z values for all range gates9 within the cell are summed and the average is obtained, and in a similar way for Zυ.
These values were computed as follows:
where Zh is in millimeters to the sixth power per meter cubed. The summation is over all range gates within the polygon for a sweep at constant elevation angle, when Zh ≥ 10 dBZ and rhohυ ≥ 0.95 (see below). Here N is the number of sample volumes that meet these criteria (N varied roughly between 25 and 2500):
where Zυ is in millimeters to the sixth power per meter cubed and all else is as defined in Eq. (1).
These thresholds for acceptable data are strict to help to ensure the quality of the Zdr values. The polarimetric value rhohυ is the correlation coefficient between Zh and Zυ in a sample volume (~ 100 pulses). As mentioned above, there are several factors that can contaminate Zdr, which include strong reflectivity gradients, ground clutter, and insects. Knight et al. (2002) suggested that insects might influence AZdr during the early phases of cloud growth when discussing results of AZ versus AZdr from Florida, and Wakimoto et al. (2004) showed excellent examples of insects being swept up from the boundary layer into clouds. From J. Vivekanandan (2010, personal communication), it is reported that the NCAR particle identification software uses a rhohυ of <0.90 as one of the member functions for identifying insects. Inspection of rhohυ values with CP2 in Queensland typically showed that, in the clear air, rhohυ values were considerably less than 0.90 whereas in rainy regions rhohυ was >0.90. Thus much of the insect contamination will be eliminated by removing echo where rhohυ < 0.95 and Z < 10 dBZ; this assumes that insects in the updrafts and clouds have correlation characteristics that are similar to those that they have in clear-air conditions.
Figure 3a is an example of a time–height profile of AZ for cell 9 (C09) on 2 February 2008. The individual values are the AZ value for each polygon at the selected radar elevation angle and time for which the polygon was drawn. The registered height is the midbeam height at the range of the polygon center. The registered time is the start of the elevation sweep upon which the polygon is entered; a sweep took ~ 15 s. Figure 3b shows the analyzed field of AZ from Fig. 3a, overlaid with the AZdr analyzed field. This was the common method used during this study for examining the evolution of AZ and AZdr. A feature that was often observed was high AZdr located near the axis of maximum mean reflectivity (see Fig. 3b).
4. Zdr evolution
a. Comparison of RICO and QCSRP
Values of AZ and AZdr from the small, maritime cumulus clouds in RICO are compared with those from QCSRP in Fig. 4. Qualitative observations at the radar (S-Pol) indicated that coalescence was occurring in the RICO clouds when they were about 2 km deep. This was presumably due to relatively low CCN concentrations (Hudson et al. 2009) leading to larger cloud droplets and a more rapid onset of coalescence than can occur in continental clouds. It is well known that continental aerosols and CCN populations often give rise to higher cloud droplet concentrations and smaller cloud droplet sizes. Differences in early cloud dynamics and thermodynamics are important additional factors that can affect precipitation formation. Nonetheless, in continental cumulus clouds one may anticipate that the earliest formation of large drops is in low concentrations as part of a bimodal drop size distribution, so that they dominate the radar reflectivity. This might result from giant or ultragiant CCN, or just from stochastic collection without the large CCN. For continental clouds then, the earliest raindrops may be present at a low concentration but will produce an observable radar reflectivity because of raindrop size while most of the cloud droplets remain very small and are mostly undetected by the S-band radar. On radar, this is identified as high Zdr values coupled with relatively low reflectivity factor. This was identified in continental cumulus by Illingworth (1988), and this characteristic of early precipitation had been anticipated using simple cloud modeling by Johnson (1982) in a study of possible effects of ultragiant aerosols. If ultragiant CCN are present in the continental clouds they may also be in maritime clouds, producing as many larger drops as in the continental case (if present in similar concentrations), but in the maritime case the larger drops may not dominate the radar reflectivity that is influenced by the many smaller rain (and drizzle-sized) drops that are also growing by coalescence. It is important to note, however, that there were no direct measurements of giant or ultragiant CCN in the Queensland data. The aerosol measurements used to characterize the air masses in this study are solely of small CCN that determine the total cloud droplet concentration.
We examined the relationship between Z and Zdr for growing isolated cumulus in QCSRP clouds and compared it with similar RICO clouds using the total-sweep (area) values of Zdr and Z through individual clouds. In Fig. 4, the red dots are data in the growing stage of clouds and the blue dots are the maximum and dissipating stages based on the Z history at 1.5 km. Of the two reference curves, the left one shows the Z–Zdr relationship for a standard Marshall–Palmer raindrop size distribution,10 and the right one represents one drop per cubic meter of the size that produces the indicated Z and corresponding Zdr. The right curve approximates a very bimodal distribution of drop sizes with large drops and the smaller size peak contributing negligibly to the reflectivity. Note that, for points to the right of the right-hand curve, AZ must be dominated by large drops at very low concentrations (less than 1 m−3). This is because the large AZdr values indicate the drops are large; however, they must be in a low concentration to produce a low-enough AZ value.
In the maritime, RICO cumulus-cloud data (Fig. 4a) there are few instances of low values of Z and high values of Zdr; Z and Zdr are well correlated, and most of what scatter (low correlation) there is comes from growing clouds. In the QCSRP cumulus clouds, there is much less correlation—in particular, from growing clouds. This lower correlation is attributed to the presence of more continental clouds. A more detailed look at the Queensland data in the next section supports that attribution.
b. Within Queensland
Figure 5 shows time–height plots of AZ and AZdr for four individual clouds on different days from Queensland that represent the four basic types of “raindrop size evolution” that were observed. The insets in Fig. 5 show the time evolution of AZ versus AZdr for a single elevation-angle sweep at approximately 1.5 km MSL, which corresponds to 0.5–1 km above cloud base. The lines connecting the individual points within the insets show time evolution, with red representing the growth phase to the maximum AZ and blue representing the dissipation phase. The plots of AZ versus AZdr will be referred to as plots of drop size evolution.
Type 1a (Fig. 5a) is very similar to the RICO data (Fig. 4a), with most of the points on the drop size evolution plots between the reference curves and similar growing and dissipating values. Type 1b (Fig. 5b) is similar to type 1a except initially a single point has abnormally large AZdr for low AZ, which is to the right of the right-hand reference curve. Requirements for type 2a (Fig. 5c) are that during the early growth stage AZdr > 2.0 dB for all AZ < 25 dBZ. Thus during the early growth phase type-2a cases have points to the right of the right-hand reference curve. For type 2b (Fig. 5d), points had to also occur to the right of the right-hand reference curve at some time during the growth stage; AZdr is <2.0 dB for AZ < 25 dBZ during the early growth stage, however, and AZdr decreases markedly as AZ reaches its maximum.
The type-1b cases are especially unusual in that they closely resemble the type-1a cases except for the large single AZdr value for the first point during the growth period. This suggests a very brief period of very few large drops when the rain first begins.11
c. Comparison of cases on the same day
Table 2 indicates consistency in the drop size evolution type within the same day, which is expected because the airmass characteristics do not usually change suddenly. Cells that began to form very close to each other in time and space often had very similar drop size evolutions. Cells C12, C13, and C14 on 23 January 2009 are such an example (Fig. 6a). These three cells initiated within 10 km and 10 min of each other. In contrast, 2 February 2008 has seven cells that initiate within 24 min of each other in an area of 35 km × 10 km that are all type-1a cells except for one type-2b cell (C06) that was in the middle of all of the other cells (Table 2). Figure 6b shows the drop size evolution for C06 and two nearby cells (C02 and C03). No explanation for this has been found in the data.
5. Interpreting the results
a. The effect of the aerosols
Our main hypothesis for explaining the difference between the types of raindrop size evolution is that the type-1 cases occur in maritime air with low aerosol concentrations whereas type-2 cases occur in continental air with relatively high aerosol concentrations. To test this, an estimate is made of the source region for air entering cloud base using a backward–air trajectory model called Hybrid Single-Particle Lagrangian Integrated Trajectory (HYSPLIT; Draxler and Hess 2004). For details of this technique see Tessendorf et al. (2011). In the analysis presented herein, air trajectories that were both below 2 km and spent 12 or more hours over land prior to reaching the cloud at cloud base are classified as continental; otherwise they are classified as maritime. The rationale for using 12 h as the breakpoint between continental and maritime is based simply on a statistical approach called maximally selected rank statistics (e.g., Lausen and Schumacher 1992) in which the in situ aerosol measurements were compared with trajectory time over land to empirically find a time that best divides the two populations of aerosol concentrations. The literature does not provide specific guidelines to quantify the time required for airmass modification because the aerosol concentrations will be very dependent on meteorological factors and aerosol-source characteristics that vary by location and time. The constraint to be below 2 km is a rough proxy for the height of the boundary layer, in which one would expect much of the aerosol population to reside. Table 2 lists the cases classified as maritime and continental according to the trajectories, as well as the raindrop size evolution type and aerosol concentration measured from aircraft with the PCASP instrument.
Table 2 shows that all but 1 of the 22 type-1 cases occurred on days classified as maritime (over land for <12 h) and all but 7 of the 23 type-2 cases occurred on days classified as continental. As expected, the available PCASP aerosol concentrations also correlate with the maritime/continental distinction based upon trajectories. In the continental column all of the PCASP measurements were >350 cm−3, and in contrast all but four of the PCASP measurements in the maritime column were <350 cm−3. The air trajectories for three of these four measurements passed over the greater Brisbane, Queensland, area, which presumably explains their high aerosol concentrations. Table 2 establishes a remarkably good correlation between raindrop size evolution type, continental or maritime air trajectory, and aerosol concentration. The more maritime clouds tend to have type-1 raindrop size evolutions, and the more continental clouds tend to have type 2.
Seven type-2b cases are in the maritime air trajectory column in Table 2. The air trajectory for two of these cases passed over Brisbane and had a more continental aerosol concentration. However, four of the other type-2b cases that had maritime air trajectories had more maritime aerosol concentrations. There were no aerosol data for the seventh type-2b case.
b. Possible influence of ice processes
The possibility that ice processes influenced the drop size evolution results needs attention, especially because the continental cases tended to have higher cloud tops than did the maritime cases. The maximum height of the 0-dBZ echo top, up to and including the time of maximum AZ, averaged 7.5 km (varying between 5.6 and 9.7 km) for the continental cases and 5.7 km (varying between 3.9 and 8.1 km) for the maritime cases. The average height of the freezing level (0°C) was 4.7 km. On that basis, some of the large drop occurrences in the growth stages of the continental cases could have been melted graupel.
To examine this possibility, echo-top heights during the cell growth stages were compared with the temperatures from the Brisbane 0000 UTC (1000 local time) soundings. The type-1a cases are left out of this analysis because they did not exhibit the early volume scans with the large drops and high AZdr along with relatively low AZ at 1.5 km. Figure 7 is a plot of the aerosol concentration versus the time difference between the time at which the echo top crossed the freezing level on the sounding and the first occurrence of the high AZdr with low AZ. For example, a value of 10 min means the echo top crossed the freezing level 10 min prior to the first large AZdr value (right of the right-hand reference curve) at 1.5-km height. Cases in which the top did not pass the freezing level until after the first large AZdr values are plotted just to the right of time 0. Of the 25 cases plotted, 19 had tops that crossed the freezing level 7 min or less before the first large drops were detected. That length of time makes it virtually impossible for ice to nucleate, grow from vapor to the size at which it can start riming, grow into a graupel, and finally fall to a height of 1.5 km, melting along the way. Ice involvement in the remaining six cases (10 and 15 min) is doubtful but cannot be ruled out.
The data as a whole support the proposition that, in general, ice processes are not the cause of the relatively large drops in the early growth stage of the clouds.
c. Possible dynamic influences
In addition to the difference in early cloud-top heights between the continental and maritime cases, the 0-dBZ echo-top rise rates in the continental cases frequently were faster, averaging 5.1 m s−1 as compared with 2.7 m s−1 for the maritime cases. This suggests that during the growth stage the updrafts in the continental cases were likely stronger. An examination of soundings does not reveal a consistent difference in the atmospheric stability, thus raising a question as to whether the updrafts were actually stronger.
At this point the explanation of higher Zdr coincident with low Z in continental clouds is that it is probably the result of the aerosol populations. Direct evidence will require better observations of concentrations of large and small aerosol and vertical velocities [e.g., similar to the type of observations reported by Hudson et al. (2011)].
d. Possible seeding influences
Sixteen of the 45 clouds in this study were seeded from aircraft with hygroscopic flares. For five of these cases the seeding occurred during the dissipation stage. For another four cases the seeding started after relatively large AZdr values to the right of the right-hand reference curve were already occurring. In another two cases large values never did occur. That leaves only five cases in which the seeding commenced early enough to have caused the relatively large AZdr values coupled with low values of AZ. These five cases have boldface cell numbers in Table 2. Three of these cases were continental clouds of type 2a. Their drop size evolution was similar to the other type-2a cases. The remaining two cases were type 1b and type 2b with maritime trajectories that passed over Brisbane and had a high aerosol concentration. The data presented here do not suggest any influence of seeding effects.
The Queensland Cloud Seeding Research Program provided the first opportunity to utilize polarimetric radar to observe raindrop size evolution in seeded clouds by making use of the radar polarimetric variable Zdr: the differential radar reflectivity. It quickly became apparent that there were significant variations in the raindrop size evolution from one cloud to another, however, thus making it essential to examine the natural variability in the drop size evolution before looking for possible cloud seeding effects.
The primary finding from this study is that there are two basic raindrop size evolutions: one associated with continental air masses that are characterized by high aerosol concentrations, long air trajectories over land, rapid cloud-top growth, and higher cloud tops and the other occurring in maritime air masses that are more often characterized by lower aerosol concentrations, long air trajectories over water, slower cloud-top growth rates, and warmer cloud tops. The difference in the two types of raindrop size evolutions is in the growth stage of the clouds when the continental clouds typically had larger Zdr (raindrop sizes) values than the maritime clouds, at relatively low values of the radar reflectivity. They were all very similar during the dissipation stage. The raindrop size evolution with many of the maritime air masses was very similar to those observed in the maritime air mass of the Caribbean Sea that was observed with the S-Pol polarimetric radar during the RICO project. Nonetheless, it is shown herein that within the same air mass there still could be considerable variability in the drop size evolution. In general, the evolutions were most similar for cells that occurred close together in time and space; occasionally there were cells with unexplained large differences from a neighboring cell, however.
Because of the large natural variability in the evolution of raindrop size distribution, this study suggests a preferred method for examining possible seeding effects would be to seed individual clouds in their very early growth stage and compare differential radar reflectivity evolutions with a “sister” cloud that is very close in time and space to the seeded cloud. Furthermore, looking for seeding effects by statistical grouping of cells without consideration of airmass characteristics and proximity in time and space would likely introduce considerable noise into the analysis and decrease the likelihood of observing any cloud seeding effect.
Although there remains uncertainty in the explanation of the early formation of large raindrops in the continental cases, the simplest explanation may be that continental clouds have greater colloidal stability (Squires 1958). In the maritime clouds, coalescence naturally starts earlier and a full raindrop size distribution develops quickly, whereas the coalescence commences more slowly in the continental clouds, with only a small concentration of “luckier” drops, perhaps getting a head start by being initiated on ultragiant aerosols that grow quickly to large size. It was shown that ice processes are probably not important for these results, although the eventual continental cloud tops can be very cold.
To gain more confidence in the cause of the different raindrop evolution types, measurements are needed of each cloud’s immediate thermodynamic environment, updraft strength, drop size distribution within the cloud, and aerosol concentrations in the updraft/inflow region. Of particular importance, and that which is missing from this study, would be the measurement of giant and ultragiant nuclei.
It is recommended that a field campaign be designed that would make significant progress toward understanding the reason for relatively large raindrops during the growth stage in continental clouds and that in general would lead to improved insight to the rain formation process. A location like Brisbane that experiences both continental and maritime clouds would be appropriate. The experiment should have available a polarimetric radar able to obtain high-quality Zdr measurements. Aircraft measurements of aerosols, thermodynamic variables, and microphysical variables would be essential together with a mobile radiosonde unit and a second Doppler radar so that dual-Doppler estimates of vertical air motions could be estimated within the clouds.
The authors thank Nancy Rehak of NCAR for developing the polygon computer program to obtain the AZ and AZdr values for the Queensland data and Jay Miller for the RICO data. We thank Rita Roberts of NCAR for supplying Fig 1. Ed Brandes of NCAR provided valuable discussions concerning data-quality issues with Zdr. Roelof Bruintjes of NCAR has been very supportive of this effort, providing many useful comments. We thank Mike Dixon of NCAR for his software modification in the field and during the analysis phase. We particularly thank Ken Glasson, the CP2 engineer, of the Australian Bureau of Meteorology, for his assistance in collecting the CP2 data. We are also grateful to Jim Hudson of DRI and two other formal reviewers whose efforts significantly improved the paper.
The National Center for Atmospheric Research is sponsored by the National Science Foundation.
The majority of the seeding undertaken in the QCSRP was with hygroscopic flares at cloud base.
In this paper, all radar reflectivities are equivalent radar reflectivity, and for simplicity the subscript e is not used. It is customary for Z without a subscript to refer to Zh; therefore, we will only use the subscript h when we want to contrast it with Zυ. They typically will be within a few decibels of each other.
Reflectivities Zh and Zυ have the units millimeters to the sixth power per meter cubed; often, however, they are expressed in terms of reflectivity decibels (dBZ), which is measured on a logarithmic scale for which the reference value is Z = 1 mm6 m−3 (dBZ = 10 logZ). When Zh, Zυ, and Zdr are expressed in dBZ, then Zdr = Zh − Zυ.
The PCASP used was a PCASP-100X with the upgraded processing system (Signal Processing Package-200), manufactured by Droplet Measurement Technologies, Inc., of Boulder, Colorado.
Negative dBZ values are possible, but by using a threshold of 0 dBZ noisy nonprecipitating and Bragg scatter echoes will be significantly reduced.
These 45 cells were selected independent of the selection used for comparison with the RICO clouds.
Only 5 of these 16 cells were seeded during the early growth stage when the seeding could have had any effect on the results presented in this paper.
The polygons were drawn to enclose the 3D structure of an individual cell. This meant the polygons followed the vertical tilt of the cell.
Range gates are 150 m in range.
For a full explanation, see the appendix of Knight et al. (2008).
As discussed previously, insect contamination of this first point cannot be completely ruled out.