cussed earlier, Doppler techniques may be used to deter-
5.3.1.5.3 Multiple-band retrieval of liquid-water con-
mine droplet size from the fall speeds, but the radar must
tent. Considerations for multiple-wavelength radars
attain a high angle scan, near zenith or nadir, to obtain a
include the use of more than two wavelengths and rela-
tionships beyond differential attenuation. For example,
reliable velocity measurement. Since the range of drop
Jameson (1994) develops theory necessary to measure
speeds in the drop sizes of interest is small, use of Dop-
rainfall rate, rain water content, and mass-weighted mean
pler techniques may be difficult. Cloud drops all fall at
less than 27 cm s1, and drizzle falls from about 27 cm
drop diameter from satellite and airborne radars using
s1 to 2.06 m s1 (Rogers and Yau 1989). Larger drops
multiple-wavelength radars. He claims that simple differ-
ential attenuation can be used to determine water content.
fall faster.
However, three wavelengths--38, 25, and 13 GHz--are
Polarization relies upon the oblateness of larger drops
necessary to measure rain rate, rain water content, and
as they fall and distort in shape due to aerodynamic drag.
Since drops smaller than 300 m in diameter are typi-
mass-weighted mean drop diameter. In his summary,
Jameson states that two frequencies are needed to meas-
cally spherical, drop shape cannot be used to determine
ure one parameter, and the measurement of two or three
droplet size in the cloud drop range (Pruppacher and
parameters requires at least three, and ideally four, fre-
Klett 1997). Depending upon the wavelengths, especially
quencies. Some of these concepts might be applied to
for multiwavelength radars, it may also be possible to
measuring icing potential. In a somewhat different
use Rayleigh vs. Mie scattering to sort drops by size.
approach, Srivastava and Tian (1996) theorize that two
5.3.2.1 Doppler radar techniques. Thomson and List
radars of the same or nearly the same wavelength, but
(1996) developed a new method to determine rainfall
physically located apart, could be used to improve com-
drop-size spectra with a vertically pointing X-band Dop-
putations of rainfall through simultaneous use of each
pler radar. Errors in raindrop fall speeds with Doppler
radar's attenuation and reflectivity.
radar occur from vertical wind, turbulence, pressure
A neural-network-based retrieval technique has been
dependence of terminal fall speed, and deviations from
developed at Quadrant Engineering and at the Univer-
Rayleigh scattering. Disdrometer measurements of drop
sity of Massachusetts based upon backscatter informa-
size during radar measurements in the Canadian Atlantic
tion from three radar bands: X, Ka, and W (Mead at al.
Storms Project indicated that the power spectrum of the
1998, Koenig et al. 1999). A neural network was trained
raindrop velocities was related to the drop-size spectra.
to estimate cloud temperature, liquid-water content, and
Vertical wind effects were removed, and the drop spec-
drop mean volume diameters (MeanVD) and mean radar
trum was calculated from reflectivity. Good agreement
reflectivity diameters (MZD) from the backscatter power
between measurements and radar calculations of drop
measurements from one, two, and three bands. Range
spectra were found in two test cases.
resolution was 2 km, and cloud parameters were syn-
Gossard (1994) proposed a method for extracting
thetically created for a wide range of conditions found
cloud droplet-size spectra information using Doppler
in precipitating and nonprecipitating stratiform and
radar in the Ka band. Doppler radar typically cannot
cumiliform clouds. The neural net was trained with
detect cloud droplet spectra from fall speed because the
10,000 cases and tested with 200 cases. Temperature
settling velocity of cloud droplets allows them to be car-
retrievals were not theoretically possible because of radar
ried by updrafts and downdrafts. Gossard developed a
noise. However, with three radar bands, liquid water was
technique, tested on a long-wavelength wind profiler
retrieved with less than 0.17 g m3 error, and MeanVD
Doppler radar, that allows extraction of the cloud drop-
and MZD with less than 16% error. The neural net is
size spectrum by measuring the spectrum shape parame-
currently being evaluated with actual cloud information
ter independently of updrafts and downdrafts. He pro-
to determine prediction accuracy.
poses that the method is effective in detecting drop
growth to the drizzle-size range in stratus clouds. The
5.3.2 Detection of mixed-phase conditions
method cannot be used in precipitation because the rain-
and drop-size spectra
drops overwhelm the cloud signal, and there is too much
In addition to cloud liquid-water content, detection
error in the technique to determine cloud liquid-water
of icing conditions requires information about the loca-
content, even with no precipitation. The technique may
tion of liquid water vs. ice particles, and about the size
be applicable at airports with ground-based radars, and
distribution of water drops. The latter is particularly
it may be useful for airborne radars if high-elevation
important for detecting whether drops are within the
scanning is performed on clouds above and below the
drizzle or raindrop size ranges.
aircraft. Gossard et al. (1997) expand on the technique
The technique most useful for detecting some of these
and indicate that errors in liquid-water content are pos-
parameters, but especially drop size, depends upon
sible within a factor of two and that relative liquid water
whether the radar is ground-based or airborne. As dis-
and rainfall flux through clouds should be accurate.
37
To Contents
Previous Page
