THREE-DIMENSIONAL VARIATIONAL ANALYSIS OF
AIRBORNE DOPPLER OBSERVATIONS
John F. Gamache
(MULTIPLE-DOPPLER HARMONIC ANALYSIS)
Wen-Chau Lee (NCAR/ATD/RSF)
Frank Marks, Jr.
To produce a wind field rapidly on the aircraft that incorporates airborne
Doppler radar data obtained in normal or FAST (Fore/Aft Scanning Technique)
Rationale: The first
priority of HRD's flying is research, including applied research that supports
the hurricane forecasting effort, and basic research that helps us to understand
physical processes in hurricanes better. With the advent of the VTD
technique, a potential conflict exists between these two needs. VTD
requires Doppler radar scans that are normal to the flight track, but more
complete information about storm structure is possible if the hurricane
is observed by Doppler from more angles as simultaneously as possible,
and this often calls for use of the Fore/Aft Scanning Technique (FAST),
where the radar scans alternately in cones fore and aft of the plane perpendicular
to the fuselage. A quick method must be found that uses FAST data.
A second rationale is that
more information concerning the radial flow and flow asymmetries can be
determined from one pass with one aircraft operating in the FAST mode.
Thus if this method can be made fast enough it will actually be an improvement
upon flat scanning VTD.
Presently only the feasibility has been tested, but the method used in
the feasibility test is:
A best guess is made for the
location of the storm circulation center location and motion
Doppler data are first interpolated
to a horizontal two-dimensional polar grid at flight level, where the polar
origin is the estimated storm center.
A wind field is determined using
MDHA, where the vertical wind is assumed zero,
and only data obtained within 60 degrees of horizontal is employed.
A new storm center is estimated
from the newest analysis.
The projections of the smooth
MDHA wind analysis upon the Doppler radials,
is compared with the observed Doppler radial motion, and where necessary
the radial observation is changed by a multiple of two times the Nyquist
velocity (also called unfolding or dealiasing), or ignored if the difference
is approximately an odd multiple of the Nyquist velocity.
A new two-dimensional analysis
is performed using the dealiased radials.
A new center location is estimated.
This is the final assumed center.
A full three dimensional interpolation
and analysis is peformed using the final center.
All the data in the analysis
volume are dealiased.
Final analysis is done in the
three-dimensional analysis using the interpolation of the dealiased radials.
Software has been written that
shows the feasibility of the method as a completely automatic one; however,
automatic removal of surface reflectivity has not been coded, and the method
as used is too time consuming. The emphasis of future work will
be on reducing the time necessary to complete the analysis.
Two guesses for the storm center
that were over ten kilometers apart were each used as a starting center
for the two-dimensional analyses. The final estimates for the two
guesses were less than one kilometer apart. This difference could
become a little higher when three-dimensional analyses are done due to
the greater effective filtering in the three-dimensional version.
Last modified: 02/01/00