RADAR / MST and ST Radars and Wind Profilers 1825
RADAR
MST and ST Radars and
Wind Profilers
R F Woodman, Instituto Geofísico del Perú, Lima, Perú
Copyright 2003 Elsevier Science Ltd. All Rights Reserved.
Short History and Definitions
MST stands for mesosphere-stratosphere-tropo-
sphere and qualifies VHF (very high frequency) radar
systems capable of observing all these regions of the
atmosphere. The technique was developed and used
for the first time in the early 1970s at the Jicamarca
Radio Observatory, an incoherent-scatter facility of
the Instituto Geofísico del Perú, located near Lima,
Perú (see Figure 1). The ñame that is now used was
coined a few years later on the occasion of the US
National Academy of Science workshop on the Use of
Radar for Atmospheric Research in the 1980s, held in
Salt Lake City, Utah, in 1978.
To reach mesospheric altitudes, the system must
normally have a very large antenna (or antennas) and a
powerful transmitter. In the case of the Jicamarca
radar, the antenna is a square array of dipoles 300 ni on
a side, and the transmitter has more than 2 MW of
peak power. Af ter the success of the Jicamarca radar in
obtaining radar echoes from these mid-atmospheric
altitudes, other radars were built specifically for this
purpose: the SOUSY radar in the Harz, Germany; the
Poker Fíat radar in Alaska (no longer in operation); the
MU radar in Shigaraki, Japan; and the Gadanki
radar in Tirupati, India. The VHF EISCAT radar, in
Troms0, Norway, although built mainly for incoher-
ent-scatter ionospheric studies, should also be includ-
ed in the list.
To qualify as an MST radar, the system has to be
capable of obtaining echoes from the mesosphere. This
requirement limits the frequency range of these radars
to the lower frequencies of the VHF band, i.e., between
30 and 300 MHz. Smaller radars, using the same
principies and techniques but capable of reaching only
the stratosphere and troposphere, are called ST radars.
Here we can cite as an example the NOAA Tropical
Pacific Profiler Network in the equatorial Pacific, and
the wind profiler network in central United States as
important múltiple radar installations that qualify as
such. These, and even smaller systems capable of
observing only the troposphere, are also called wind
profilers, because of their main use as operational
radars dedicated to measuring the winds aloft in a
continuous manner. At these lower altitudes it is
possible to use frequencies higher than VHF frequen-
cies. Nowadays, many radars designed for other
purposes, including operational air traffic, meteoro-
logical Doppler radars, and ionospheric incoherent-
scatter radars, are also used as ST radars.
Figure 1 Panoramic view of the Jicamarca Radio Observatory and its 300 m x 300 m antenna array. The building houses a powerful
transmitter with approximately 3 MW of peak power. It was the first MST radar and it is still the most powerful.
and ST Radars and Wind Profilers
At polar and near-polar latitudes a very special
mesospheric geophysical phenomenon manifests itseJf
with a greatly enhanced radar reflectivity and permits
the use of smaller VHF systems. These smaller radars,
when operated at these latitudes, could also be
classifíed as MST radars. The Mobile-SOUSY radar
operating in Svalbard, Norway, the Peruvian radar
installed at its Machu-Picchu Antarctic station, and
the Cornell University mobile radar (CUPRI) beJong to
this category. The Jist is not exhaustive and it includes
only some of the early systems. The high-reflectivity
phenomenon was discovered by the Poker Fíat MST
radar and is referred to as PMSE, for polar meso-
spheric summer echoes, with occurrence limited to the
summer months.
Lastly, we are now seeing the proliferation of even
smaller radars that use the same techniques but are
limited to study of the lower troposphere; these are
referred to as boundary layer radars (BLRs).
The Basic Technique
The MSTradar technique involves the illumination of
a target with a pulsed electromagnetic (EM) wave. The
target, in this case the clear atmosphere, scatters a very
small fraction of the energy of the wave in all
directions, includíng the direction to the receíving
antenna, which in most cases is the same antenna used
for transmission. We should underline here 'clear
atmosphere' to differentiate the technique from meteo-
rological radars that depend on hydrometeors (rain,
snow, hail) to act as targets. At reception, the returned
echoes are arnplified and discriminated in range by
sampling them at different delays with respect to the
time at which the transmitted pulse was emitted,
taking advantage of the fínite speed of propagation of
the probing wave (the speed of light). The sampled
signal forms a two-dimensional discretely sampled
complex random-process S(h, t), such as the one
depicted in Figure 2. One of the dimensions is the
sample delay, or range h, and the other is the discrete
time t corresponding to the emission of the transmitter
pulse. Given the practica! statistical índependence of
the process for different h's, it is preferable to think of
the process as múltiple random processes in t, S¡, (t),
one for each h.
The complex nature oíS/,(t) is a consequence of the
coherent nature of the detectíon scheme used. The
magnitude of Sf,(t) is proportional to the amplitude of
the received radío signal, and its phase is that of the
received signal with respect to the transmitted one. ín
practice, the signal is obtained and represented by its
real and imaginary components.
The received signáis are digitized and fed to a digital
processing system. There they are grouped and statis-
S(h,t)
-1
Figure 2 Schemafic representaron of the radar received signal,
Sh(t) = S(h, t), as a two-dimensional random process. The
process is sampled at the altitude of the echoing región, h, and a
time t, corresponding to the time of the transmitted pulse. The
process is complex. Here we are representing only the real
(or imaginary) component.
tically processed as múltiple independent time-series,
one for each range (h = 1,2,... ,N). The statistical
parameters obtained for each series are related to the
properties of interest corresponding to each particular
range, h, as discussed in the following sections.
One characteristic of the MST or ST echoes is the
slowness of the process as compared to the typical
pulse repetition rate, i.e., the sampling time interval.
The corresponding interpulse period is determined by
the máximum altitude (delay) of the región under
study. This period is of the order of a fraction of a
millisecond, whereas the characteristic time of the
process is of the order of a second or more. This
permits the coherent addition of many samples of
almost ídentical valué before any further processing is
done. This is a common pre-processing scheme in all
MST radars and has a twofold advantage: it enhances
the signal-to-noise ratio of the signal prior to its later
processing and, more ímportantly, it reduces the
digital statistical-processing speed requirements by
about two orders of magnitude.
Underlying Principies
The MST radar technique is based on the scattering
properties of minute fluctuations in Índex of refraction
at the altitudes of interest. At tropospheric and
stratospheric altitudes, these fluctuations in refractive
Índex come about as a consequence of fluctuations in
the density of the médium. At mesospheric altitudes,
the médium is ionized and it is the density of free
electrons that determines its Índex of refraction.
Henee, the scattering is produced by fluctuations in
RADAR / MST and ST Radars and Wind Profilers 1827
electrón density. In all cases, the fluctuations in Índex
of refraction, small as they are, must be enhanced well
above the ever-existing thermal fluctuation level. This
enhancement is usually produced directly or indirectly
by turbulence.
At tropospheric heights the MST radar technique,
although it does not exclude fluctuations produced by
the presence of hydrometeors, depends mainly on
clear-air fluctuations of the dielectric properties of the
médium. This dependence on clear-air fluctuations
differentiates the MSTor ST techniques from the older
and better-known meteorological radars, which do
depend on the existence of hydrometeors. The fact that
MST radar technique does not require the presence of
hydrometeors extends the usefulness of the technique
by permitting continuous 24-hour observations of the
stratosphere and troposphere. At mesospheric alti-
tudes the MST radar technique is useful mainly during
(mesospheric) daytime hours when free electrons
exist. At polar latitudes the time is extended during
the summer almost to a full day, depending on the
latitude.
Fluctuations in the Índex of refraction in the
atmosphere are of a stochastic nature; consequently,
the radar echoes they produce are also stochastic. Both
have to be characterized statistically. If we denote the
fluctuating component (zero mean) of the Índex of
refraction of the médium at position x (vectors are
represented by bold letters) and time t by «(x, í), an
important statistical property of the médium is given
by its un-normalized space-time autocorrelation
function, p(r, r), defined in eqn [1].
p(r, -L] = {«(x, í) «(x + r, t + T)) [1]
Here, «(x + r, t + r) denotes the fluctuation in Índex of
refraction measured at a displaced position r and time
lag i. Both displacements refer to the point, x, and
time, í, where «(x, i] is initially measured. The angle
brackets represen! an ensemble average. We have
assumed that the médium is homogeneous within a
large space around x, and stationary within an
observational time centered around í, henee the
independence of p(r, T) on these variables. The statist-
ical properties of the received signal, S¿,(í), as defined
before, scattered from a región of range h, are
characterized in turn by its temporal correlation
function Q,(T). Furthermore, it can be shown that
these two statistical averages are, under simplifying
assumptions, related as in eqn [2].
[2]
where A is a constant of proportionality that depends
on the radar system parameters, including - among
others - the radar power and its antenna gain. The
vector k is called the scattering Bragg wave vector and
stands for the vector difference k¡ — ks. The terms k¡
and ks stand, in turn, for the incident and scattered
electromagnetic wave vectors, respectively. Note that
we have relaxed the assumption of homogeneity of the
médium, allowing p¿(r. T) to vary slowly as a function
of the range h.
In most MST radar systems the transmitter and
receiving antenna are in the same place. Then, the
incident and the scattered wave vectors are aligned
with the line of sight but with opposite signs. In such
a case the Bragg vector k is equal to 2k¡ and has
a magnitude equal to twice the wavenumber corre-
sponding to the radar frequency, co, i.e., k = 2x
2n/l = 2co/c, where c is the speed of light and /, is
the radar wavelength.
The potential of the MST radar technique, as well as
of other radars used to observe the atmosphere, is
based on the simple functional relationship of eqn [2].
Note that the integral operation on p¿(r, T) is its three-
dimensional Fourier transform evaluated at a single
wave vector k The first conclusión we can derive from
eqn [2] is that the radar is sensitive only to one of the
Fourier components of p¿(r, T), that corresponding to
k, and, henee, on the basis of the Wiener-Khinchin
theorem, to the same Fourier component of the
random fluctuations n(x,t). The fluctuations repre-
sented by M(X, í) can be thought of as the Fourier
superposition of many sinusoids of different wave-
lengths. However, the radar will see only one compo-
nent of the fluctuations, ñ(k. í), i.e., the one with a
wavelength equal to half the wavelength of the radar
and a direction aligned with the line of sight (in the
backscatter case) of the radar. Furthermore, the power
of the echo returns, given by C¿,(0), is proportional to
the power spectral density of this fluctuation compo-
nent, {«(k, f) }. Additionally, the temporal dynamics
of Q,(T) depends on the dynamics of this spatial
Fourier component of the fluctuations.
We have characterized above the statistical proper-
ties of the received signáis by the autocorrelation
function, Q,(T), of the time-series they genérate. They
can also be characterized by their frequency power
spectra, Fh(m), since this is the temporal Fourier
transform of Q,(T) (eqn [3]).
Fh(w) = ¡3]
In fact, in most MST radar processing systems, the
frequency spectrum is evaluated directly from the
series it characterizes, taking advantage of the speed of
the fast Fourier transform (FFT) algorithm. The
frequency spectrum is also preferred by many users
1828 RADAR / MST and ST Radars and Wind Profilers
- Noise contribution
Signal power
F(oi)
(arbitrary units)
Ground clutter
Noise level'
-3
(B)
-2 -1 1
co/k
2 3
Figure 3 Statistical properties of the time series at a given
altitude, /?: (A) its complex, amplitude and phase, temporal
autocorrelation function C/,(r); (B) its corresponding frequency
spectrum, Fh(to).
of the technique for discussíng the physical meaning of
the observations. Figure 3 shows schematically the
typical shape of these two functions, Q,(T) and Ff,(co).
We will discuss later the physical significance of their
characterizing parameters.
On the Orígin of the Scattering
Fluctuations
To interpret the MST radar results we need a model of
the physical mechanisms responsible for the fluctua-
tions in Índex of refraction that are responsible for the
radar echoes. In most cases, the small-scale fluctua-
tions are produced by turbulent processes. Acoustic
waves can also produce scattering fluctuations, but
only if excited significantly by artificial means.
Turbulence occurs in the atmosphere as a conse-
quence of static or dynamic instabilities. Under the
presence of a gradient in the undisturbed Índex of
refraction, turbulence will mix pareéis of high índex
into regions of low índex of refraction, and vice versa.
Once these primary fluctuations are formed, smaller
secondary fluctuations are produced as a consequence
of smaller-scale mixing working on the larger scale
fluctuating gradients. In the case of the neutral
atmosphere the índex of refraction depends on the
neutral density, which for a given pressure depends on
temperature.
The required original gradient in Índex of refraction
has to deviate from that produced by the normal static
exponential vertical pressure gradient of the atmos-
phere produced by gravity. One actually requires an
initial gradient in potential temperature (temperature
at a constant reference pressure). Potential tempera-
ture behaves as a passive scalar, i.e., stays ínvariant in a
Lagrangian description of the turbulent motions,
despite the presence of the exponential static pressure
gradient; thus it fluctuates as any other passive scalar
as a consequence of turbulence mixing.
In the mesosphere, the índex of refraction depends
on the electrón density. For turbulence to produce the
necessary fluctuations, we need a gradient in this
density, or more properly in the potential electrón
density (i.e., electrón density at a given reference
pressure). The behavior of electrón density as a passive
scalar can be questioned because of production and
recombination processes, but these can be partially
ignored given their relatively large characteristic times
at these altitudes.
Apart from the fluctuations produced directly by
turbulent mixing, other indirect effects may also
produce scattering structures. Here we need to con-
sider that turbulence in the atmosphere, with excep-
ción of that produced by tropospheric thermal
convection processes, is stratified. This is a conse-
quence of the fact that dynamic instabilities produce
turbulence only at altitudes where the vertical shear of
horizontal velocities exceeds a critical level. Thus, the
transition región from turbulent to nonturbulent can
be very thin, producing essential discontinuities in the
vertical profile of Índex of refraction with a large
horizontal extent. If the discontinuity has a horizontal
dimensión larger than a Fresnel radius, it can produce
a pardal reflecting structure with almost specular
properties. Alternatively, these structures can be
incorporated in the formalism expressed by eqns [1]
and [2] above as very fíat, disklike, local correlation
functions to which corresponds an almost unidirec-
tional spatial wave vector (k) spectrum. We then speak
of 'aspect-sensitive' structures with enhanced scatter-
ing properties in the vertical direction.
Furthermore, the transition from turbulent to non-
turbulent is a región where the initial vertical shear is
concentrated. This shear will take an initially isotropic
blob and deform it into an elliptical shape, also
producing almost horizontal aspect-sensitive struc-
tures with similar scattering effects.
RADAR / MST and ST Radars and Wind Profilers 1829
The existence of these aspect-sensitive structures
complicates the interpretation of the radial velocities
measured by the radar since it effectively introduces a
bias in the effective pointing direction of the antenna
beam. On the other hand, the enhancement in reflec-
tivity produced by these aspect-sensitive structures is
used by some ST techniques, by pointing only in the
vertical direction, to reduce the required power and/or
antenna size while maintaining sufficient sensitivity.
Physical Interpretation of the Radar
Echo Characteristics
The frequency spectrum Fh(w) of the radar echo
returns is usually a bell-shaped function, especially
when the radar has sufficient altitude resolution to
sample a locally homogeneous región. As such, the
spectrum can be well characterized by its first three
moments, i.e., its total power or área under the
spectrum, its displacement with respect to zero
frequency (transmitter frequency), and its width. The
contributions of noise and ground clutter, which are
always present, have to be subtracted before the
analysis.
According to the previous discussion, the total
power, Q,(0) =/Ffc(cu) dco, is proportional to
{«(k, í) }, i.e., the spatial spectral power density at
wave vector k, and is a proxy of the total fluctuation
power {«(x, t) ). In fact, under the assumption that the
fluctuations are a consequence of turbulent mixing,
turbulence theory allows us to infer the latter from the
former. We need only to assume k to be within the
inertial subrange of turbulence, i.e., a known spectral
shape, and have an estimate of the outer scale, which in
many cases can be obtained from the measurements
themselves (e.g., layer thickness). Thus the technique
allows us to obtain the altitude and temporal distri-
bution of turbulence intensity in the atmosphere.
Therefore, the MST power measurements have been
used to determine the morphology of turbulence in the
atmosphere. Under the assumption of turbulent mix-
ing, MST radar measurements have also been used to
estimate turbulent diffusivity and the energy dissipa-
tion rate at different altitudes. Regions of high aspect
sensitivity can be discriminated by looking off-vertical
or with interferometer techniques (see below). Aspect
sensitivity can be used, at least in a qualitative fashion,
as an indication of the presence of anisotropic turbu-
lence or as an indication of sharp horizontally strat-
ified structures in the Índex of refraction such as the
ones discussed above.
The second characteristic parameter of the frequen-
cy spectrum is the displacement of the 'center' of the
spectrum with respect to zero frequency (i.e., the
transmitter frequency). As intuitively expected, this is
a Doppler displacement and is related to the radial
velocity of the target. We will show next that this is
indeed the case.
Density, and henee dielectric, fluctuations at the
scale sizes of interest have two main modes of
dynamical behavior - acoustic and diffusive - depend-
ing on whether or not there are pressure fluctuations
associated with them. Acoustic fluctuations propágate
at the speed of sound; diffusive fluctuations do not
propágate, they move with the bulk motion of the
médium in which they are imbedded. Turbulence-
induced fluctuations are diffusive and therefore can be
used as tracers of the bulk motions of the atmosphere.
Acoustic fluctuations are also capable of scattering if
excited at a sufficiently high level. They are the
artificially introduced tracers of the radio acoustic
sounding system (RASS) technique, which we will
briefly mention later.
If a time-varying random fluctuation field does not
show any preferred direction with respect to an
observer, there is no relative velocity between the
observer and the médium. In this case the correspond-
ing space-time autocorrelation function is symmetric
under a direction interchange, i.e., p0(r, T) = p0(—r, T).
A radar observing from this frame of reference will
obtain a real and symmetric spectrum F0(co) =
F0(—co). If the same médium moves with velocity v
with respect to the same observing radar, then
p(r, x) = p0(r - VT, T). It can be shown using eqns [2]
and [3] above that the resulting spectrum would then
be given by eqn [4].
F(a))=F0((o-k-v) [4]
This is the Doppler effect and means that the point of
symmetry of the spectrum would be shifted by a
frequency Q = k • v, proportional to the velocity of the
médium, v, projected along the Bragg wave vector, k.
Since the first moment of the symmetric spectrum
F0(co) is zero, the first moment of the displaced
spectrum, F(a>), is given by the displacement
Q = k • v. In practice, owing to deviations from médi-
um homogeneity within the scattering volume, one
may find that there is no perfect center of symmetry for
either F(a>) or F0(ai). Nevertheless, one can always find
the first moment and use it as a measure of the
(projected) velocity of the médium.
Backscatter radar measures the médium vector
velocity component along the line of sight of the radar,
at all ranges. Figure 4 shows an example of a spectrum
profile obtained with the Arecibo, Puerto Rico,
430 MHz radar in ST mode. The radial velocity profile
can be practically read directly from the graph. By
using at least three different and non-coplanar pointing
1830 RADAR / MST and ST Radars and Wind Profilers
12July1978 19:25AST
ar^
30
o
1
-16-12-8 - 4 0 4 8 12 16
Eastward velocity (m s~1)
Figure 4 Typical rawstatistical data obtained every minute bythe
430 MHz radar of the Arecibo Observatory operating in a ST mode.
The zenith angle for this observation was 15°. Contours are drawn
every 3dB in a logarithmic scale. Signal contours have been
darkened. The center contours correspond to ground clutter. (From
Woodman RF. High altitude resolution stratospheric measurements
with the Arecibo 430-MHz radar. Radio Science 15(2): 417^22,
March-April 1980. Copyright (1980) American Geophysical Union.
Reproduced by permission of American Geophysical Union.)
directions one could determine the full velocity vector
of the médium, including the vertical component,
provided that the three scattering volumes are cióse
enough to sample the same velocity field.
The MST (or ST) technique is the only technique
capable of measuring directly the very small vertical
component of the atmospheric motions. Accuracies of
the order of a few centimeters per second are obtained
for the on-line-of-sight velocities with only a minute of
integration. Another important advantage of the
technique is its high temporal resolution and continu-
ity of observation. The time dependence of full vector
velocity fields can be obtained showing the fastest
temporal fluctuations of buoyancy waves at and below
Brünt-Váissalá periods as well as the longest planetary
wave periods, and beyond.
As far as the horizontal components of the wind are
concerned, there are a variety of radiowave techniques
that can be employed. At least half a dozen different
acronyms have been coined for these methods, but in
reality they are based on only two different principies.
Some can be grouped as techniques based on múltiple
projections of the Doppler shift, as described above.
The others are based on the relationship between the
velocity of the diffraction pattern cast on the ground
by the scattered EM field. These are measured by three
or more non-collinear spaced antennas, that respond
to the horizontal velocity of the médium responsible
for the scattering. Figure 5 depicts velocity parameters
obtained with a BLR using this latter technique.
Another technique belonging to the second group is
the interferometric technique. Here antenna signáis
are processed in pairs. The cross-spectral correlation
of these two received signáis gives additional infor-
mation, including the angle of arrival of the echo its
angular velocity and the angular spread of the scatter-
ing región. But even this grouping is artificial, since it
can be shown that the angular distribution of the
scattered waves (pointing techniques) and the spatial
distribution of the field on the ground (spaced receiver
techniques) form a spatial Fourier transformpair-i.e.,
they are not independent.
For the interpretation of the spectral width, it is
convenient to think of the scattering signal as coming
from different independent subvolumes of the total
scattering volume. The size of the subvolumes should
be much smaller than the total volume but slightly
larger than the Bragg scattering wavelength. Each of
the subvolumes will scatter with a Doppler shift
corresponding to its radial velocity. If we assume the
scattering properties of the médium to be homogene-
ous, the Doppler frequency spectral width would be a
measure of the radial velocity distribution of the
subvolumes. In a turbulent mixing model, the spectral
width is then directly related to the variance of the
RADAR / MST and ST Radars and Wind Profilers 1831
Windbarb field-Largest valué : 24.1 knots
20
•o
c
I
oI
20:00 0:00 4:00 8:00 12:00 16:00
Universal time-12/12/2000 16:30:00 to 13/12/2000 16:30:00
Figure 5 Windbarb velocity field as a function of altitude and time obtained with a VHF boundary layer radar. The vertical component
corresponds to the NS component of the wind, and the horizontal to the EW component of the same. (Courtesy of the Atmospheric Physics
group at the University of Adelaide.)
turbulent velocities and henee a measure of the
turbulent kinetic energy and, through turbulence
theory, is related also to the turbulence dissipation
rate. Figure 6 shows an experimental example of the
three spectral moments, including the spectral width,
as a function of altitude and time obtained by the
ALWIN radar, in Andenes, Norway. In this case, they
correspond to PMSE echoes.
The validity of the above interpretation of the
spectral width depends on the validity of some implicit
assumptions. We have implicitly assumed that the
subvolumes have a frozen structure. In reality there are
smaller-scale motions within the subvolume, even
smaller than a wavelength. The measured spectrum is
actually the convolution of the velocity distribution of
this smaller structure with the distribution of the
subvolume velocities. This assumption has no practi-
cal consequence, since the variance of the larger
structures is much larger than that of the smaller
scales and, in any case, the measured spectra actually
include the contributions of the smaller scales.
We have also implicitly assumed that the average
radial velocities of the subvolumes are all the same and
that the mean Doppler shift is a measure of it. In
practice the antennas have a finite beam width and the
horizontal component of the wind will project differ-
ent average valúes depending on the pointing angle to
the subvolume with respect to the center pointing
direction of the antenna. This effect can be very
important, considering that the horizontal velocities
are very large, of the order of tens of meters per second,
while the turbulent velocities are only of the order of
10 cm s ~1. The effect is known as the beam broaden-
ing effect. Fortunately, unless the beam width is of the
order of ten degrees or more, the beam broadening
effect can be corrected by deconvolving the spectrum
with the known beam broadening effect. To do this
one needs to know the horizontal velocity, which is
measured by the technique.
Other sources of spectral broadening are the exist-
ence of shear and gravity waves. Shear is important
only in oblique pointing directions when the altitude
resolution is not capable of resolving the existence of a
shear in the horizontal velocity (see for instance the
spectrum at 15.2 km in Figure 4). In this case, the
subvolume average radial velocities have systematic
contributions from the different horizontal velocities
at different altitudes. On the other hand, if there is a
concurrent vertical pointing beam, shear broadening
can be used to infer the valué of the very same shear
that produced it, even when the instrument is not
capable of resolving it. Gravity waves can also
contribute to the broadening, but only when the
spectra have been obtained with an integration time
that is longer than a significant fraction of the shortest
(Brünt-Váissálá) periods in the wind fluctuations.
We have mentioned that acoustic wave density
fluctuations are also capable of producing EM wave
scattering. In this case the Doppler shift is given by the
acoustic velocity and henee it can be used to determine
1832 RADAR / MST and ST Radars and Wind Profilers
Signal power
60
50
40
30
20
10
13:00 14:00 15:00 16:00 17:00 18:00
Radial velocity
19:00 20:00
12:00 13:00 14:00 15:00 16:00 17:00 18:00 19:00 20:00
Spectral width
12:00 13:00 14:00 15:00 16:00 17:00 18:00
10 Jul. 2000, Time (UTC)
19:00 20:00
Figure 6 Colour shade display of the power, radial velocity, and spectral width of polar mesospheric summer echoes (PMSE) obtained
with the ALWIN radar in Andenes, Norway. (Courtesy of the Institute for Atmospheric Physics, Kühlungsborn, Germany.)
the temperature of the médium, but the natural
background level of the fluctuations is too weak to
be of practical use: therefore artificial excitation of the
waves by acoustic sources is necessary. The technique
has been used successfully to measure tropospheric
temperature profiles and is known as the radio
acoustic sounding system (RASS).
See also
Boundary Layers: Neutrally Stratified Boundary Layer.
Clear Air Turbulence. Data Analysis: Time Series
Analysis. Dynamic Meteorology: Overview. General
Circulation: Overview. Mesosphere: Polar Summer
Mesopause. Radar: Doppler Radar; Incoherent Scatter
Radar; Precipitation Radar.
Further Reading
Balsley BB (1981) The MST technique - a brief review.
Journal of Atmospheric and Terrestrial Physics 43:
495-509.
Cho JYN and Rottger J (1997) An updated review of polar
mesosphere summer echoes: observations, theory, and
their relationship to noctilucent clouds and sub-visible
aerosols. Journal of Geophysical Research 102:
2001-2020.
Gage KS (1990) Radar observations of the free atmosphere
structure and dynamics. In: Atlas D (ed.) Radar in
Meteorology, pp. 534-565. Boston, MA: American
Meteorological Society.
Hocking WK (1997) Recent advances in radar instrumen-
tation and techniques for studies of the mesosphere,
stratosphere, and troposphere. Radio Science 32:
2241-2270.
RADAR / Precipitaron Radar 1833
Róttger J (1991) MST radar and incoherent scatter radar
contributions to studyíng the middle atmosphere. Journal
of Geomagnetism and Geoelectricity 43: 563-596.
Róttger J and Larsen MF (1990) UHF/VHF radar techniques
for atmospheric research and wind profiler applications.
In: Atlas D (ed.) Radar in Meteorology, pp. 253-281.
Boston, MA: American Meteorological Society.
Woodman RF (1991) A general statistical instrument theory
of atmospheric and ionospheric radars. Journal of Geo-
physical Research 96: 7911-7928.
Woodman RF and Guillen A (1974) Radar observations
of winds and turbulence in the stratosphere and
mesosphere. Journal of the Atmospheric Sciences 31:
493-505.
Precipitation Radar
S E Yuter, University of Washington, Seattle, WA, USA
Copyright 2003 Elsevier Science Ltd. All Rights Reserved.
Introduction
Precipitation radars are widely used to determine the
location, size, and intensity of precipitating storrns.
Ground-based scanning precipitation radars are used
in short-term weather and flood forecasting, and to
estímate the distribution and amount of cumulative
rainfall over a región. The weather services of many
industrial countries have networks of operational
radars that monitor precipitation near population
centers. The output from these operational radar
networks can be combined to provide a picture of the
distribution of precipitation over synoptic-scale re-
gions. Mobile precipitation radar on aircraft provides
pilots with information to navigate safely around
dangerous regions of hail and turbulence. Precipita-
tion radars are also used to map the three-dimensional
structure of storms. Spaceborne precipitation radar on
low-orbit satellite maps the structure and distribution
of precipitation around the globe over periods of
months and years.
The British and Americans first developed weather
radar during World War II. The precipitation radar
transmits a pulse of electromagnetic energy via an
antenna. When the transmitted energy encounters a
reflector, such as a raindrop, part of the transmitted
energy is scattered back toward the antenna, where it
is received and amplified. The time delay between the
original pulse transmission and the receipt of the
backscattered energy is used to deduce the distance to
the reflector. The relationship between the range-
corrected, backscattered, returned power and the
size and number of the reflecting targets is the
physical foundation for interpreting precipitation
radar data.
The frequency of electromagnetic waves f in Hz
(s~ ) is defined as f — C/Á, where c is the speed of light
and / is the wavelength. Radar frequencies are divided
into several bands, as shown in Table 1. The choice of
frequency for precipitation radar is a tradeoff between
the practical constraints of size, weight, cost, and the
relation between the wavelength and the size of the
target hydrometeors. Theoretical considerations favor
the choice of the longer S-band and C-band wave-
lengths for many precipitation applications. However,
use of these longer wavelengths is not always practical.
The beam wídth for circular antennas is proportional
to /./d, where d is the antenna diameter. In comparison
to shorter wavelengths, longer wavelengths necessi-
tate a larger antenna to obtain a focused beam of the
same angular beam width. Larger antennas are heav-
ier, require more powerful motors to rotate them, and
are more expensive than smaller antennas. Shorter X-,
Ku-, and Ka-band wavelengths are often utilized in
mobile precipitation radars deployed on spacecraft,
aircraft, and ships where size and weight are more
constrained as compared to stationary ground-based
radars.
Table 1 Precipitation radar frequencies and wavelengths
(adapted from Skolnik, 1990, and Rinehart, 1991)
Band
designaron
S
C
X
Ku
Ka
Nomina/
frequency (GHz)
2-4
4-8
8-12
12-18
27-10
Nominal
wavelength
(cm)
15-8
8-4
4-2.5
2.5-1.7
1.1-0.75
Applications
Surface-based
radars
Mobile and
surface-
based
radars
Mobile and
surface-
based
radars
Mobile and
spaceborne
radars
Mobile and
spaceborne
radars
1834 RADAR / Precipitation Radar
Precipitation Radar Components
The precipitation radar consists of a transmitter,
receiver, transmit/receive switch, antenna, and display
(Figure 1). In this simplified diagram, the processing of
the electromagnetic signáis into output suitable for
display is included in the display block. A phase
detector may be included to measure Doppler velocity.
The radar transmitter contains a modulator that
switches the transmitter on and off to form discrete
pulses. The radar sends out a pulse and then switches
to the receiver to listen. The range to the targets is
obtained by comparing the time of pulse transmission
to the time the backscattered pulse is received. In
precipitation radars, the pulses are transmitted at a
pulse repetition frequency (PRF) of ~ 300-1300 Hz
and each pulse is order 1 ^ s (10 ~6 s) in duration. The
time between transmitted pulses limits the máximum
range (rmax) the electromagnetic pulse can travel
before the next pulse is transmitted (eqn [1])
[1]2PRF
The receiver detects and amplifies the received signáis
and averages the characteristics of the returned pulses
over defined time periods. Typical peak transmitted
power for an operational precipitation radar is
105-106W. Typical received power is 10~10W. The
transmit/receive switch protects the sensitive receiver
from the powerful transmitter. Without the switch, the
radar transmitter would burn out the receiver. In
practice, the transmit/receive switch is not perfect and
a small amount of transmitted energy leaks into the
receiver.
Radar antennas focus transmitted energy and direct
it along a narrow angular beam. For scanning radars,
Figure 1 Simplified hardware block diagram of a precipitation
radar. The non-Doppler portion of the system yields the range r and
equivalent reflectivity factor (Ze) of the target. Dashed lines
connect parts of the system included in a Doppler radar that
additionally measures the radial velocity of the target (V,).
(Adapted from Houze, 1993.)
this direction is often described in terms of an elevation
angle relative to the ground and an azimuth angle
relative to north. Moving the antenna points the axis
of the beam in different directions, and permits
scanning of two- and three-dimensional regions of
the atmosphere. The antenna shape determines the
radar beam size and shape. The radar energy is
máximum along the center of the beam and decreases
outward with increasing angular width. The beam
width is defined as the angular width where the power
is exactly half the máximum power. Most precipita-
tion radars utilize a circular parabolic antenna for
both transmission and reception.
The main purpose of the display is to distinguish
scatterers at different ranges. A basic scanning radar
display will usually indícate the compass angle and
range to the radar echo ín polar coordinates. Figure 2
shows the range-corrected received power as a func-
tion of range along a single pointing direction of the
antenna. This example ¡Ilústrales that not all the
energy received at the radar is backscattered from
meteorological targets. Transmitter leakage and
ground clutter from nearby nonmeteorological targets
such as trees and buildings are present at cióse ranges
in the display. The signal from a point target such as a
radio tower is present at 50 km range. The wider signal
associated with meteorological echo is present in the
range between 90 and 115 km.
The Radar Equation
The radar equation expresses the relationship between
the transmitted power and backscattered received
power from precipitation targets in terms of the
radar's hardware characteristics and the distance
between the transmitter and the target. In this section,
the radar equation and radar reflectivity will be
derived by first making some simplifying assumptions
and then gradually refining the terms to more accu-
rately represen! the electromagnetic theory underpin-
ning precipitation radars.
Isolated Scatterers
The amount of power incident (P¡) at an isotropic
target of cross-sectional área At at range r\m an
isotropic transmitter is given by eqn [2], in which Ft is
the transmitted power (Figure 3A).
Transmitted power and received power are commonly
expressed in units of watts or dBm. The latter
represents the ratio of power (?) in watts relative
to 1 milliwatt (101og10[P/10"3W]). In a typical