Introduction

Tropical convective systems play a major role in our climate by affecting the earth's heat, moisture, and radiation budget. Current attempts to model and predict tropical convection must adequately represent the microphysical structure of these clouds, which is critical to a proper understanding or simulation (e.g., Grabowski et al. 1999). In spite of the importance of these clouds, comparatively few studies of their microphysical structure are available and these are limited to a few geographical regions and cloud types. In this paper we report on microphysical measurements made in two tropical regions: a continental region in the southern Amazon and a midoceanic region.

These measurements were made as part of field campaigns for validating and improving the measurements from the National Aeronautics and Space Administration (NASA) Tropical Rainfall Measuring Mission (TRMM) satellite, which was launched in 1997. The TRMM satellite provides the first precipitation measuring radar in space as well as a multichannel passive microwave imaging sensor [the TRMM Microwave Imager (TMI)], a visible and infrared sensor (VIRS) to measure rainfall (Kummerow et al. 1998), and a lightning imaging sensor (LIS) to measure total lighting. These instruments were designed to provide quantitative large-scale measurements of tropical rainfall (Simpson et al. 1996; Kummerow et al. 1998).

Perhaps the most critical factor needed to interpret the data from the TRMM satellite is a comprehensive understanding of the nature of the hydrometeors that are observed by the satellite sensors. The hydrometeor types, concentration, sizes, shapes, phase, and location determine what the satellite measures. The TRMM field campaigns were designed to provide this understanding by measuring cloud and precipitation particles in cloudy regions that were concurrently sampled by the satellite and by multiparameter Doppler radars and other remote sensors. Detailed measurements to characterize the mesoscale and synoptic meteorology of the storms were also made. Airborne in situ and remote sensor data were collected during two major field campaigns. These were held in the southwestern Amazon in the state of Rondonia, Brazil [TRMM-Large-Scale Biosphere–Atmosphere Experiment in Amazonia (LBA), January and February 1999; Silva Dias et al. 2001, hereinafter SD01] and near Kwajalein, Republic of the Marshall Islands [Kwajalein Experiment (KWAJEX), August and September 1999]. These locations were chosen to provide validation for TRMM in a tropical continental and tropical oceanic region, respectively. An initial test program was conducted in Texas [Texas–Florida Underflights Experiments (TEFLUN)-A] and eastern Florida (TEFLUN-B) during 1998.

A number of recent studies (e.g., Nesbitt et al. 2000; Petersen and Rutledge 2001) have shown that the Amazon region displays convective characteristics that are intermediate between continental locations like the African Congo and maritime regions such as the west Pacific warm pool. Moreover, during the Amazon wet season, convective intensity appears to be modulated by shifts in the low-level wind, similar to the maritime continent region of northern Australia (Cifelli et al. 2001, hereinafter Cif01; Halverson et al. 2002; Petersen et al. 2002; Rickenbach et al. 2001, hereinafter Ric01; Williams 2001, manuscript submitted to J. Geophys. Res., hereinafter Wil01). Because of the correspondence between convective intensity and the vertical structure of latent heating, it is important for TRMM field campaigns to validate hydrometeor distributions in these different regimes.

The airborne measurement program was designed to obtain data on the location, sizes, types, and phase of hydrometeors in the convective and stratiform rain-producing regions of tropical cloud systems. One of the major goals of the TRMM program is to use the satellite-measured rainfall and vertical reflectivity profiles in the Tropics as validation for cloud resolving models. Cloud resolving models, such as the Goddard Cumulus Ensemble Model (GCE), are used to derive vertical profiles of hydrometeors and to relate this information to the profile of vertical heating rate (Tao et al. 1990 and references therein). These vertical hydrometeor profiles are also used for passive microwave rainfall retrievals (Kummerow and Giglio 1994a,b; Kummerow et al. 1996; Kummerow 1998), which are strongly affected by the form of the droplet size distribution (e.g., Coppens and Haddad 2000). Thus, it is important to validate the cloud resolving models and satellite retrieval algorithms with in situ microphysical measurements provided by the aircraft platforms.

After briefly reviewing some existing microphysical studies of tropical clouds, we describe the instruments and measurement strategy and present results from case studies that illustrate the microphysical nature of the tropical clouds during the field campaigns. We conclude with a summary and plans for future work.

Unlike most midlatitude clouds, tropical clouds have warmer cloud bases and weaker vertical velocities in their lower levels, which provides enough time for rain formation by the warm rain process without the presence of the ice phase. However, the physical mechanisms that control the rate of rain production and the form of the raindrop size distribution are not well understood. (Drops larger than 0.1 mm in radius are referred to as raindrops; those smaller as cloud droplets). Beard and Ochs (1993) reviewed our present state of understanding of this process.

The rate of warm rain formation is likely to vary geographically, due to differences in cloud condensation nuclei (CCN) and other factors that might influence the formation rate of precipitation (e.g., Rosenfeld and Lensky 1998; Wil01). For example, one would expect differences between continental and maritime (e.g., oceanic) tropical convection due to differences in their CCN population (e.g., Rosenfeld and Lensky 1998). However, previous microphysical studies of tropical convection have focused primarily on oceanic systems, so comparisons have not been possible.

Studies of deeper tropical oceanic convection indicate that clouds are usually glaciated above the freezing level. Recently, Tesmer and Wilheit (1998) reviewed previous studies describing the amount and vertical distribution of liquid water in oceanic tropical clouds. The mid-to-lower parts of these clouds are expected to have rather low vertical velocities (LeMone and Zipser 1980; Zipser and LeMone 1980; Jorgensen and LeMone 1989; Jorgensen et al. 1985; Lucas et al. 1994), with most regions sampled in these studies exhibiting peak values less than 10 m s⁻¹, even in hurricanes. However, Black et al. (1994) reported unusually strong updrafts in Hurricane Emily. They observed sustained updrafts greater than 20 m s⁻¹ near the freezing level and downdrafts as strong as 19 m s⁻¹.

Although most of the differences in updraft velocities between continental and oceanic convection appear to be well-documented, the reasons for these differences are not clear (e.g., Michaud 1996; Lucas et al. 1996). Wei et al. (1998) concluded that entrainment and mixing played the major role in reducing the buoyancy of clouds in the Tropical Ocean and Global Atmosphere Coupled Ocean–Atmosphere Response Experiment (TOGA COARE). A consequence of the low-level weak updrafts is that there is ample time for the warm rain process to produce precipitation prior to air reaching the freezing level. This results in a large fractional rainout prior to the freezing level and, above that level, ample time for glaciation to occur. Szoke et al. (1986) examined 296 vertical profiles of radar reflectivity during the Global Atmospheric Research Program (GARP) Atlantic Topical Experiment (GATE) with surface radar reflectivities of 40 dBZ or greater. They concluded that large particles are rare above the freezing level, which they attributed to weak updrafts above the freezing level. Since a likely mechanism for electrification of thunderstorms involves riming of large ice particles suspended in strong updrafts, these observations are consistent with the relatively reduced amount of lightning that occurs in deep tropical oceanic clouds (Zipser 1994; Petersen et al. 1996; Boccippio et al. 2000).

Black and Hallett (1986, 1999) have studied the microphysics of Atlantic hurricanes. They found supercooled droplets only in convective updrafts greater than 5 m s⁻¹, and not all updrafts greater than 5 m s⁻¹ contained supercooled droplets (Black and Hallett 1986). However graupel was found in all updrafts colder than −2°C, suggesting the presence of supercooled droplets at some point in the lifetime of hurricane updrafts. Columnar ice crystals and aggregate ice particles were found in stratiform regions of the storms. Similar observations of aggregate ice particles and low or nonexistent supercooled droplets in stratiform regions are provided by McFarquhar and Heymsfield (1996) for tropical anvil clouds observed in the central equatorial pacific, and by Houze and Churchill (1987) for a mesoscale system near the Bay of Bengal. Black and Hallett (1986) suggested that circulation of ice from high levels of the cloud to lower regions was a factor in the rapid glaciation of the clouds. Black and Hallett (1999) concluded that in symmetrical, mature hurricanes, supercooled droplets usually occur only in regions warmer than about −5°C. Jorgensen et al. (1985) found that cloud liquid water (in droplets less than about 40-μm diameter) reached a maximum of about 1 g m⁻³ near the freezing level of four hurricanes and decreased rapidly at higher altitudes. They also found a decrease in radar reflectivity above the freezing level in these storms.

Liu and Curry (1998) used an ice water path retrieval algorithm to study the relationships between ice water path (IWP), total liquid water path (LWP, including cloud and rainwater), and cloud-top temperature for tropical clouds during TOGA COARE. They found that cloud-top temperature was well-correlated with IWP, but poorly correlated with total LWP. They also found that precipitating clouds have greater IWP than nonprecipitating clouds with the same cloud-top temperature. In a similar study (Liu and Curry 1999), they found that rainfall rate has a strong correlation with IWP. This suggests that the ice phase is important in producing rain from deep tropical convective systems, even though an active warm rain process is also present.

Instrumentation and measurement strategy

Several research aircraft were used during the TRMM campaigns. For the TRMM-LBA campaign, the University of North Dakota (UND) aircraft (Citation) was used for in situ microphysical sampling, while the NASA ER-2 studied the same clouds from above. During the KWAJEX campaign the UND Citation sampled the middle regions of the clouds while the University of Washington Convair aircraft sampled lower levels and the NASA DC-8 obtained both in situ and remotely sensed data from the upper levels. The campaigns were very successful and a large dataset characterizing tropical cloud hydrometeors is now available.

The focus of this study is to illustrate the primary microphysical features of clouds in the two campaigns, using data primarily from the UND Citation. Future studies are planned using the larger dataset.

The Citation instrumentation consisted of standard instruments for state parameters (pressure, temperature, dewpoint, position), a radome gust probe–inertial navigation system (INS) wind and turbulence measurement system, a set of Particle Measuring Systems, Inc., (PMS) probes and two Stratton Park Engineering Company (SPEC) probes for microphysical measurements. Instruments for state parameters included a Rosemount total temperature probe, an EG&G cooled mirror dewpoint instrument, Rosemount dynamic and static pressure sensors, and a Litton LTN-76 model INS.

For the wind instrumentation, a set of maneuvers was done to calibrate the radome system. These consist of a series of pitch, yaw, and reverse track maneuvers. These are used to calibrate the system and derive three-dimensional winds, based on methods described in Lenschow (1972). Even though the radome is heated to prevent most freezing, water did occasionally freeze in some of the pressure sampling lines, resulting in large excursions in measured wind. These periods of large deviations were obvious and are excluded from the analysis.

PMS instruments utilized on the Citation during the field campaigns included a forward-scattering spectrometer probe (FSSP)-100 (TRMM-LBA), a two-dimensional cloud probe (2DC) with 30-μm diode spacing, a one-dimensional precipitation probe (1DP) (TRMM-LBA), and a one-dimensional cloud probe (1DC) (KWAJEX). Calibration of the PMS probes is done by using glass beads for the FSSP and 1DC probes and by using a spinning disk calibrator for the 2DC and 1DP probes. SPEC probes included a cloud particle imager (CPI) for particle microphotography and a high-volume particle sampler (HVPS), which has a large sample volume to capture the largest hydrometeors. Instruments for measuring liquid water included a PMS Commonwealth Scientific and Industrial Research Organization (CSIRO) King-type hot wire probe, and a Rosemount icing detector for supercooled liquid water. The King instrument is a standard instrument for measuring liquid water; however, it has been observed to respond somewhat to ice crystals (e.g., Cober et al. 1995). A review of data collected in glaciated clouds (those with high concentrations of ice observed by the 2DC probe and no response by the icing detector) in the TRMM campaigns revealed an ice response as high as 0.3 g m⁻³. Collisions with dense graupel or frozen droplets made it difficult to operate this instrument during KWAJEX, so rather limited data on liquid water is available for that project. The icing detector was available to detect the presence of supercooled liquid water, but is not used here for quantitative liquid water measurements. This instrument detects the freezing of supercooled liquid water on a vibrating rod. As ice builds up on the rod, the output (volts) from the instrument increases. A heating cycle heats the rod, melting the ice and decreasing the voltage. In high continuous liquid water, the output consists of a series of steps as ice builds up and is melted by the heating cycle. In brief encounters with supercooled water of low concentration, a variation in voltage is observed as the ice builds up and sublimates, without the cycling of the instrument. We refer to these regions as having traces of supercooled liquid water.

For plotting time series of data, 1-Hz averages of data recorded at 4 Hz are used. For the 2DC the total counts (referred to as the “shadow-or”) were used to provide 2DC time series concentrations. This parameter is usually dominated by the presence of smaller droplets or ice particles that have grown to detectable sizes (greater than 30 to 50 μm). Hence, this variable is generally an indicator of small precipitation particles. Size distribution data are computed from the 2DC data using the reconstruction method of Heymsfield and Parrish (1978). Particle size refers to the reconstructed size using this method. We have not included the HVPS measurements here because there were considerable problems during TRMM-LBA and it was significantly modified for KWAJEX. Its performance, and the effects caused by this change are currently being evaluated; these data will be included in future studies. For the present study, we use the 2DC, which was the same for each program, to provide comparisons without the bias caused by different instruments.

Aircraft flight patterns included spiral ascents and descents and horizontal legs stepping up or down through different cloud temperatures. These patterns were flown in both convective and stratiform regions, although spirals were mostly limited to stratiform cloud regions. Only precipitating cloud systems, or cloud turrets near them, were selected for study, based on the presence of radar returns near the sampling location. However, both precipitating and nonprecipitating portions of these systems were generally sampled. The aircraft could safely penetrate most oceanic cloudy regions during KWAJEX, but with the exception of a few occasions, was not able to sample the more intense regions of the continental Amazonian storms. We present examples of data from six storms, three from TRMM-LBA, and two from KWAJEX and one from TEFLUN-B. It is likely that some of these will become the focus of major case studies in the future.

Microphysical characteristics of convective regions in three Amazonian storms

Microphysical characteristics of convective regions in two Kwajalein storms

Two convective cases from Kwajalein (17 and 22 August 1999) are presented here for comparison with the Brazilian cases. Both cases contained relatively isolated convective cells with substantial updrafts noted by the flight crew and measured values briefly exceeding 10 m s⁻¹. Cloud base was not measured in this study, but LCLs were estimated to be very close to the surface: 23°C (975 hPa) and 21°C (950 hPa) for 17 and 22 August, respectively.

The results from passes at 0°, −11°, and −17.5°C on 17 August are given in Figs. 10 and 11. Liquid water measurements from the King probe were not available on 17 August. The concentrations of small raindrops were correlated with updraft regions at 0°C (Fig. 10). Large droplets were found outside of the updraft regions or in regions of weaker updraft (e.g., at 2225:15 UTC). A similar correlation between small particles and updrafts was found at the colder temperatures (Fig. 11). Irregular rimed or aggregate ice particles were found outside of the updraft regions. (CPI data are not available for this case.) The icing detector responded briefly in colder regions, but did not cycle (Fig. 11b), indicating traces of liquid water.

Figures 12, 13, and 14 show the results of passes made at approximately 0°, −6°, and −11°C through similar cells on 22 August. CPI imagery is given in Fig. 15. Figure 12 shows higher concentrations of small droplets in the updraft regions at 0°C and melting ice particles outside of the updraft regions. Slightly less than 1 g m⁻³ of liquid water was found (Fig. 12b).

At −6°C a wide variety of particle types were found (Fig. 13). Upon entering the cloud, columnar crystals were found on the 2DC images, which appeared as sheath ice particles in the corresponding CPI images (Fig. 15). The highest concentrations of small particles correlate with the updraft in Fig. 13; these particles were a mix of droplets and ice particles, with some large drops on the edge of the second updraft. About 0.8 g m⁻³ of liquid water was found in the updraft regions (Fig 13b). Between the updrafts, millimeter-sized graupel was observed. (Shortly after this pass, graupel or frozen droplets broke the hot wire on the King liquid water instrument.)

At −11°C the highest concentrations of small particles were also found in updraft regions (Fig. 14) and larger, rimed particles were found in weaker updraft regions. The icing detector indicated only traces of liquid water.

The results from these measurements illustrate the development and form of precipitation in moderately strong (for oceanic cumuli) updraft regions. At the 0°C level, small drizzle-sized drops are evident, most likely due to the effects of collision and coalescence of cloud droplets at warmer temperatures. Many of these are carried up to colder temperatures and continue to grow as supercooled raindrops (e.g., the raindrop images from the pass at −6°C shown on the right-hand side of Fig. 13a or in Fig. 15). Between about −6° and −18°C most drops of all sizes freeze, as significant quantities of supercooled liquid water were found only in the lower portions of the updraft at warmer temperatures. This is in accord with earlier measurements of supercooled liquid water in oceanic convection. Thus, one would expect riming growth to be limited to lower regions of the cloud. However, trace amounts of liquid water were found at −17.5°C and warmer, which suggests that there were enough unfrozen drops present to maintain water saturation in stronger updrafts.

The development of precipitation-sized ice particles (between 0.1 and 1 mm) is illustrated by the changes in their particle size distribution from warmer to colder temperatures (Fig. 16). For example, the concentration of 0.5-mm sizes increases by over an order of magnitude between the freezing level and −17.5°C. Thus, the conditions present in these clouds (e.g., the amount of water vapor and liquid available to form ice particles and the number of small ice particles available for growth) favors the formation of relatively large numbers of ice particles in the 0.1- to 1-mm size range.

The particle images in the updraft at −17.5°C on 17 August appear very similar to the images collected in TRMM-LBA updrafts at similar temperatures (Figs. 5 and 11). The size distributions for these cases (Fig. 17) were also very similar.

Microphysical characteristics of three stratiform regions

The characteristics of three deep stratiform regions associated with convective storms are presented in this section. These were from cases in Florida, Brazil, and Kwajalein on 20 August 1998, 17 February 1999, and 11 September 1999, respectively. Several other cases were available, but these were selected for comparison because they were all deep clouds that were sampled over most of their vertical extent. More extensive analysis of stratiform cloud regions from the TRMM field campaigns is provided in a companion paper (Heymsfield et al. 2001, manuscript submitted to J. Atmos. Sci.). Data are presented as vertical profiles (with temperature as the vertical coordinate). The concentrations of larger precipitation particles (larger than 0.3 mm) and their mean volume diameter were computed from the 1-Hz measured 2DC size distributions. The results are given in Fig. 18. Several features are evident. The profiles for Florida and Kwajalein are rather similar. The profile for 17 February is also similar, but with more variability. Examination of radar cross sections for this case revealed the presence of small convective features embedded in the stratiform region. These likely contributed to the variability for this case. The concentration of large precipitation particles reaches a maximum at the midlevel of the clouds at approximately a temperature of −18°C, and decreases at lower altitudes. These particles were all ice at altitudes above the freezing level. The mean volume diameter of the ice particles increases nearly monotonically with warmer temperatures, then decreases below the freezing level as the particles melt.

Average ice particle size distributions were computed for the upper (<−25°C), mid- (−20° to −10°C), and lower (−5° to 0°C) regions of the clouds (Fig. 19). In each case increases in the concentration of particles larger than about 0.8 mm become evident as warmer temperatures are sampled. Examination of particle imagery (Fig. 20) for each of these cases reveals an increase in the size of aggregates at warmer temperatures, and may explain observed decreases in the concentration of precipitation at warmer temperatures, where aggregation is expected to be most active (e.g., Hobbs et al. 1974). Riming sufficient to be identified on the 2DC imagery was not observed. Examination of the icing detector data also indicates no evidence for supercooled liquid water, suggesting that riming was not occurring in these stratiform regions.

Summary

These results suggest several characteristics and similarities of the TRMM tropical clouds:

• As expected, the warm rain process dominates precipitation formation in shallow convective clouds with weak updrafts. In these clouds, precipitation formation occurs by the time the cloud has reached the freezing level in both the southwest Amazon clouds and clouds near Kwajalein.

• Ice formation and growth is the primary process producing precipitation-sized particles above the freezing level in deeper convective clouds and in regions with strong updrafts.

• A number of similarities are evident in the microphysical characteristics of clouds in the Amazon and near Kwajalein. In both cases, stronger updrafts (greater than about 5 m s⁻¹) contained higher concentrations of small water or ice particles than weaker updrafts, and this was true for both warm and cold (midlevel) portions of the clouds. The size distributions and particle images in comparable midlevel updrafts were similar (Figs. 5, 11, and 17). In both regions, significant amounts of supercooled liquid water were found at temperatures warmer than about −7°C, and decreased substantially at colder temperatures. (Of the other cases that have been examined to date, the coldest temperature where significant liquid water occurred was in TRMM-LBA on 13 February 1999, where 0.51 g m⁻³ was found at a temperature of −12°C.) Traces of supercooled water were found at colder temperatures (e.g., −18°C), suggesting that water saturation can be maintained at colder temperatures, even if not enough supercooled water is present for significant riming growth. Graupel was found in weaker updrafts. These results are generally in accord with earlier microphysical observations in tropical convection, except that we found supercooled water at somewhat colder temperatures than reported in earlier studies.

• The results from the strong updraft on 26 January suggest that the intersection region between a cloudy updraft and a cloudy downdraft may be favorable for the growth of precipitation particles, as droplets from the older cloud region in the downdraft may be entrained into the liquid water-rich updraft region, where further growth by accretion is likely. A similar mechanism for the growth of precipitation in continental clouds has also been proposed (Dye et al. 1976).

• In the upper regions of clouds, aggregates were observed to have chainlike structures (e.g., Figs. 7 and 8). Electrical forces, or some other mechanism, might cause this type of aggregation in the cold regions where these particles were found (e.g., −43°C in Fig. 8), as compared with aggregation caused by differences in particle fall speed at warmer temperatures (e.g., Fig. 15).

• Ice particles sampled in deep stratiform clouds from the three geographically separated regions also exhibited a number of similarities. In each case, particle size increased with decreasing altitude because of the effects of aggregation. Riming did not appear to play a significant role in particle growth in the stratiform regions sampled, as supercooled liquid water was not detected. The particles may have experienced some riming at an earlier stage in their lifetime; however, graupel was not found in significant amounts in the stratiform cases (the degree of riming for lightly rimed particles is difficult to determine from the 2DC imagery). The presence of aggregates and low liquid water in these cases is similar to observations made elsewhere in tropical regions (e.g., Houze and Churchill 1987; McFarquhar and Heymsfield 1996).

Discussion

There are likely to be major differences in the dynamics of continental (TRMM-LBA) and oceanic (KWAJEX) tropical regions (e.g., differences in updraft strengths). Differences in cloud active nuclei are also likely to influence the initial formation of precipitation. However, in our observations, certain parts of the clouds with similar kinematic features (e.g., updrafts near −18°C; stratiform areas) from the different geographical regions displayed similar microphysical features (e.g., particle size distributions, particle images, and liquid water content). These are encouraging results for the TRMM simulation efforts, because they suggest that similar microphysical retrieval schemes (e.g., those for handling ice) might work for both regions, provided the kinematic structure of the storms can be properly simulated. In both regions the convective clouds exhibited high concentrations of small- to moderate-sized (0.1–1 mm) ice particles when they were sampled between −11° and −18°C. Many of these smaller ice particles are likely to be detrained into stratiform anvil regions, where further aggregation of the particles occur.

Tao et al. (1990) used a composite sounding from three GATE squall-type mesoscale convective systems as input to the GCE. The model output provided vertical profiles of hydrometeors for convective and stratiform regions, which was then used to derive vertical heating profiles. The major ice hydrometeor type in the simulation for their mature anvil and their decaying anvil stratiform cases was graupel. Our observations and those from earlier studies suggest that their simulation did not reproduce the proper hydrometeor type for these cases. Because of the sensitivity of numerical model-derived precipitation to the assumed particle fall speed relationship (e.g., Ferrier et al. 1995), this is likely to significantly affect the rainfall retrieval algorithms and the calculation of vertical heating profiles.

Because of the presence of graupel and droplets larger than 23–25 μm, these clouds should be ideal candidates for Hallett–Mossop secondary ice production by riming at temperatures between −3° and −8°C. However, graupel and supercooled water were found together primarily in weaker updrafts where precipitation is likely to be well-developed (by the warm rain process) by the time that the clouds reach these temperatures, which probably limits the importance of this process. Figure 13 provides an example from KWAJEX at −6°C. Different portions of the cloud, at constant altitude, contain cloud water, raindrops, drizzle drops, graupel, and sheath ice particles. The sheath ice particles are found near the edge of the cloud in a slight updraft. Since this type of habit grows in the Hallett–Mossop temperature range (and slightly colder) and the other necessary conditions are found nearby, these are likely to have originated by this process. Yet, it is clear that these types of ice crystals predominate in only a small portion of the cloud (Fig. 13) at this temperature.

Near the freezing level and slightly colder, many updrafts contained substantial amounts of drizzle (Figs. 6a,b; 10; and 12). The concentrations of the drizzle, as measured by the 2DC instrument, were typically in the range of about 100–600 L⁻¹. In colder updraft regions (Figs. 5, 6c, 11, 14), where ice was the primary particle sampled by the 2DC, the maximum concentrations were significantly greater (about a factor of 2) than the maximum concentrations found in the lower drizzle regions. This enhancement in concentration is roughly in agreement with what might be expected due to the shattering and splintering of these drizzle droplets as they freeze (e.g., Pruppacher and Klett 1997).

On average, droplet concentrations in low-precipitation updrafts in Brazil (e.g., Figs. 2, 3, 6) were a few hundred per cubic centimeter, which is reasonable for a moderately clean continental regime. Future studies will examine variability in the droplet concentration as a function of synoptic regime (e.g., low-level westerly regime periods were observed to be associated with more oceanic-like convection; Cif01; Ric01; Wil01).

More detailed studies are currently in progress using the multiplatform measurements available from the TRMM field campaigns. These will be compared with simulations of these clouds and measurements from the TRMM satellite. Our results to date suggest that these are likely to produce substantial improvements in our understanding of tropical clouds, their effects on the vertical heating profile of the atmosphere, and our ability to monitor and predict their effects on the climate.