Local weather forecast by
"City, St" or zip code |
|
Search by city or zip code. Press enter or select the go button to submit request
|
|
You are at NWS Houston/Galveston »
Research Projects » VWP Responses to Changes in VAD OSR
Impacts of Optimum Slant Range
On 88D VAD Wind Profiles
Prepared by Steve Allen, SOO, Houston/Galveston NWSO; last updated 4/27/96.
Notes or comments should be addressed to
steve.allen@noaa.gov
It has often been noted by 88D operators that changes in the Optimum Slant Range (OSR) adaptable parameter of the Velocity-Azimuth Display (VAD)
algorithm can have profound effects on the apparent wind field (the VAD Wind Profile, or VWP product). It is the purpose of this study to examine how changes
in the OSR may produce these effects, and how operators may use the OSR adaptable parameter to optimize VAD wind estimates and VWP output.
As an example of how dynamic the VAD response can be to changes in the OSR, consider the VWP product above produced by the Houston/Galveston 88D
on October 16, 1994. (Click here to see the base reflectivity and velocity products at the time the OSR was changed from 10 to 150 km.)
Obviously, the change in OSR from 10 to 150 km has here induced a dynamic change in the indicated wind field of the VWP. And although it may
be argued that this particular change from 10 to 150 km was extreme (the OSF default OSR is 30 km), the actual critical change in the VWP product actually
takes place somewhere between the ranges of 30 and 45 km, as will be demonstrated below. Still, these vacillations in the VWP as a function of the
OSR continue to be something of a mystery, especially since VWP responses to OSR changes are not, strictly speaking, repeatable. That is, effects noted on any
given day due to a change in the OSR value may appear to have no effect at all on another. To examine how and why this may be, it is first necessary to revisit
just how the 88D VAD algorithm works:
The objective of the VAD algorithm is to sample radial velocities in a circle about the radar at constant elevation and slant range where the combined beam
elevation and slant range intersect a desired reporting altitude. (These "reporting altitudes" constitute the Y axis of the VWP products generated by the 88D.) The
algorithm applies a discrete Fourier transform on this azimuthal set of radial velocities, the result of which is a plot of radial velocity vs. azimuth, from
which the VAD gets its name. (Click here to view a sample of this product.) The application of the
Fourier transform permits accurate wind estimates even when the radial velocity sample is discontinuous about the sample circle.
The VAD technique was first introduced by Lhermitte and Altas (1961) who demonstrated that this method could provide an accurate assessment of the
horizontal wind in regions of relatively homogenous precipitation. Later work by Browning and Wexler (1968) and Rabin and Zrnic (who applied the
technique in clear air, 1980) have expanded on the original theme, but the final 88D algorithm (also by Rabin and Zrnic) is most closely related to the method
first applied by Lermitte and Atlas for two primary reasons: the first is that it presumes a nearly homogenous horizontal flow, and the second is that it allows
for the radial velocity sample to be taken at any of a number of slant ranges and beam elevations.
While allowing these parameters to vary permits a broad range of sampling possibilities, the algorithm is not specifically designed to take advantage of the
best available sample (i.e. where the greatest number of radials about the circle have measurable returns), nor is there any provision in the algorithm to avoid
some types of errors noted by previous researchers. As for the former issue, the "best sample" is effectively the choice of the radar operator, and will be
addressed below. As for the latter, Browning and Wexler (1968) proposed that combined effects of vertical wind shear and inhomogeneities in precipitation fall
speed were likely to create unacceptable errors in VAD estimates when slant range exceeded 23 km and/or elevation angles exceeded 9 degrees (in rain) or
27 degrees (in snow). While their work was intensely concerned with non-linear wind fields, some results of the current study suggest that a few of the
peculiarities of the 88D VAD products may be related at least in part to sampling in the higher elevation angles available to the radar.
In the current 88D algorithm, the only preprocessing performed on the radial velocity data is the (attempted) elimination of bad data. This includes returns
from ground clutter and anomalous propagation as well as range folded data. The only enhancement currently at work in the algorithm to optimize sample quality is to
average across three adjacent sample volumes when the radar is operating in clear air (to minimize the scatter of velocities by weak echoes). The only cautions currently
provided by OSF regarding the algorithm are that the use of very short slant ranges may acquire radial velocity samples contaminated by ground clutter and that Doppler
resolution inherently decreases with range.
As has been previously mentioned, the provision of adaptable parameters in the 88D VAD algorithm permits considerable latitude concerning the range and
beam elevation at which samples are taken for the desired reporting altitudes. But as has also been shown, selection of the OSR, which controls these
parameters, can be of critical importance to the representativeness and the accuracy of the VWP displayed by the radar. It is therefore necessary to understand what
role the OSR actually plays in the VAD algorithm, and what changes take place when the OSR is changed.
To begin with, it must be understood that the OSR set by the operator in the VAD adaptable parameters is by no means the actual slant range used
by the radar for all the winds displayed in the VWP. This should be clear from geomotretic considerations alone:
If we take as an example the OSR of 150 km from the VWP previously viewed, it is clear from the above diagram that if the OSR were an absolute
control on the VAD, no winds below about 9,000 feet could be sampled at all. Conversely, the OSR of 10 km could not possibly take wind samples above
12,000 feet. Yet we have seen that the radar reported winds through a deep tropospheric layer at both of these OSRs
(Click here to review the VWP product), so something else must set the actual slant range used by the radar.
In fact, the 88D has an internal set of VAD analysis slant ranges, or VASRs, which it may use for the actual VAD assessment. The VAD algorithm merely
checks the operator supplied OSR against the VASR table for any given reporting elevation, and then chooses the VASR that is nearest to the OSR for
the closest possible fit. Because the radar has a finite set of elevation angles available, there is only rarely an exact fit between OSR and VASR. As a
consequence of these OSR/VASR relations, in conjunction with reporting altitudes (also operator specified) and elevation angle, the principal effect of
changing the OSR is to force the VAD algorithm to use radial velocity samples from higher elevation angles for low OSRs, and from lower elevation angles for high
OSRs. That is, the actual elements being altered are the beam elevation of the radial velocity sample and the horizontal domain over which it is taken.
(Click here to view a graphic depiction of how the horizontal domains and beam elevations of the VAD
sample vary as a function of OSR.)
It goes without saying that one of the most important factors in radar theory is the back-scattering of energy from targets present in the radar beam.
Hence it is no surprise that the most crucial element affecting the accuracy of the VAD algorithm is the radial velocity sample available for evaluation. Thus, to optimize
VAD, it is important to select a slant range that will use a sample circle that has a large number of radial velocity samples (high echo density) but within a range domain
over which the flow can be expected to be relatively homogenous. Unfortunately, the current 88D VAD algorithm has no "quality control" mechanism for comparing
the radial velocity samples of the various VASRs for the purpose of seeking this optimal balance. Instead, it merely takes the radial velocity sample of the VASR that
is closest to the operator specified OSR for each given reporting elevation. The result of this arrangement can hardly be understated: ONLY THE RADAR OPERATOR
CAN DEFINE THE REGION OF THE OPTIMUM RADIAL VELOCITY SAMPLE. This is what the "optimum" part of "optimum slant range" is all about, and leaving this
element permanently assigned to the 88D's default settings can frequently lead to VWP products which seem curiously unrepresentative or innaccurate.
To examine empirically how the VWP changes as a function of OSR, the RADS (WATADS) software developed by NSSL was used to examine Houston 88D
data from a two-hour period on October 16, 1994 (the same date as the images viewed above). The same two hours of archive level II data was re-run 8 times, with
OSRs of 2, 10, 30, 45, 60, 75, 120, and 150 km. While it would have been preferable to examine the exact same time period, this data was not available from the
archive II data tapes, and the closest available contiguous 2 hours of data have been used instead. Select from the list below to examine the RADS VWP output:
What this data clearly shows is that an operator selected OSR of 45 to 75 km would have definitely been the best "optimum" slant range to return accurate VAD/VWP
winds in this particular case. (Shortly after the real-time OSR switch to 150 km, the OSR was reset to 60 km to enhance VWP output.)
The beneficial impacts of such "operator aware" changes to OSR should be self evident. Some of the possible detrimental effects of leaving OSR fixed at the OSF
default of 30 km are as follows: in the not-too-distant-future, the VAD output (which goes into the 88D's radar coded message (RCM)) is slated to be an input to the rapid
update cycle (RUC) model currently in use at NMC. The sensitivity of meteorological models to initial conditions is a well demonstrated fact and will not be entertained in
this study. However, it is important to consider the downstream impact (on RUC output) represented by as broad a range of possible initial kinematic conditions as
we have seen in the RADS output. Secondly, there is the natural "second-nature" trend of NWS meteorologist to accept PUP displays almost unconsciously. That is, when
a radar operator clicks the 88D puck on VWP, he or she is likely to give some credibility to the PUP image that subsequently appears, regardless of how contrary
it may be to considerations of radial velocity and/or reflectivity data available on alternate displays. Thirdly, the VWP may be the ONLY product viewed by some users. For
example, the Storm Prediction Center (SPC, formerly SELS) frequently uses VWP for a quick evaluation of shear and low level jet evaluations.
It is the hope of this study to clarify the need for conscientious selection of appropriate slant range by forecasters in the operational environment. The method
suggested is simple: Have the radar analyze the wind where the Lagrangian sample is apt to be the best. In the very simplest sense, this merely means setting the OSR
to the range of best radar returns. The caveats to this method are simply that forecasters must understand that in non-linear wind regimes the radar may produce a wind
profile (given the constraints of VASR/OSR discussed above) which is not representative of the horizonal wind at the RDA site.
Acknowledgements:
Julie Hall, David Himes, Pat Parrish, Heidi Lindenlaub, and Peggy Bruehl (at COMET) helped make the initial version of this document (Nov. 1995) possible with
their technical support
My COMET Mentor Dr. Wen Chau Lee directed my studies in the evolution of the VAD technique and its subsequent applications in the NEXRAD system.
References:
- Dr. Wen Chau Lee (conversations at COMET, September/October, 1995)
- Captain Jerry Davis of OSF, phone conversations, 88D VAD algorithm ennuncation language, various details on the actual performance of the current 88D
VAD algorithm.
- Subsynoptic-Scale Vertical Wind Revealed by Dual Doppler-Radar andVAD Analysis, Rabin & Zrnic, Journal of the Atmospheric Sciences, 1979, p.644-654
- The Determination of Kineatic Properties of a Wind Field Using Doppler Radar, Browning & Wexler, Journal of Applied Meteorology, Feb 1968, p105-113.
- Precipitation Motion by Pulse Doppler, Lhermitte & Atlas, 9th Conference on Radar Meteorology, 1961, p.218-223
- An Improved Version of the Extended Velocity-Azimuth Display Analysis of Single-Doppler Radar Data, Matejka & Srivastava, Journal of Atmospheric and Oceanic Technology, Aug. 1991, p. 453-466.
Miscellaneous Items
|
|
weather.gov 
National Weather Service Forecast Office
