• Albrecht, B. A., 1989: Aerosols, cloud microphysics, and fractional cloudiness. Science, 245 , 12271230.

  • Battan, L. J., 1953: Observations on the formation and spread of precipitation in convective clouds. J. Meteor., 10 , 311324.

  • Berry, E. X., , and R. L. Reinhardt, 1974: An analysis of cloud drop growth by collection. Part IV: A new parameterization. J. Atmos. Sci., 31 , 21272135.

    • Search Google Scholar
    • Export Citation
  • Caylor, I. J., , and A. J. Illingworth, 1987: Radar observations and modelling of warm rain initiation. Quart. J. Roy. Meteor. Soc., 113 , 11711191.

    • Search Google Scholar
    • Export Citation
  • Hudson, J. G., 1993: Cloud condensation nuclei. J. Appl. Meteor., 32 , 596607.

  • Hudson, J. G., , and H. Li, 1995: Microphysical contrasts in Atlantic stratus. J. Atmos. Sci., 52 , 30313040.

  • Hudson, J. G., , and S. S. Yum, 2001: Maritime–continental drizzle contrasts in small cumuli. J. Atmos. Sci., 58 , 915926.

  • Illingworth, A. J., 1988: The formation of rain in convective clouds. Nature, 336 , 754756.

  • Illingworth, A. J., , J. W. Goddard, , and S. M. Cherry, 1987: Polarization radar studies of precipitation development in convective storms. Quart. J. Roy. Meteor. Soc., 113 , 469489.

    • Search Google Scholar
    • Export Citation
  • Keeler, R. J., , P. H. Herzegh, , and C. F. Frush, 1984: Measurements of CP-2 co-polar and cross-polar antenna illumination functions at S- and X-band. Preprints, 22d Conf. on Radar Meteorology, Zurich, Switzerland, Amer. Meteor. Soc., 287–291.

  • Knight, C. A., , and L. J. Miller, 1998: Early radar echoes from small, warm cumulus: Bragg and hydrometeor scattering. J. Atmos. Sci., 55 , 29742992.

    • Search Google Scholar
    • Export Citation
  • Knight, C. A., , J. Vivekanandan, , and S. G. Lasher-Trapp, 2002: First radar echoes and the early ZDR history of Florida cumulus. J. Atmos. Sci., 59 , 14541472.

    • Search Google Scholar
    • Export Citation
  • Mielke Jr., P. W., , K. J. Berry, , and G. W. Brier, 1981: Application of multi-response permutation procedures for examining seasonal changes in monthly mean sea-level pressure patterns. Mon. Wea. Rev., 109 , 120126.

    • Search Google Scholar
    • Export Citation
  • Sax, R. I., , and J. G. Hudson, 1981: Continentality of the south Florida summertime CCN aerosol. J. Atmos. Sci., 38 , 14671479.

  • Wilson, J. W., , T. M. Weckwerth, , J. Vivekanandan, , R. M. Wakimoto, , and R. W. Russell, 1994: Boundary layer clear-air radar echoes: Origin of echoes and accuracy of derived winds. J. Atmos. Oceanic Technol., 11 , 11841206.

    • Search Google Scholar
    • Export Citation
  • View in gallery

    (a) A sequence of RHI scans through a cloud’s maximum X-band radar reflectivity. The cloud developed on 5 Aug 1995, 20 km east-southeast of the radar. The 0-dBZ contour line is highlighted in a thin black line, and maximum and minimum heights of this contour line are indicated. (b) Corresponding time–height cross section showing the cloud’s temporal evolution. Negatively slanted reflectivity isolines are related to the bulk total fall velocity (terminal velocity plus vertical air movement) of raindrops. The relationship between the slope of the reflectivity contours and bulk fall velocities are plotted as black lines. CT indicates the characteristic time for precipitation formation as defined in the text.

  • View in gallery

    Mean bulk fall velocities (dots) obtained from the slope of each contour line at the time of the first occurrence of the (a) −2.5-, (b) 0-, (c) 2.5-dBZ, etc., contour. The standard deviations are depicted as horizontal lines. The solid vertical line indicates a fall velocity of 6.5 m s−1, which corresponds to a drop diameter of 2 mm falling at sea level pressure.

  • View in gallery

    Profile of the adiabatic liquid water content (LWC) on 5 Aug. This figure illustrates how the CTWC for the cloud depicted in Fig. 1 was determined.

  • View in gallery

    Chronology of the SCMS project showing flight averaged CCN concentrations (dots) activated at 1% supersaturation and averaged cloud droplet concentrations (dots) (diameter 2 to 50 μm) with their standard deviations depicted as vertical solid lines (Hudson and Yum 2001). Our classification of air masses based on cloud motion is compared to the classification provided by Hudson and Yum.

  • View in gallery

    Cross-peninsula profile of CCN and Aitken nuclei concentrations acquired at cloud-base level on 21 Jul 1975 with onshore flow on both coasts (from Sax and Hudson 1981).

  • View in gallery

    Horizontal cloud movements (black arrows), obtained by tracking radar reflectivity echoes, indicate the prevailing wind direction: (a) Day with westerly (offshore) winds and clouds developing over land and (b) day with southeasterly (onshore) winds and clouds developing over the ocean. The location of the radar is indicated as black + sign, and the 50-km range as a black circle.

  • View in gallery

    Scatterplot of CT vs CTWC. (a) Dots indicate clouds that developed during days with mean relative humidities >75% within the cloud layer; × symbols indicate clouds that developed during days with mean relative humidities <75% within the cloud layer. (b) Dots indicate clouds that developed during days with standard deviations of the mean wind speed within the cloud layer <1.0 m s−1; × symbols indicate clouds that developed during days with standard deviations of the mean wind speed within the cloud layer >1.0 m s−1. (c) Dots indicate clouds that developed before local noon (<1600 UTC); × symbols indicate clouds that developed after local noon (>1600 UTC).

  • View in gallery

    Scatterplot of CT vs CTWC. (a) Dots indicate onshore moving clouds that developed over land; × symbols indicate onshore moving clouds that developed over the ocean. (b) Dots indicate offshore moving clouds that developed over land; × symbols indicate offshore moving clouds that developed over the ocean.

  • View in gallery

    Scatterplot of CT vs CTWC for all 38 clouds. Symbols indicate the different growth location and gray shades indicate the cloud movement. Data points from 10 Aug are noted.

All Time Past Year Past 30 Days
Abstract Views 0 0 0
Full Text Views 40 40 5
PDF Downloads 20 20 3

Radar Analysis of Precipitation Initiation in Maritime versus Continental Clouds near the Florida Coast: Inferences Concerning the Role of CCN and Giant Nuclei

View More View Less
  • 1 Department of Atmospheric Sciences, University of Illinois at Urbana–Champaign, Urbana, Illinois
© Get Permissions
Full access

Abstract

A method of analyzing radar data is developed and applied to determine whether the X-band radar reflectivity evolution of clouds observed during summertime on the northeast Florida coast during the Small Cumulus Microphysics Study (SCMS) shows distinct differences in precipitation development that can be associated with the clouds’ maritime or continental characteristics. For this study, the entire National Center for Atmospheric Research CP2 radar dataset from SCMS was examined, and 38 clouds were used. For these clouds the evolution in X-band radar reflectivity, from the clouds’ earliest detection through precipitation, was clearly documented and met specific requirements concerning the clouds’ location relative to the coastline and direction of movement. Since cloud condensation nuclei (CCN) and giant and ultragiant nuclei (GN) measurements were not available for the specific clouds used in this study, proxies were used to partition the clouds into four groups based on the cloud location and direction of movement. Specifically, it was assumed that clouds forming over the ocean during onshore flow had maritime characteristics (group 1: low CCN, high GN), clouds forming over land during onshore flow would have modified maritime characteristics (group 2: high CCN, high GN), clouds forming over land during offshore flow would have continental characteristics (group 3: high CCN, low GN), and clouds forming over the ocean during offshore flow would have modified continental characteristics (group 4: high CCN, high GN). These assumptions are based on past measurements presented by Sax and Hudson. Then, these populations were statistically compared using the nonparametric multiresponse permutation procedure developed by Mielke et al. A comparison of groups 1 and 2 provided a test of the role of CCN concentrations in precipitation development in these cloud populations. A comparison of groups 3 and 4 provided a test of the role of GN concentrations in precipitation development in these cloud populations. The two cloud populations that were disjoint at a statistically significant level were groups 1 and 2. For these groups, the analysis showed that the median characteristic total water content of the truly maritime clouds (group 1) was about half that of the modified maritime clouds (group 2) at the time of precipitation formation. The characteristic time to precipitation formation was about 60% smaller for the truly maritime clouds. Thus, the characteristic reflectivity threshold for precipitation development was reached at a much lower altitude above cloud base in a much faster time in the truly maritime clouds. This result supports the conclusions of Hudson and Yum that precipitation development in the SCMS clouds was primarily controlled by CCN concentrations rather than GN concentrations.

* Current affiliation: Department of Physical Sciences, University of Helsinki, Helsinki, Finland

Corresponding author address: Dr. Sabine Göke, Division of Atmospheric Sciences, University of Helsinki, P.O. Box 64 (Gustaf Hällstömin katu 2), 00014 Helsinki, Finland. Email: sabine.goeke@helsinki.fi

Abstract

A method of analyzing radar data is developed and applied to determine whether the X-band radar reflectivity evolution of clouds observed during summertime on the northeast Florida coast during the Small Cumulus Microphysics Study (SCMS) shows distinct differences in precipitation development that can be associated with the clouds’ maritime or continental characteristics. For this study, the entire National Center for Atmospheric Research CP2 radar dataset from SCMS was examined, and 38 clouds were used. For these clouds the evolution in X-band radar reflectivity, from the clouds’ earliest detection through precipitation, was clearly documented and met specific requirements concerning the clouds’ location relative to the coastline and direction of movement. Since cloud condensation nuclei (CCN) and giant and ultragiant nuclei (GN) measurements were not available for the specific clouds used in this study, proxies were used to partition the clouds into four groups based on the cloud location and direction of movement. Specifically, it was assumed that clouds forming over the ocean during onshore flow had maritime characteristics (group 1: low CCN, high GN), clouds forming over land during onshore flow would have modified maritime characteristics (group 2: high CCN, high GN), clouds forming over land during offshore flow would have continental characteristics (group 3: high CCN, low GN), and clouds forming over the ocean during offshore flow would have modified continental characteristics (group 4: high CCN, high GN). These assumptions are based on past measurements presented by Sax and Hudson. Then, these populations were statistically compared using the nonparametric multiresponse permutation procedure developed by Mielke et al. A comparison of groups 1 and 2 provided a test of the role of CCN concentrations in precipitation development in these cloud populations. A comparison of groups 3 and 4 provided a test of the role of GN concentrations in precipitation development in these cloud populations. The two cloud populations that were disjoint at a statistically significant level were groups 1 and 2. For these groups, the analysis showed that the median characteristic total water content of the truly maritime clouds (group 1) was about half that of the modified maritime clouds (group 2) at the time of precipitation formation. The characteristic time to precipitation formation was about 60% smaller for the truly maritime clouds. Thus, the characteristic reflectivity threshold for precipitation development was reached at a much lower altitude above cloud base in a much faster time in the truly maritime clouds. This result supports the conclusions of Hudson and Yum that precipitation development in the SCMS clouds was primarily controlled by CCN concentrations rather than GN concentrations.

* Current affiliation: Department of Physical Sciences, University of Helsinki, Helsinki, Finland

Corresponding author address: Dr. Sabine Göke, Division of Atmospheric Sciences, University of Helsinki, P.O. Box 64 (Gustaf Hällstömin katu 2), 00014 Helsinki, Finland. Email: sabine.goeke@helsinki.fi

1. Introduction

One of the goals of the Small Cumulus Microphysics Study (SCMS), which was conducted in the summer of 1995 near Cape Canaveral, Florida, was to study warm cumuli in their earliest stages, with the goal of understanding the factors that control the time to onset of precipitation. Although the SCMS clouds exhibited little variation in cloud-base height and cloud-base temperature, there were significant variations from day to day in the cloud condensation nuclei (CCN) concentrations, ranging from more maritime conditions (359 ± 142 cm−3) to continental conditions (1411 ± 388 cm−3) (Hudson and Yum 2001, hereafter HY01). HY01 associated these variations with the prevailing wind direction, which in different time periods was either onshore or offshore of the east Florida coast. They noted that the small cumulus clouds in continental air were characterized by consistently higher concentrations of smaller cloud droplets and reduced concentrations of drizzle droplets. They found that there was much more drizzle (which they defined as droplets with diameter >50 μm) in the maritime clouds, which were characterized by larger mean cloud droplet sizes, higher concentrations of large cloud droplets, and greater amounts of cloud liquid water. According to HY01, an apparent cloud droplet mean size threshold for the onset of drizzle was almost never exceeded in the continental clouds but was often exceeded in the maritime clouds, especially at higher altitudes. They concluded that in the SCMS clouds higher CCN concentrations suppressed drizzle formation. HY01 suggested that giant and ultragiant nuclei (GN) apparently had little effect on the formation of drops >50 μm diameter in these clouds. They also argued that these differences in CCN could impact cloud radiative properties through the second Twomey effect, which links CCN to the cloud lifetime (Albrecht 1989; Hudson 1993).

Contrasting evidence concerning the influence of GN on the early evolution of small warm cumulus clouds comes from analysis of dual polarization radar measurements. Polarization radar data obtained during the Precipitation Project (PRECIP-98) in Florida showed that a column of high differential reflectivity (ZDR) extended from near cloud top through cloud base as the early radar reflectivity developed near cloud top (Knight et al. 2002). In the middle to lower part of the cloud, the ZDR signal was characteristic of raindrops with diameters between 1 and 3 mm. These studies confirmed and extended earlier work of Caylor and Illingworth (1987), Illingworth et al. (1987), and Illingworth (1988) that used ZDR to show that low concentrations of large raindrops are present simultaneously with the early weak radar reflectivity echoes in cumulus. In each of these studies, the authors hypothesized that these early large raindrops were a result of growth on GN.

HY01’s findings imply that one should be able to detect a significant difference in the rate at which precipitation develops in continental versus maritime clouds in SCMS using radar. Since the radar reflectivity factor (hereafter reflectivity) is proportional to the sixth power of the diameter of the precipitation particles, differences in the time evolution of the reflectivity should be easily detectable for these two different classes of SCMS clouds. However, if early radar reflectivity development is sufficiently influenced by accretional growth on GN in both the continental and maritime Florida clouds, then the larger droplets forming on the GN may sufficiently dominate the radar reflectivity so that little difference may exist in the early evolution of their radar reflectivity signatures. In this paper, we develop and apply a method of analyzing radar data to determine whether the X-band radar reflectivity evolution of SCMS clouds supports the evidence presented by HY01 that distinct differences in precipitation development can be associated with the clouds’ maritime or continental characteristics.

2. The Small Cumulus Microphysics Study

During SCMS, small cumuli were studied using radar, three instrumented aircraft, atmospheric soundings, and time-lapse videos. Small cumuli typically first appeared in midmorning and developed precipitation when their tops reached 3 to 4 km MSL. Precipitation typically developed within 20 to 30 min of the initial detection of the clouds with radar. The clouds in this study were all isolated and had tops with temperatures >0°C. During SCMS, radar data were collected on 44 consecutive days from 3 July through 19 August 1995. Aircraft operation began on 19 July and continued through 19 August. The early portion of the project was a radar-only period. Special project soundings were launched one to four times per day at the radar site early in the project (3–16 July) or 22 km southwest of the radar later in the project (20 July–14 August).

The data used in this paper come from both the radar-only and the aircraft-supported periods of the project. Data were acquired with the National Center for Atmospheric Research (NCAR) CP2 dual-wavelength radar. The radar, which was located on a barrier island, operated at 3- (X-band) and 10-cm (S-band) wavelengths, with the antennas collocated on the same pedestal and adjusted to the same pointing angle (Keeler et al. 1984). The dual-wavelength measurements were influenced by hydrometeor, Bragg, and nonmeteorological (insect, bird, aircraft, and ground targets) scattering. Bragg scattering dominated the S-band cloud echoes early in their lifetime, producing a characteristic mantle echo defining the cloud outline (Knight and Miller 1998). The X-band cloud signal was primarily attributable to Rayleigh scattering associated with hydrometeors. In Florida, there is a significant contribution to the Rayleigh signal in the boundary layer from insects (Wilson et al. 1994). Clear air boundary layer radar reflectivity echoes range typically between −5 and 10 dBZ.

During SCMS, radar studies of clouds were conducted almost exclusively using range–height indicator (RHI) scans over a narrow range of azimuth (typically 20°–30°). The RHI scans were typically spaced about 1°–1.5° apart. Radar volumes were repeated approximately every 2.5 min. These volumes were interrupted for a 360° surveillance scan approximately every 10 min. The time between RHI radar volumes when interrupted by a surveillance scan was about 4 min. Range gates were every 100 m and the vertical beam spacing ranged from 0.3° to 0.7°. The beamwidth in both the azimuthal and elevation directions was 0.9° for both S- and X-band. The scanning strategy was designed primarily to coordinate radar and aircraft operations. For this reason, the radar was typically fixed within a narrow sector where the aircraft operated. This had the unfortunate consequence that most clouds entered the sector and exited the sector without complete documentation of their life cycle from their earliest appearance through precipitation.

For this study, we examined the entire SCMS dataset to identify all clouds for which the first appearance of any echo (S- or X-band) could be identified and for which the life cycle of the cloud through precipitation development was completely documented. In practice, each cloud that reached at least 10-dBZ maximum X-band reflectivity in its lifetime was tracked backwards in time to the earliest appearance of an X-band echo. The S-band mantle echo from Bragg scattering was further used to confirm the existence of the early cloud at the point of development of the first Rayleigh signal in X-band. Figure 1a shows an example of a series of RHI scans through a cloud’s maximum reflectivity, which represents a typical temporal evolution of an SCMS cumulus cloud in X-band radar reflectivity. The corresponding time–height cross section is depicted in Fig. 1b. This time–height approach was first used by Battan (1953), and later by Knight and Miller (1998) to study precipitation development in convective clouds. Forty-five clouds were identified throughout the experiment that met this criterion.

The clouds observed with radar used in this analysis were not sampled by aircraft because aircraft produce strong radar echoes that render the data unusable. Even if we could have used these clouds, the additional requirement that we observed them from their earliest appearance through precipitation development would have reduced the sample size to zero.

3. Methodology

Our objective is to test a method of using radar observables and sounding data to compare the rate at which microphysical processes generate precipitation in clouds ingesting different CCN and GN concentrations. The basis for this method is the parameterization of the production of rainwater in warm clouds found in Berry and Reinhardt (1974). Their Eq. (40a),
i1520-0469-64-10-3695-e1
relates the rate of change of the natural log of the rainwater content to the cloud water content. In this equation, L2 is the rainwater content, L1 is the cloud water content, and b is a constant related to the collection kernel. In the early stages of precipitation development when L2 is small, L1 can be approximated as the total water content and a piecewise linear approximation to (1) is
i1520-0469-64-10-3695-e2
Equation (2) suggests that relevant parameters that can be used to compare precipitation development in populations of clouds are some measure of the liquid water content and some measure of the time it takes for the rainwater to evolve to a given state. Therefore, we hypothesize that, in a population of clouds in which entrainment is similar, differences in the rate of precipitation development caused by variations in CCN and GN may be detected by examining a characteristic total water content (CTWC) and a characteristic time (CT) as determined from radar and sounding data. Our goal in this paper is to determine a CT and CTWC for each cloud for which we have observed the complete radar lifetime. We will then statistically compare cloud populations that we hypothesize have different CCN and/or GN characteristics where each cloud in the dataset is represented by its CT and CTWC. A statistical comparison will be used to determine if the cloud populations are distinct.

a. Characteristic time for precipitation development

In this paper, we will define the CT for precipitation development as the length of time between the first occurrences of two X-band reflectivity values. In Florida, developing cloud-base echoes are nearly always mingled to some extent with the top of the clear-air boundary layer echo, which is dominated by backscattering from insects (Wilson et al. 1994). At and above the top of the well-mixed boundary layer, Rayleigh scattering dominates at X-band. In our analysis the first occurrence of the −5 dBZ X-band radar reflectivity could be clearly detected. The time that this value was first observed was chosen as the beginning of the CT interval.

The end of the CT interval was chosen so that the time interval was sufficiently long to permit differences in precipitation development to be brought out in the statistical analyses, while still meeting the constraint that the precipitation content of the clouds at the end of the time interval was small. An aid in determining when precipitation develops is the appearance of negatively sloped reflectivity contours in the time–height cross sections (e.g., Fig. 1b). Negatively sloped contour lines in these time–height cross sections provide information on the echo descent rate, which can roughly be associated with the bulk fall velocity of raindrops (Knight et al. 2002; see the diagram in the upper right corner of Fig. 1b). The panels in Fig. 2 show the mean bulk fall velocities from all 38 clouds (dots), obtained from the slope of each contour line, as a function of the first occurrence of progressively larger reflectivity values. The standard deviations are depicted as horizontal lines. The fall velocity of 6.5 m s−1, corresponding to the terminal fall velocity of a 2-mm diameter raindrop falling at mean sea level pressure, is indicated by the vertical lines in Fig. 2. The mean reflectivity-weighted fall velocities exceeded 6.5 m s−1 at one contour level at the first occurrence of 7.5 dBZ and three contour levels at the first occurrence of 10 dBZ. We therefore chose 7.5 dBZ as the upper reflectivity value as the end of the CT interval. As we shall show, the statistical analyses that follow indicate that this was a reasonable choice.

b. Cloud liquid water content

The cloud liquid water content is fundamental in controlling the rate at which the warm rain process proceeds. It is necessary to choose a CTWC that is physically related to CT. The height at which the CTWC is chosen must also be related to the height at which a reflectivity threshold is observed within the CT interval. We chose to use the adiabatic liquid water content at the altitude of the first occurrence of the 7.5-dBZ reflectivity value, which is the end of the CT interval. This choice is arbitrary in the sense that we could have chosen any fraction of the adiabatic value and the results of the statistical analysis would not change. However, it is important to remember that the test is only valid if the entire population of clouds has similar entrainment characteristics.

Note that the updraft speed, which was not measured, is not necessary to characterize the evolution of precipitation using the technique described above. This is because the analysis implicitly includes vertical velocity by considering the time between the occurrence of two Rayleigh reflectivities and the CTWC, which is only a function of height above cloud base (given knowledge of cloud base pressure and temperature). Thus, the analysis can be viewed as considering a characteristic height and time, and thus a characteristic vertical velocity.

Cloud-base pressure and temperature were calculated for each day from either the project or aircraft sounding closest in time to the cloud development. The CTWC was calculated using an adiabatic ascent from the lifting condensation level (LCL; e.g., Fig. 3). Temperature and pressure at the LCL on all days used in this analysis varied from 20.3° to 24.3°C and 908.0 to 979.5 mb, respectively (Table 1). Table 1 also shows the mean and standard deviation of the wind speed and relative humidity within the cloud layer for each day in which clouds with sufficient radar history for our analysis were found. The cloud layer was defined here as the layer between cloud base and the altitude of the first occurrence of the 7.5-dBZ X-band radar reflectivity. To ensure that differences in the radar cloud evolution are not related to variations in the entrainment rate or surface forcing, we provide an analysis of the effects of variations in the environmental relative humidity, wind, and the daily heating cycle in section 4.

c. Designation of cloud populations

For the aircraft-supported portion of the project, HY01 determined the average CCN concentrations (activated at 1% supersaturation) and average cloud droplet concentrations from each flight of the NCAR C-130 aircraft. Their analysis, which is reproduced in Fig. 4, showed that three dominant periods occurred, two near either end of the aircraft period of the project when CCN concentrations characteristic of continental clouds were present (1411 ± 388 cm−3), and a long period in the middle when modified maritime CCN concentrations were observed (369 ± 142 cm−3). According to HY01, the continental periods were associated with boundary layer winds that were primarily westerly (offshore of the east coast of the Florida peninsula), while the “maritime” conditions were associated primarily with easterly (onshore) flow. The wind measurements used by HY01 were based on twice daily (1200 and 0000 UTC) soundings taken about 65 km west of the shoreline. Since only a portion of the clouds studied in this paper occurred during the period of aircraft measurements, and none were directly sampled by aircraft, it was necessary to use other approaches that serve as reasonable proxies for categorizing clouds as either maritime or continental.

Our approach is based on work by Sax and Hudson (1981, hereafter SH81), and Hudson and Li (1995). SH81 measured CCN and Aitken nuclei concentrations across the Florida peninsula during the Florida Area Cumulus Experiment. Figure 5 shows CCN and Aitken nuclei concentrations at cloud-base level measured on a east–west transect across the Florida peninsula by the National Oceanic and Atmospheric Administration (NOAA) DC-6 on 21 July 1975. According to SH81, the flow at each coast at the time of the measurements had an onshore component. The most remarkable aspect of these measurements is the sharp gradient in CCN and Aitken nuclei concentrations that occurred on both coasts. The concentration of Aitken nuclei, which were measured at a finer resolution, jumped by about a factor of 5 as the aircraft crossed each coast. CCN concentrations between successive measurements jumped over an order of magnitude at the east coast and by a factor of 2 at the west coast. SH81 noted that changes in CCN concentrations from low values over water to very high concentrations over land surfaces in Florida were typically abrupt with onshore flow. During the daytime in summer, both coasts typically experience a sea breeze circulation, so that clouds forming over the ocean are likely to ingest maritime air with low CCN concentrations. As noted by SH81, the Florida coast is a prodigious source of aerosol and CCN, due to both natural and anthropogenic sources. As a result, maritime air masses, as they move over land, quickly acquire continental characteristics, and clouds forming over the Florida peninsula rapidly take on a continental character (HY01). In contrast, when continental air masses move over the ocean, the air retains its aerosol and does not undergo a rapid transition from continental to maritime (Hudson and Li 1995). However once air moves over the ocean it can also begin to acquire GN produced by wave action at the ocean surface.

Based on these studies, we segregated the radar data into four groups based on the location of clouds (onshore or offshore) and the wind direction (over land or ocean). We also assumed that clouds forming over the ocean during onshore flow will have maritime characteristics (low CCN, high GN), and clouds forming over land in offshore flow will have continental characteristics (high CCN, low GN). We further assumed that air over land with onshore flow quickly acquires higher CCN concentrations but retains higher GN concentrations, so that clouds forming over land will be influenced by both high CCN and high GN concentrations. For clouds forming over the ocean with offshore flow we assumed that the air will retain its CCN but acquire GN as it passes over the ocean surface, so that clouds forming over the ocean will be influenced by both high CCN and high GN concentrations. In this paper, we tested the hypothesis that the CT and CTWC pairs determined from each analyzed cloud were distinct for these different cloud populations because of the different aerosol properties.

Each analyzed cloud was tracked to determine whether the cloud moved in an offshore direction (160° to 340°; e.g., Fig. 6a), an onshore direction (from 340° to 160°; e.g., Fig. 6b), or parallel to the shoreline. Clouds that moved parallel to the shoreline (seven clouds) were ignored in the analysis. The cloud motion was also compared with the surface to 700-mb winds from the nearest project sounding to confirm that the wind profile was consistent with the cloud track. Finally, the cloud tracking method was compared with the data from HY01 for the period where both datasets were available. The results of this analysis appear on Fig. 4. With one exception, 10 August 1995, the two classifications agreed. HY01 noted that the transition between the continental and maritime regime was gradual and was different at various altitudes at any given time. The 10 August case is near a transition time and may represent one of the more difficult cases to classify.

4. Results

Figures 7 and 8 show two-dimensional diagrams of CTWC versus CT for the 38 clouds that featured radar histories suitable for this analysis. The table in the appendix lists important characteristics for each of these clouds. Each point represents one cloud in which the X-band radar reflectivity was tracked from the first occurrence of −5 to 7.5 dBZ. In each panel, the clouds are segregated into two populations. We then used the nonparametric multiresponse permutation procedure developed by Mielke et al. (1981) to examine possible differences between these two populations. This procedure uses a resampling technique to determine the level of significance of data clustering in datasets for which the underlying distribution is unknown (e.g., the data are not normally distributed). Table 2 summarizes the results of this statistical analysis for each panel in Figs. 7 and 8. In each case we tested the null hypothesis that the observed clustering of the data was a chance random event; that is, the average distance between members of each observed population was not significantly different from all other possible groupings of the same data.

Before we examine the effects of GN and CCN on the radar evolution of the clouds, it is important to show that there were no significant environmental effects that could potentially skew the analysis. Two possible first-order effects that may cause variations in the radar reflectivity evolution are entrainment and forcing (in particular, solar heating). To ensure that precipitation development in the sampled clouds was not biased by variations in entrainment, we examined parameters related to the moisture distribution and shear. Specifically, we segregated the cloud population according to the mean relative humidity and the standard deviation of the wind speed over the cloud layer. The standard deviation of the wind speed was used instead of the vertical wind shear because the vertical wind speed profile varied from case to case with the maximum wind speed occurring at different levels within the cloud layer.

Figure 7a shows the distribution of CT and CTWC for clouds growing in an environment with mean relative humidities >75% compared to <75%. Table 2, column 2, shows the probability that the two populations of clouds have significantly different characteristic behavior, that is they separate into disjoint populations. The probability value (p value) for the data depicted in Fig. 7a is 0.3133, indicating that the two populations cannot be distinguished on the basis of environmental moisture variation. Columns 3 and 4 of Table 2 show the median CT and CTWC for the two populations, respectively. The difference between the median CTs for the two populations was 66 s and for the CTWCs was 0.6 g m−3. These values become important when we compare them later to similar values for CCN and GN effects.

Figure 7b shows the distribution of CT and CTWC for clouds growing in environments where the standard deviation of the wind speed within the cloud layer exceeded 1 m s−1 compared to <1 m s−1. Table 2, column 2, shows that these two populations of clouds are not statistically disjoint (p = 0.3321). However, the median difference in the CT for the two populations was larger (165 s) compared to the relative humidity analysis. This was also true for the difference in median CTWC (1.2 g m−3). Taken together these data suggest that there were probably slight variations in the evolution of the sample clouds due to variations in entrainment. However, these variations were not sufficiently large to lead to statistically significant partitions in the cloud populations.

It is possible that variations in the evolution of the clouds may be related to the magnitude of the forcing. For example, clouds forming early in the day may be less buoyant than clouds forming in midafternoon. To test this possibility for the clouds used in this study, we partitioned the data into two groups, those clouds that formed prior to local noon (1600 UTC) and those clouds forming after local noon. Figure 7c shows the distribution of CT and CTWC for these two cloud populations. The p value for the data depicted in this panel is 0.8560 (Table 2) indicating that the probability that these two cloud populations are disjoint is very low. The lack of significant differences for the partitions of the data depicted in Fig. 7 primarily reflect the fact that the environment along the Florida east coast during SCMS did not vary significantly on the days and times when these clouds were sampled.

We consider now the maritime versus continental nature of these clouds using the methods for segregating the clouds described in the previous section. Figure 8 and Table 2 summarize the results. Figure 8a segregates a subset of the 38 clouds into two populations, those that formed over the ocean during onshore flow (truly maritime air) and those that formed over land with onshore flow (modified maritime air). Based on previous arguments we might expect the air ingested into these clouds to have different CCN concentrations but similar GN concentrations. In clouds with low CCN concentrations, the available liquid water is distributed among a smaller number of drops, so that individual drops will be larger and the probability that collision and coalescence will occur sooner is greater. In clouds with high CCN concentrations the opposite is true.

The p value for the data depicted in Fig. 8a was 0.0051, indicating that the likelihood that the two partitions are disjoint is very high. The difference in the median CT for the two populations was 237 s and for the median CTWC, 3.1 g m−3. These values are both much larger than the differences quoted for the partitions in Fig. 7 indicating that variations in CCN dominated potential environmental influences on CT and CTWC. The four truly maritime clouds in the sample have distinctly smaller CT and CTWC implying that these clouds reached the reflectivity threshold at a lower altitude above cloud base and produced precipitation more quickly than the corresponding modified maritime clouds. This test provides direct support for the conclusions of HY01 that variations in the CCN concentration have a strong influence on the rate at which precipitation forms in the SCMS clouds. We can also examine the effect of GN by segregating the dataset in a different manner.

Figure 8b shows a second subset of the 38 clouds segregated into two different populations, those that formed over land during offshore flow (truly continental air) and those that formed over the ocean with offshore flow (modified continental air). Based on previous arguments we might expect the air ingested into these clouds to have similar CCN concentrations but different GN concentrations. The p value for the data depicted in Fig. 8b was 0.2330, indicating that the likelihood that the two partitions are disjoint is low. This particular subset of data (15 clouds) was the smallest used in any of these analyses, which may have contributed to the relatively large p value. We note that there was a large difference in the median CTWC (3.4 g m−3) supporting the idea that the GN might have influenced the cloud evolution. However the median CT was greater for clouds developing offshore (589 s) compared to onshore (418 s), counter to the behavior expected if GN were accelerating the precipitation process. These data provide further support for the conclusions of HY01.

Figure 9 shows the entire population of 38 clouds segregated into the four groups discussed previously in Fig. 8. With few exceptions, these data show a gradual transition from lower left corner of the diagram to the upper right corner. The truly maritime clouds have, on average, the lowest CTWC and CT implying that these clouds reached the characteristic reflectivity threshold for precipitation formation at lower altitudes above cloud base in shorter times. The transitional clouds (modified maritime and modified continental) occupy a band on the diagram extending from low CT and high CTWC to high CT and low CTWC. The modified continental clouds on average have low CTWC and high CT while the modified maritime clouds on average have higher CTWC and a wider range of CT. Four clouds occupying the upper right portion of the diagram are all truly continental clouds that formed over land with offshore flow. These clouds had to rise to a high altitude above cloud base to achieve the threshold radar reflectivity. However there are several truly continental clouds that occupy other portions of the diagram. These clouds deserve special consideration.

Truly continental clouds that occupy other portions of the diagram occurred on 10 August. As noted in the previous section, 10 August was the day in which our cloud tracking method disagreed with HY01’s classification (Fig. 4). The disagreement reflects the fact that the wind direction varied during the general period of observation on 10 August. For this reason the type of air mass ingested into specific clouds on 10 August had a larger degree of uncertainty. This uncertainty is reflected in Fig. 9. Clouds denoted as 10 August appear along a line extending from low CTWC and CT to high CTWC and CT. These data are consistent with an airmass transition on this particular day.

5. Conclusions

In this paper, we developed a method of using radar data to discern microphysical differences between cloud populations. The strength of the method is that a comparison can be accomplished using easily measurable quantities. We applied this method to examine whether the X-band radar reflectivity evolution of clouds observed during the Small Cumulus Microphysics Study supports the evidence presented by Hudson and Yum (2001) that distinct differences in precipitation development can be associated with the Florida clouds’ maritime or continental characteristics. For this study we examined the entire SCMS radar dataset and identified 38 clouds in which a sufficient radar history of the cloud from its earliest detection through precipitation was clearly documented, and the cloud could be clearly classified as either moving in an onshore or offshore direction. Specifically, we investigated whether significant differences in precipitation development could be detected between maritime and continental cloud populations using a characteristic time and total water content derived from the evolution of the X-band radar reflectivity field and sounding data. Since CCN and GN measurements were not available for the specific clouds used in this study, proxies were used to partition the clouds into four groups based on the cloud location and direction of movement. Specifically, we assumed that clouds forming over the ocean during onshore flow had maritime characteristics (group 1: low CCN, high GN), clouds forming over land during onshore flow would have modified maritime characteristics (group 2: high CCN, high GN), clouds forming over land during offshore flow would have continental characteristics (group 3: high CCN, low GN), and clouds forming over the ocean during offshore flow would have modified continental characteristics (group 4: high CCN, high GN). We then statistically compared these populations using the nonparametric multiresponse permutation procedure developed by Mielke et al. (1981). A comparison of groups 1 and 2 provides a test of the role of CCN concentrations in precipitation development in these cloud populations. A comparison of groups 3 and 4 provides a test of the role of GN concentrations in precipitation development in these cloud populations.

The two cloud populations that were disjoint at a statistically significant level were groups 1 and 2. For these groups, whose comparison provides a test of the role of CCN concentrations, the analysis showed that the median characteristic total water content of the truly maritime clouds (group 1) was about half that of the modified maritime clouds (group 2). The characteristic time was about 60% smaller for the truly maritime clouds. Thus the characteristic reflectivity threshold for precipitation development was reached at a much lower altitude above cloud base in a much faster time in the truly maritime clouds. This result supports the conclusions of Hudson and Yum (2001), which were based on analysis of aircraft data.

The partition of the data that compared groups 3 and 4 provided a test of the role of GN in precipitation development. No firm conclusions could be drawn from the comparison since the results were not statistically significant. The role of GN in these clouds was ambiguous. Although the median characteristic total water content was about half for group 4 (high GN), the characteristic time was about 40% greater compared to group 3 (low GN). Thus we cannot determine if GN significantly influenced the precipitation process in the Florida cumulus clouds investigated in this study.

A weakness of this study is the small sample of clouds on which it is based, a consequence of the radar scanning strategy employed during SCMS. Future studies of small cumulus in conditions similar to the SCMS experiment should focus on documenting the entire evolution of clouds from the earliest radar echo through precipitation. Also, the use of dual-polarization radar capability would allow a more complete description of the precipitation development, which would help to further understand the roles of CCN and GN in the warm rain process.

Acknowledgments

This work was supported by the National Science Foundation under the research Grant ATM-0121517. Opinions expressed are those of the authors and not necessary those of the NSF. We wish to thank Neil Laird for providing the liquid water content calculations used in this study. We also wish to thank Dr. James Hudson and two anonymous reviewers for their insightful comments.

REFERENCES

  • Albrecht, B. A., 1989: Aerosols, cloud microphysics, and fractional cloudiness. Science, 245 , 12271230.

  • Battan, L. J., 1953: Observations on the formation and spread of precipitation in convective clouds. J. Meteor., 10 , 311324.

  • Berry, E. X., , and R. L. Reinhardt, 1974: An analysis of cloud drop growth by collection. Part IV: A new parameterization. J. Atmos. Sci., 31 , 21272135.

    • Search Google Scholar
    • Export Citation
  • Caylor, I. J., , and A. J. Illingworth, 1987: Radar observations and modelling of warm rain initiation. Quart. J. Roy. Meteor. Soc., 113 , 11711191.

    • Search Google Scholar
    • Export Citation
  • Hudson, J. G., 1993: Cloud condensation nuclei. J. Appl. Meteor., 32 , 596607.

  • Hudson, J. G., , and H. Li, 1995: Microphysical contrasts in Atlantic stratus. J. Atmos. Sci., 52 , 30313040.

  • Hudson, J. G., , and S. S. Yum, 2001: Maritime–continental drizzle contrasts in small cumuli. J. Atmos. Sci., 58 , 915926.

  • Illingworth, A. J., 1988: The formation of rain in convective clouds. Nature, 336 , 754756.

  • Illingworth, A. J., , J. W. Goddard, , and S. M. Cherry, 1987: Polarization radar studies of precipitation development in convective storms. Quart. J. Roy. Meteor. Soc., 113 , 469489.

    • Search Google Scholar
    • Export Citation
  • Keeler, R. J., , P. H. Herzegh, , and C. F. Frush, 1984: Measurements of CP-2 co-polar and cross-polar antenna illumination functions at S- and X-band. Preprints, 22d Conf. on Radar Meteorology, Zurich, Switzerland, Amer. Meteor. Soc., 287–291.

  • Knight, C. A., , and L. J. Miller, 1998: Early radar echoes from small, warm cumulus: Bragg and hydrometeor scattering. J. Atmos. Sci., 55 , 29742992.

    • Search Google Scholar
    • Export Citation
  • Knight, C. A., , J. Vivekanandan, , and S. G. Lasher-Trapp, 2002: First radar echoes and the early ZDR history of Florida cumulus. J. Atmos. Sci., 59 , 14541472.

    • Search Google Scholar
    • Export Citation
  • Mielke Jr., P. W., , K. J. Berry, , and G. W. Brier, 1981: Application of multi-response permutation procedures for examining seasonal changes in monthly mean sea-level pressure patterns. Mon. Wea. Rev., 109 , 120126.

    • Search Google Scholar
    • Export Citation
  • Sax, R. I., , and J. G. Hudson, 1981: Continentality of the south Florida summertime CCN aerosol. J. Atmos. Sci., 38 , 14671479.

  • Wilson, J. W., , T. M. Weckwerth, , J. Vivekanandan, , R. M. Wakimoto, , and R. W. Russell, 1994: Boundary layer clear-air radar echoes: Origin of echoes and accuracy of derived winds. J. Atmos. Oceanic Technol., 11 , 11841206.

    • Search Google Scholar
    • Export Citation

APPENDIX

Characteristics of Clouds Used in Analysis

i1520-0469-64-10-3695-ta01

Fig. 1.
Fig. 1.

(a) A sequence of RHI scans through a cloud’s maximum X-band radar reflectivity. The cloud developed on 5 Aug 1995, 20 km east-southeast of the radar. The 0-dBZ contour line is highlighted in a thin black line, and maximum and minimum heights of this contour line are indicated. (b) Corresponding time–height cross section showing the cloud’s temporal evolution. Negatively slanted reflectivity isolines are related to the bulk total fall velocity (terminal velocity plus vertical air movement) of raindrops. The relationship between the slope of the reflectivity contours and bulk fall velocities are plotted as black lines. CT indicates the characteristic time for precipitation formation as defined in the text.

Citation: Journal of the Atmospheric Sciences 64, 10; 10.1175/JAS3961.1

Fig. 2.
Fig. 2.

Mean bulk fall velocities (dots) obtained from the slope of each contour line at the time of the first occurrence of the (a) −2.5-, (b) 0-, (c) 2.5-dBZ, etc., contour. The standard deviations are depicted as horizontal lines. The solid vertical line indicates a fall velocity of 6.5 m s−1, which corresponds to a drop diameter of 2 mm falling at sea level pressure.

Citation: Journal of the Atmospheric Sciences 64, 10; 10.1175/JAS3961.1

Fig. 3.
Fig. 3.

Profile of the adiabatic liquid water content (LWC) on 5 Aug. This figure illustrates how the CTWC for the cloud depicted in Fig. 1 was determined.

Citation: Journal of the Atmospheric Sciences 64, 10; 10.1175/JAS3961.1

Fig. 4.
Fig. 4.

Chronology of the SCMS project showing flight averaged CCN concentrations (dots) activated at 1% supersaturation and averaged cloud droplet concentrations (dots) (diameter 2 to 50 μm) with their standard deviations depicted as vertical solid lines (Hudson and Yum 2001). Our classification of air masses based on cloud motion is compared to the classification provided by Hudson and Yum.

Citation: Journal of the Atmospheric Sciences 64, 10; 10.1175/JAS3961.1

Fig. 5.
Fig. 5.

Cross-peninsula profile of CCN and Aitken nuclei concentrations acquired at cloud-base level on 21 Jul 1975 with onshore flow on both coasts (from Sax and Hudson 1981).

Citation: Journal of the Atmospheric Sciences 64, 10; 10.1175/JAS3961.1

Fig. 6.
Fig. 6.

Horizontal cloud movements (black arrows), obtained by tracking radar reflectivity echoes, indicate the prevailing wind direction: (a) Day with westerly (offshore) winds and clouds developing over land and (b) day with southeasterly (onshore) winds and clouds developing over the ocean. The location of the radar is indicated as black + sign, and the 50-km range as a black circle.

Citation: Journal of the Atmospheric Sciences 64, 10; 10.1175/JAS3961.1

Fig. 7.
Fig. 7.

Scatterplot of CT vs CTWC. (a) Dots indicate clouds that developed during days with mean relative humidities >75% within the cloud layer; × symbols indicate clouds that developed during days with mean relative humidities <75% within the cloud layer. (b) Dots indicate clouds that developed during days with standard deviations of the mean wind speed within the cloud layer <1.0 m s−1; × symbols indicate clouds that developed during days with standard deviations of the mean wind speed within the cloud layer >1.0 m s−1. (c) Dots indicate clouds that developed before local noon (<1600 UTC); × symbols indicate clouds that developed after local noon (>1600 UTC).

Citation: Journal of the Atmospheric Sciences 64, 10; 10.1175/JAS3961.1

Fig. 8.
Fig. 8.

Scatterplot of CT vs CTWC. (a) Dots indicate onshore moving clouds that developed over land; × symbols indicate onshore moving clouds that developed over the ocean. (b) Dots indicate offshore moving clouds that developed over land; × symbols indicate offshore moving clouds that developed over the ocean.

Citation: Journal of the Atmospheric Sciences 64, 10; 10.1175/JAS3961.1

Fig. 9.
Fig. 9.

Scatterplot of CT vs CTWC for all 38 clouds. Symbols indicate the different growth location and gray shades indicate the cloud movement. Data points from 10 Aug are noted.

Citation: Journal of the Atmospheric Sciences 64, 10; 10.1175/JAS3961.1

Table 1.

Environmental conditions on days where clouds with sufficient radar history were found.

Table 1.
Table 2.

Median CT and CTWC and the calculated probability (p value) that the data partitions depicted in Figs. 7 and 8 are not disjoint.

Table 2.
Save