356
Eugene E. Clothiaux, Thomas P. Ackerman, David M. Babb
cloud base histogram (Figure 13.21a) shows distinct maxima in the upper troposphere around
8 km and in the lower troposphere around 3 km, with a pronounced minima in between.
The former are, of course, cirrus clouds, while the latter is a combination of boundary layer
and deep convective clouds. Due to the extreme sensitivity of millimeter-wave radar to small
particles, they often obtain significant return from boundary layer aerosol, particularly under
humid conditions. These radars are even more sensitive to the larger precipitation drops,
making it difficult to devise an algorithm that can automatically differentiate between the
visual cloud base and the constituents below cloud base. For radiative and microphysical
purposes, we generally want to know the visual cloud base, i.e., the base of the cloud droplet
distribution, rather than the precipitation base. In our ongoing research we are exploring the
use of concurrent ceilometer or lidar cloud base data to identify periods of precipitation, thereby
allowing us to correct the radar cloud base statistics so that they contain only information about
clouds and not the precipitation emanating from them. Another possibility for inferring the
cloud base height in the presence of drizzle is bimodality in the radar Doppler spectra, where
one mode corresponds to the cloud particle returns and the other to the drizzle drop signals.
Cloud top heights are generally much easier to infer from radar reflectivity time-height cross
sections. The cloud top distribution from FIRE II (Figure 13.21 b) shows a distinct maximum
around 9 km with a secondary maxima below about 6 km. The upper maximum is largely
cirrus but includes a few occurrences of deep convective clouds.
The cloud base and top data can be differenced to produce frequency distributions of cloud
thickness (Figure 13.21c). These distributions are dominated by cloud thicknesses on the order
of 0.5 to 3 km. While we expect this result to be robust, there is some exaggeration of the
effect in these histograms because the radar was not operated during periods of the most severe
convection. There are cloud thicknesses that span the complete troposphere (9 to 12 km),
indicating that some deep convective clouds were sampled.
A unique aspect of cloud radar is its ability to detect multiple cloud layers. Using the cloud
mask as a starting point, we can count the number of cloud boundary bottoms and tops profileby-profile to estimate the frequency of occurrence of multiple cloud layer systems. During the
FIRE II experiment, clouds occurred approximately 60% of the time (Figure 13.22). (For these
statistics, we did not include layers with cloud tops below 800 m in order to avoid confusion
with aerosol layers, insects, etc.) When clouds did occur over the radar, they were multilayered
33% of the time. Using data from two radars, as well as a lidar, Uttal et al. (1995) concluded
that during the FIRE II observational period when clouds occurred they had a better than 50%
chance of being multilayered. We conclude that multiple layered cloud systems are not rare
and to accurately estimate their vertical distribution cloud radar is essential.
13.8.4 Cloud properties in a vertical atmospheric column
Cloud Liquid Water Path from Passive Radiometry
Given that cloud boundaries can be determined, the most important cloud bulk property that
we would like to obtain from our observations is the column amount of condensed water, either
the liquid water path LW P or ice water path IW P. Measurement of the condensed liquid
water column amount is perhaps most easily inferred from a determination of the atmospheric
optical depth at some wavelength, where the optical depth is defined as
roo N.
T>.. = Jo :Lk>..,;(z)p;(z)dz.
o ;=1
(13.79)
Eugene E. Clothiaux, Thomas P. Ackerman, David M. Babb
cloud base histogram (Figure 13.21a) shows distinct maxima in the upper troposphere around
8 km and in the lower troposphere around 3 km, with a pronounced minima in between.
The former are, of course, cirrus clouds, while the latter is a combination of boundary layer
and deep convective clouds. Due to the extreme sensitivity of millimeter-wave radar to small
particles, they often obtain significant return from boundary layer aerosol, particularly under
humid conditions. These radars are even more sensitive to the larger precipitation drops,
making it difficult to devise an algorithm that can automatically differentiate between the
visual cloud base and the constituents below cloud base. For radiative and microphysical
purposes, we generally want to know the visual cloud base, i.e., the base of the cloud droplet
distribution, rather than the precipitation base. In our ongoing research we are exploring the
use of concurrent ceilometer or lidar cloud base data to identify periods of precipitation, thereby
allowing us to correct the radar cloud base statistics so that they contain only information about
clouds and not the precipitation emanating from them. Another possibility for inferring the
cloud base height in the presence of drizzle is bimodality in the radar Doppler spectra, where
one mode corresponds to the cloud particle returns and the other to the drizzle drop signals.
Cloud top heights are generally much easier to infer from radar reflectivity time-height cross
sections. The cloud top distribution from FIRE II (Figure 13.21 b) shows a distinct maximum
around 9 km with a secondary maxima below about 6 km. The upper maximum is largely
cirrus but includes a few occurrences of deep convective clouds.
The cloud base and top data can be differenced to produce frequency distributions of cloud
thickness (Figure 13.21c). These distributions are dominated by cloud thicknesses on the order
of 0.5 to 3 km. While we expect this result to be robust, there is some exaggeration of the
effect in these histograms because the radar was not operated during periods of the most severe
convection. There are cloud thicknesses that span the complete troposphere (9 to 12 km),
indicating that some deep convective clouds were sampled.
A unique aspect of cloud radar is its ability to detect multiple cloud layers. Using the cloud
mask as a starting point, we can count the number of cloud boundary bottoms and tops profileby-profile to estimate the frequency of occurrence of multiple cloud layer systems. During the
FIRE II experiment, clouds occurred approximately 60% of the time (Figure 13.22). (For these
statistics, we did not include layers with cloud tops below 800 m in order to avoid confusion
with aerosol layers, insects, etc.) When clouds did occur over the radar, they were multilayered
33% of the time. Using data from two radars, as well as a lidar, Uttal et al. (1995) concluded
that during the FIRE II observational period when clouds occurred they had a better than 50%
chance of being multilayered. We conclude that multiple layered cloud systems are not rare
and to accurately estimate their vertical distribution cloud radar is essential.
13.8.4 Cloud properties in a vertical atmospheric column
Cloud Liquid Water Path from Passive Radiometry
Given that cloud boundaries can be determined, the most important cloud bulk property that
we would like to obtain from our observations is the column amount of condensed water, either
the liquid water path LW P or ice water path IW P. Measurement of the condensed liquid
water column amount is perhaps most easily inferred from a determination of the atmospheric
optical depth at some wavelength, where the optical depth is defined as
roo N.
T>.. = Jo :Lk>..,;(z)p;(z)dz.
o ;=1
(13.79)