Printer Friendly Version
Back to Intensity Change projects | Back to Main Projects Page

Real-Time Doppler Winds

Principal Investigator
John Gamache

Team Members:
Peter Dodge (HRD)
Nancy Griffin (HRD)
Joe Griffin (HRD)
Paul Leighton (HRD)
Frank Marks (HRD)
Sundararaman.G.Gopalakrishnan (HRD)

Collaborating Scientists
Naomi Surgi (NCEP/EMC)
Qingfu Liu (NCEP/EMC)


  • To control the quality of airborne Doppler radar data.
  • To analyze the quality-controlled data in real time.
  • To deliver the data and analyses in 'real time' to the Environmental Modeling Center and National Hurricane Center.
  • To examine how these data may be incorporated into numerical simulations.
  • To determine the best flight strategies for obtaining data that will initialize the simulations.

Rationale and Methodology:

As operational hurricane forecasts attain finer resolution, the models must be initialized with finer-resolution data. Airborne Doppler radar observations provide the most widespread observations of the hurricane core at a 1.5 km or better horizontal resolution. At the Hurricane Research Division we have been developing methods to produce research-quality analyses from the P-3 airborne Doppler radar for over 20 years. Within the last decade we have been refining a three-dimensional variational method to produce three-dimensional wind analyses, as well as analyses of winds in a vertical cross-section defined by the flight track of the aircraft. The three-dimensional analysis scheme was described by Gamache et al. (1997), and has been modified in such a way as to be nearly identical to the method described by Gao et al. (1999). The variational scheme simultaneously minimizes

  1. the difference between the projection of the wind analysis on the Doppler radial direction cosines, and the observed Doppler radial velocities,
  2. the error in the 4-th order discretized three-dimensional mass divergence of the wind field, and
  3. the various second derivatives of wind components.

The third term acts as a filter of 2 delta x noise resulting from solving the discretized mass-continuity equation, and is kept as small as practically possible. Although this method of solution is computationally expensive it works best in situations where the aircraft must fly through the middle of the feature it is observing, as in aircraft penetrations of the hurricane eyewall.

Until 2004, the data analyzed in the aforementioned method were manually quality controlled, which is a time consuming process, although possible when doing research. To quality control an hour of airborne Doppler data can take one person approximately 2-4 weeks. To produce analyses in 'real time,' that is within minutes to an hour of the observations, quality control must be performed automatically, and methods were developed in 2003 and 2004 to perform such quality-control tasks as, 1) to remove noisy data, 2) to remove reflections of the main and side lobes by the sea surface, and 3) to correct improperly de-aliased Doppler radial velocities. Because of present limitations of the airborne Doppler radar signal processing, a three-step process is used. In the first step, the single-PRF (single pulse repetition frequency) data are de-aliased using a single-radial method developed by Bargen and Brown (1980), followed by a two-dimensional method developed at HRD. Any data beyond a radial gap of 1 km or greater are discarded, since the Bargen-Brown method can be unreliable with greater gaps in the high-gradient eyewalls of hurricanes.Those data left are analyzed in a cylindrical analysis centered on the storm center, but with wavenumber 0 and 1 Fourier components in the azimuthal direction. This low-wavenumber analysis is then used to improve the de-aliasing in a second pass through the Doppler data, where a three-dimensional Cartesian analysis of the wind is produced. The lowest levels of this analysis are then transmitted back to the National Hurricane Center. A second type of analysis uses the same quality-controlled data, but incorporates Doppler data well upwind and downwind of the flight track, and looking both fore and aft of the flight track, to produce higher resolution vertical cross-sections of wind speed, radial wind (wind component toward or away from storm center), vertical wind, and reflectivity.Examples of both the three-dimensional analysis and the vertical profiles are shown below.

Starting in 2008, the data used in the analysis will be transmitted to the Environmental Modeling Center for assimilation into the operational hurricane model (HWRF-Hurricane Weather Research and Forecasting). In 2008, the data may still only be assimilated into a 'parallel' run at EMC. Efforts continue in 2007 at HRD to make the actual running of the real-time quality control aboard the aircraft, and the transmission of data from the aircraft, as automatic (with as little onboard human interaction) as possible. Further development and improvement will occur in 2008, when the Sigmet RVP-8 Doppler-radar processing will be operating on 2 NOAA P-3s (N42RF and N43RF) and the NOAA Gulfstream-IV (N49RF) aircraft.We also hope to examine independently how data could best be collected to meet the model needs (in OSSEs).

A new modeling effort is starting up at HRD to test different procedures in hurricane simulations.One such effort will involve studying different methods in which the Doppler radials or Doppler wind analyses may be used to initilize numerical simulations.


  • Method tested successfully on NOAA WP-3D aircraft in 2004.

  • Doppler analyses transmitted to NHC in real time in 2005 (in particular in Wilma of October 2005)

  • Real-time quality controlled analyses of airborne Doppler radar data accepted by JHT (NHC) for operational use.

  • Real-time analyses transmitted in real time from Tropical Storm Ernesto and Hurricane Helene of 2006.

  • New more automatic determinations of storm center, and some quality control parameters developed in spring 2007.

  • A simple GUI menu was developed to help set up parameters for automatic quality control and analysis

  • Real-time analyses transmitted in real time from Tropical Storm Erin, Hurricane Felix, Tropical Storm Ingrid, Tropical Depression 10, and Hurricane Karen

Goals for FY 2008:

  • Develop I/O software for RVP-8 Sigmet data in quality control/analysis software

  • Develop and test a completely automatic quality control program and try it in 2008 hurricane field season


Bargen, D. W., and R. C. Brown, 1980:Interactive radar velocity unfolding. Proc. 19th Conference on Radar Meteorology, Miami, Amer. Meteor. Soc., 278-283.

Gamache, J. F., 1997: Evaluation of a fully three-dimensional variational Doppler analysis technique. Preprints, 28th Conf. On Radar Meteorology , 9-13 Sept. 1997, Austin, AMS.

Gamache, J. F., 2005:Final report on JHT project entitled: Real-Time Dissemination of Hurricane Wind Fields Determined from Airborne Doppler Radar.

Gao Jidong, Ming Xue, Alan Shapiro, Kelvin K. Droegemeier, 1999: A Variational Method for the Analysis of Three-Dimensional Wind Fields from Two Doppler Radars. Mon. Wea. Rev: 127, 2128-2142.

Total wind-speed vertical cross-section

Radial wind vertical cross-section

Total wind speed at 500 meters

Examples of analyses produced from automatically quality-controlled airborne Doppler observations collected in Hurricane Katrina on 28 August 2005, when it was a category 5 hurricane. Top and bottom panels on left are total wind speed and radial wind, respectively at 1725-1818 UTC, while the analysis on the right is of the total wind speed at 500 meters above sea level.

Last modified: 10/24/2007

Stay Connected