pdf online search and view
Loading...

 Report broken link

System Requirements for Phased Array Weather Radar
Dusan S. Zrnic and Richard J. Doviak
National Severe Storms Laboratory, Norman OK

Table of Contents
Preamble ........................................................................................................................................ 1
1. INTRODUCTION..................................................................................................................... 1
2. WEATHER RADAR SPECIFICATIONS ............................................................................. 2
2.1 Surveillance ................................................................................................................. 3
2.1.1 Range ............................................................................................................. 3
2.1.2 Time................................................................................................................ 3
2.1.3 Volumetric coverage ...................................................................................... 4
2.2 Signal to Noise Ratio (SNR) ....................................................................................... 5
2.3 Wavelength .................................................................................................................. 6
2.4 Angular resolution ...................................................................................................... 6
2.5 Range resolution ......................................................................................................... 7
2.6 Accuracy of measurements ........................................................................................ 7
2.7 The WSR-88D specifications...................................................................................... 8
2.7.1 Transmitted power ......................................................................................... 8
2.7.2 Transmitted pulse widths ............................................................................... 9
2.7.3 Transmitter phase noise ................................................................................ 9
2.7.5 Dynamic range............................................................................................. 10
2.7.7 Ground clutter canceling ............................................................................. 11
2.7.9 Estimation errors ......................................................................................... 16
3. REFLECTIVITY FACTOR FOR VARIOUS SCATTERING MEDIA ........................... 16
3.1 Non precipitating clouds .......................................................................................... 16
3.2 Precipitation .............................................................................................................. 17
3.3 Cloudless skies........................................................................................................... 18
4. POLARIMETRIC REQUIREMENTS................................................................................. 19
4.1 Match of H and V patterns ...................................................................................... 21
5. REFERENCES........................................................................................................................ 22

System Requirements for Phased Array Weather Radar
Preamble
This is the final report to Lockheed Martin on the requirements for a phased array
weather radar. Discussed herein are those meteorological requirements that need to be met by
weather radars having either reflector type antennas or active phased array elements. These
requirements impact the design of the antenna, the transmitter-receiver, and signal processor.
Justifications for these meteorological requirements (presented in this report) are based, for the
most part, upon meteorological considerations, but sometimes practical limitations of the radar
dictate what can be done. Because microwaves penetrate cloud and rain, they are in an
unchallenged position to remotely survey the atmosphere.
The Doppler radar is the only remote sensing instrument that can detect tracers of wind to
reveal a storm’s internal structure and the hazardous phenomena harbored therein. The recent
inclusion of polarimetric capability to research Doppler weather radars has added a new
dimension to radar observations by providing information that can be used to considerably
improve the estimation of rainfall, as well as allow for the classification of precipitation type.
Thus, included in this report is discussion on the requirements for polarimetric radar. The unique
capabilities of the polarimetric Doppler radar make this the instrument of choice to survey
weather.
1. INTRODUCTION
Weather observations have been part of radar technology since its very beginning.
Although military applications have driven this technology and are often credited with its
inception, first radar-like instruments were made in the 1920s to remotely probe the ionosphere.
Today a similar technology is used to produce vertical profiles of winds throughout the
troposphere (to over 15 km above ground). These wind profiling radars operate in the UHF
(e.g., 400 MHz) and VHF (e.g., 50 MHz) bands and serve mainly the research community.
Nonetheless, specialized applications exist to monitor conditions at launch or landing of space
vehicles, and there is even a small experimental network of profiling radars in the central USA.
Over the decades the weather radars have undergone considerable change, principally
because of the advances in solid state technology that allows for real time quantitative analysis
of the weather signal, the continuous stream of echoes from scatterers that can be distributed
over hundreds of kilometers. These advances in solid state technology, however, have
principally impacted the receiver, signal processing, and display components of the radar; the
transmitter and antenna have essentially been unchanged, and the transmitted signals to this day
are pulsed sinusoids of relatively short duration and high power. The advent of the phased array
antenna, and the potential to distribute the transmitted power over an array of elements, as well
as provide agility in positioning the beam, offers the potential of improving weather observations
and the timeliness of warnings. Thus, this report presents those basic requirements that, for the
most part, are met by the present day weather radars, and therefore must be met on modern
phased array radars.
1

Of all weather radars, the most commonly used are airborne surveillance radars. Yet the
general public is mainly aware of ground based surveillance radars whose images are displayed
on television sets. The prime purpose of ground based radars is to estimate the amount of rain or
snow reaching the ground. Equally important for the National Weather Service is to detect
severe storms that harbor hazardous winds and damaging hail so that timely warnings can be
issued. Three frequency bands are allocated to weather surveillance radars. These are the S (~
10 cm wavelength), C (~ 5 cm wavelength) and X (~ 3 cm wavelength) bands. The shorter
wavelength radars are less expensive and have smaller size but are significantly affected by
attenuation, briefly discussed in Section 2.3 of this report. Hence the USA national network of
Doppler weather radars (i.e., the WSR-88Ds) operates at about 10 cm wavelengths. Throughout
this report, stated requirements are for 10 cm wavelength weather surveillance radars.
2. WEATHER RADAR SPECIFICATIONS
We start with the specifications given in Table 2.1 for observing weather with radar.
These are obtained from Tables 3.1 and 6.1 in Doviak and Zrnic (1993) which are based upon
years of experience with the USA’s national network of non-Doppler radars (i.e., the WSR-57),
as well as the experience gained with Doppler radar observations by the research community. It
is important to distinguish between the requirements that are based on proven meteorological
grounds and those which are consider desirable, but are still in the process of exploratory
development.
Table 2.1
Requirements for weather radar observations
1.1. Surveillance
1.1.1 Range:
1.1.2 Time:
1.1.3 Volumetric coverage:
1.2. SNR:
1.3. Angular resolution:
1.4. Range sample interval Δr
1.4.1 for reflectivity estimates:

460 km
< 5 minutes
hemispherical
> 10 dB, for Z . 15 dBZ at r = 230 km
≤1o

Δr < 1 km; 0 < r <230 km
Δr < 2 km; r < 460 km
1.4.2 for velocity and spectrum width estimates (r < 230 km):
Δr = 250 m
1.5. Estimate accuracy:
1.5.1 reflectivity:
≤1 dB
1.5.2 velocity:
≤1 m s-1; SNR> 8 dB; σv = 4 m s-1
1.5.3 spectrum width:
≤1 m s-1; SNR>10 dB; σv = 4 m s-1

2

The specifications in Table 2.1 have, for the most part, been accepted by the
meteorological community. The principal parameters are those that determine the weather
radar’s SNR required to estimate reflectivity, Doppler velocity, and spectrum width with
specified accuracy for scatter from various classes of weather (principally precipitation), and the
required spatial resolution. These fundamental requirements of SNR and spatial resolution are
met by the WSR-88D radar and thus should serve as a baseline for the phased array radar.
2.1 Surveillance
The surveillance range, time, and volumetric coverage are routed in practical
considerations, previous experience with weather radar, and the sizes and lifetimes of some
significant meteorological phenomena.
2.1.1 Range
Surveillance range is limited to about 460 km because storms beyond this range are
usually below the horizon. Without beam blockage, the horizon is at 12.5 km altitude at 460 km;
thus only the tops of strong convective storms can be detected. Although these storms can be
detected at ranges between 230 and 460 km, quantitative measurements of precipitation are only
required for storms at ranges less than 230 km. Nevertheless, in the region beyond 230 km, storm
cells can be identified and their tracks can be established. Even at the range of about 230 km, the
lowest altitude that the radar can observe under normal propagation conditions is about 3 km.
Extrapolation of rainfall measurements from this height to the ground is subject to large errors,
especially if the beam is above the melting layer and is detecting scatter from snow or melting
ice particles. In this regard satellite borne radars have an advantage because they can have nearly
uniform vertical resolution over large areas. But satellite borne weather radars are, for now,
limited to orbiting satellites, and thus they cannot continuously monitor weather phenomena.
2.1.2 Time
Surveillance time is determines by the time of growth of hazardous phenomena as well as
the need for timely warnings. Five minutes for a repeat time is sufficient for detecting and
confirming features with lifetime of about 15 min or more. Typical mesocyclone life time is 90
minutes (Burgess, et al., 1982). Ordinary storms last tens of minutes but microbursts from these
storms can produce dangerous shear in but a few minutes.
The principal driver to decrease the surveillance time is the prompt detection of the
tornado and the need to have timely warnings of its presence. Presently, the lead time for
tornado warnings (i.e., the time that a warning is issued to the time the tornado does damage) is
about 12 minutes. Tornadoes can rapidly develop from mesocyclones and it takes about six
minutes for the reflector antenna to return to the position of the first radar sighting of the
tornado. Because two consecutive radar observations of tornado vortex signatures (TVS) are
required before a tornado is confirmed to be observed, a beam agile phased array has a definitive
advantage; it can return to location of the first detection of a TVS within the minute. Thus about
5 minutes can be added to the lead time of tornado warnings.

3

For over ten years meteorologists have been developing schemes to retrieve winds
transverse to the radar beam. Recent results at the University of Oklahoma (Fig. 1) indicate that
a rapid increase of errors ensues if the update time increases beyond 1 min. Figure 1 suggests
that further improvement is possible for even shorter update times.

Figure 1. Root mean square error of transverse wind as a function of time between scans.
Figure was provided by Alan Shapiro.

2.1.3 Volumetric coverage
There is no firm requirement for the highest elevation the radar should cover, and in
practice a compromise is made between the highest elevation, the scan time, and the number of
tilts in the scan pattern. The volume scan patterns currently available on the WSR-88D have
maximum elevations of 20 deg. For example the VCP-11 has 14 elevations and ends at 19.5 deg
in about 5 minutes. Brown and Wood (2000) have examined alternative scan patterns. The
demanding one (Fig. 2), termed VCP-γ, is quite similar to the VCP-11, it has 14 elevations, it
ends at 19.5 deg, but it has to be completed in 4.1 min. Meteorologists from the NWS have
expressed a desire to extend the coverage to higher elevations to reduce the cone of silence.
Therefore, NSSL has proposed and collected data in a pattern that extends to 30 deg, has 14
elevations and can be executed in less than 5 min. Furthermore, there are proposed patterns that
update at 5 min and end at 50 deg, but no data have been collected using these. It is fair to state
that a maximum elevation of 50 deg will likely remain an upper limit, and the 30 deg might be a
practical upper limit for the WSR-88D. Top elevations higher than 20 deg have not been justified
by strong meteorological reasons. In NSSL’s recently proposed patterns, the data (i.e., spectral
4

moments) are provided at 0.5 deg in azimuth in order to extend the range for detection of
mesocyclones and violent tornadoes (Wood et al. 2001). This improvement in range of detection
is at a small expense of velocity accuracy.

Figure 2. A demanding scanning strategy for the WSR-88D, from Brown and Wood (2000).

2.2 Signal to Noise Ratio (SNR)
Using the relations between rainfall rate and equivalent rainfall rate for snowfall given by
Doviak and Zrnic (1993, Section 8.4) for Z = 15 dBZ, the SNR listed in Table 1 provides the
specified accuracy of velocity and spectrum width measurements to the range of 230 km for both
rain and snowfall at a rate of about 0.3 mm of liquid water depth per hour. The SNR specified in
Table 1 is consistent with the SNR = 0 dB for a reflectivity factor Z = -8 dBZ at 50 km specified
in the Nexrad Technical Requirements (NTR 1984) document; 10 dB or more is added to insure
that Doppler velocity and spectrum width estimates are made with specified accuracy
independent of SNR given the surveillance time constraints. Furthermore, at rainfall rates
smaller than about 1 mm h-1, the differential reflectivity factor is practically unresolvable
(Doviak and Zrnic 1993, Fig.8.21), so that polarimetric measurements would not add any
additional information to improve rainfall or to classify precipitation. In conclusion, the SNR
specified in Table 1 provides quantitative measurements of precipitation to ranges of about 230
km, and potential, with polarimetric Doppler radar, to improve rainfall measurements for rain
rates larger than about 1 mm h-1 (corresponding to Z values of about 25 dBZ).
5

2.3 Wavelength
Accurate rainfall estimates depend upon accurate Z measurements which require
corrections for propagation loss. Although, for a given antenna size, angular resolution improves
in proportion to the wavelength, and thus shorter wavelengths are favored, a 0.1 m wavelength
has been selected for weather radars in the USA because attenuation through rain is relatively
negligible compare to that at shorter wavelengths. Correcting for attenuation becomes
excessively inaccurate if propagation losses exceed 10 dB (Hitschfeld and Bordan 1954). To
have accurate measurements to ranges of 230 km requires that the specific attenuation be less
than 0.02 dB km-1.
But at wavelengths much shorter than 10 cm (e.g., 5 cm, a weather radar band) specific
attenuation is higher than 0.02 dB km-1 for Z > 40 dBZ. Severe storms along squall lines have Z
values as high as 60 dBZ and specific attenuation up to 0.6 dB km-1. Thus 5 cm wavelength
radars will not meet the requirements under severe conditions (i.e., Z > 40 dBZ). Furthermore, it
has been observed that storms aligned along the radial can be significantly obscured by
attenuation, and important observations of tornado cyclones can be lost if 5 cm wavelength
Doppler radars are used. Nevertheless, countries at higher latitudes, where tornadoes are
infrequent, generally have opted for 5 cm wavelength radar because precipitation usually forms
in stratiform conditions, reflectivity values are typically less than 40 dBZ, and storms are
smaller.
2.4 Angular resolution
The cited spatial resolutions are principally determined by the need to resolve
meteorological phenomena such as tornados and mesocyclones to ranges of about 230 km, and
the practical limitations imposed by antenna size at wavelength of 0.1 m. To achieve angular
resolution significantly less than 1o at wavelengths of 10 cm is cost prohibitive due to the size of
the reflector. Thus the angular resolution of 1o has been accepted as a practical limit for weather
radars operating at wavelengths of 10 cm.
Even though a beamwidth of 1o is considered to provide relatively high resolution (e.g.,
the WSR-57 weather radar=s angular resolution was 2.2o), the spatial resolution at a range of
230 km is 4 km. This is larger than many mesocyclone diameters, and thus these important
weather phenomena, precursors to many tornadoes, can be missed. Tornadoes have even smaller
diameters and therefore many of these cannot be resolved at ranges to 230 km.
Because the beam of the WSR-88D weather radar is scanning azimuthally, the effective
angular resolution in the azimuthal direction is somewhat larger (Doviak and Zrnic 1993, Section
7.8). Typically, the effective azimuthal resolution is about 40% larger at the 3 RPM scan rates of
the WSR-88D. The advantage of the phase array radar is that the beam can be stationary during
the time that echo samples are being processed to estimate the moment (i.e., reflectivity factor Z,
Doppler velocity v, and spectrum width) fields. Thus, the phase array radar could have an
antenna that is 40% smaller in width without compromising the present performance experienced
by NWS radar meteorologists. But the vertical size of the array needs to be about same as that of
6

the WSR-88D radar (i.e., about 8.5 m) in order to maintain the same vertical resolution. On the
other hand, for the same size aperture, the phased array radar would provide an azimuthal
resolution that is 40% better than presently achieved with the scanning reflector antenna
provided the beam is pointed in the bore sight direction.
In conclusion, a phased array radar with beamwidths in el < 1 deg, and az < 1.4 deg
should be acceptable. Because the atmosphere is vertically stratified a narrower beam width in
elevation is typically needed. For example the melting layer is few hundred meters deep and
should be resolved to as large a range as possible so that it does not contaminate rainfall
measurements.
2.5 Range resolution
The range resolution is somewhat controlled by the angular resolution; there is
practically no gain in having range resolution much finer than the angular one. For example, the
angular resolution equals the 250 m range resolution at a range of about 15 km. Thus, improving
the range resolution can significantly enhance observations only if the phenomena are closer
than 15 km. But, better range resolution can provide additional shear segments and therefore
improve detection of vortices at larger distance.
Although angular resolution limits, for the most part, the resolving power of the radar,
maintaining a higher resolution in range for most of the range of measurements is useful on two
accounts. First of all, detail estimates of tornadic winds and size can be calculated from the
Doppler spectrum (e.g, Doviak and Zrnic 1993; Figs 9.29 and 9.31) at appropriate ranges.
Another reason for this relatively fine range resolution is to allow accumulation of independent
samples in a range interval of 1 km. The average of these samples reduces statistical uncertainty
of estimates.
The range resolution for reflectivity is coarser for two reasons: (1) reflectivity is
principally used to measure rainfall rates over watersheds and thus spatial resolution as high as
that required to resolve mesocyclones is not needed, and (2) reflectivity samples at a resolution
of 250 m are averaged in range (Doviak and Zrnic, 1993, Section 6.3.2) to achieve the required
accuracy of 1 dB.
2.6 Accuracy of measurements
Using the reflectivity-factor/rainfall-rate relation for stratified rain given by Doviak and
Zrnic (1993, Eq.8.22a), the specified 1 dB accuracy of reflectivity measurements provides about
a 15% relative error of rainfall rate. This has been accepted by the meteorological community.
Nevertheless, errors larger than this are often found because the radars, except polarimetric ones,
cannot provide sufficient information about the drop size distribution required to estimate
rainfall rate at the ground.
The specified accuracies of velocity and spectrum width estimates are those derived from
observations of mesocyclones with research radars which provided the accuracies specified in
Table 2.1. The 8 dB SNR is roughly that level at which the accuracy of velocity and spectrum
7

width estimates do not improve significantly for increases in SNR (Doviak and Zrnic, 1993;
Sections 6.4, 6.5). But, it is possible that lower accuracies can be tolerated and benefits can be
derived therefrom. For example, it has been proposed (Wood et al. 2001) that velocity estimates
be made with less samples (e.g., by a factor of two) in order to improve the azimuthal resolution.
Although this decreases the accuracy of the Doppler velocity estimates by the square root of
two, the improved angular resolution can increase the range, by about 50% (Brown et al. 2002
and 2005), to which mesocyclones can be detected. With more signal processing power,
automated algorithms can search the higher resolution data fields for tornado signatures, while
the data field with normal resolution and accuracy can still be available for other algorithms that
might require velocity fields with higher accuracy. But even if data at the finer resolution but
worse accuracy is only available, the increased accuracy can often be obtained by simply
spatially averaging the data under the assumption that in most regions the spatial resolution is
finer than the spatial scale of the weather.
2.7 The WSR-88D specifications
To achieve the specifications presented in Table 1, the WSR-88D radar has been
designed to have the parameters listed in Table 2.2.

Table 2.2
WSR-88D parameters
2.1 Peak transmitted power:
2.2 Transmitted pulse widths:
2.3 Transmitted phase noise:
2.4 System noise power:
2.5 Dynamic range:
2.6 PRFs:
2.6.1 short pulse (8 selectable):
2.6.2 long pulse:
2.7 Ground clutter canceling:
2.8 Antenna
2.8.1 Gain:
2.8.2 Beamwidths (equal in azimuth and elevation):
2.8.3 Sidelobes:

475 kW
1.57 and 4.57 μs
<0.2o
-113 dBm
93 dB
320 to 1300 Hz
320 to 450 Hz
>50 dB
45.6 dB
1o
see Fig. 3

The parameters values listed in Tables 2.1 and 2.2 can be used to derive a reasonable set of
requirements applicable to phase array radar. We now discuss these parameters.
2.7.1 Transmitted power
The peak transmitted power level is chosen to provide the required SNR specified in
Table 2.1; pulsed sinusoids are transmitted and their length (also listed in Table 2.1) determines
the range resolution. Probably of greater importance than the peak power is the average
8

transmitted power. The duty factor of the WSR-88D radar is 0.002, and thus the transmitter
needs to supply an average power as high as 1 kW. Lower peak powers can be tolerated if larger
pulse widths are transmitted to maintain the average power and SNR requirements, under the
condition that pulse compression techniques are used to preserve range resolution, and range
sidelobes are smaller than 50 dB.
2.7.2 Transmitted pulse widths
The transmitted pulse width of the WSR-88D is simply that required to meet the range
resolution listed in Table 2.1. As noted in the previous section, longer pulse widths can be
tolerated if pulse compression techniques are implemented. Note that the radar has a non storm
observation mode whereby a long pulse (4.57 μs) is transmitted. Again this can be increased to
longer pulse widths and, contrary to the need for pulse compression to maintain range resolution
for storm observations, these techniques are not required for fair weather observations because
there is no specified range resolution in this mode; further most non storm weather phenomena
do not require the range resolution needed to observe hazards in storms.
Although this mode of operation is not the principal driver of the specifications of the
WSR-88D radar, it has been observed that this radar can detect returns from refractive index
irregularities in the convective boundary layer, and it has been suggested that weather radar
could as well monitor the wind throughout the troposphere (Doviak and Zrnic, 1993, Section
11.8.1.2). The principal motivation for making observations in clear air is to obtain wind profiles
to complement those obtained from rawinsonds and wind profilers. Thus the pulse width in this
mode will only be limited by the need to obtain a height resolution of about 300 m for soundings
below 7 km and of the order of 1 km for higher altitudes; these are the specifications for wind
profilers (Doviak and Zrnic, 1993, Fig.11.26). Thus, for radar beam elevation angles of about
30o, range resolution could be as large as 600 m for wind measurements to 7 km altitude, and
2,000 m for measurements at higher altitudes. To detect clear air refractive index irregularities,
the radar must operate in its most sensitive mode (i.e., longest pulse widths and smallest
bandwidths, and largest average power).
2.7.3 Transmitter phase noise
The performance of range velocity mitigation techniques proposed by the NSSL
(Sachidanada et al., Parts 1-4, 1997-2000), depend strongly upon the phase noise of the radar.
For example, if there were no phase noise in the system, the proposed systematic phase coding
schemes could retrieve velocity (with accuracies better than 2 m s-1) of signals 60 dB below the
signal power of a stronger overlaid out-of-trip echo (i.e., the logarithm of power ratio,
10log(ps/pw) = 60 dB, assuming that spectrum width of each signal is 4 m s-1 or less, and the
unambiguous velocity is at least 32 m s-1). However, if the phase noise is about 0.2o, the weaker
signal has to be within 40 dB of the stronger signal (i.e., 10log(ps/pw) < 40 dB). The phase error
of the WSR-88D has been measured to be less than 0.2o.
2.7.4 System noise power
The system noise power listed herein is the one that meets the SNR requirements (Table
2.1) for the given transmitted power. Lower system noise power is possible if a lower noise
9

Info

Published on : 2018-07-19 23:25:28
File Name : LMCO_Consult2.pdf
File Type :
Publisher :www.nssl.noaa.gov

File Source