Automated Quality Control and Analysis Methods for Real-Time and Research Processing of NOAA Airborne Doppler Radar Data

Tropical-cyclone airborne Doppler data have been collected aboard the NOAA WP-3D aircraft since 1982, and the NOAA Gulfstream IV (G-IV) since 2012. In 2003 the Hurricane Research Division (HRD) of NOAA/AOML (National Oceanographic and Atmospheric Administration/Atlantic Oceanographic and Meteorological Laboratory) was funded by the Joint Hurricane Testbed (JHT) to automate the quality control (QC) and analysis of airborne Doppler radar velocities collected by the NOAA aircraft. The purpose was to provide real-time analyses of the wind field in tropical cyclones for hurricane specialists at the National Hurricane Center (NHC), and quality-controlled Doppler radial velocities to NCEP (National Centers for Environmental Predicition) Central Operations (NCO) for assimilation in real-time runs of the Hurricane Weather Research and Forecasting (HWRF) model of the Environmental Modeling Center (EMC). At the conclusion of the JHT project, low-level horizontal and vertical-profile analyses of wind were being sent by script to the AOML ftp site, where they were polled by NHC/JHT software for viewing under NAWIPS (NCEP Advanced Weather Interactive Processing System). At the time (2005), EMC was not ready to assimilate the airborne Doppler data into HWRF, but the means to produce a QC'ed data stream was in place.

Through the cooperation of HRD, EMC, NCO, AOC (NOAA Aircraft Operations Center), and Remote Sensing Solutions (RSS), a format was determined for transmitting data to the National Weather Service (NWS) gateway, leading to the first assimilation in 2010 during Hurricane Tomas of airborne Doppler data in the real-time, but parallel, run of HWRF. Subsequently, after testing, the data have been assimilated into the operational runs of HRWF since 2013. Presently, only data from the WP-3D aircraft are assimilated in the operational HWRF run, while data from the G-IV continue to be evaluated in parallel or post-season runs of HWRF.

For each analysis period, a java application developed at HRD is used to determine the parameters needed to run the analysis and quality control. The very first QC step, performed before a flight, uses both precipitation returns and sea-surface returns to correct for small mounting errors of the antennas. The first real-time QC step removes suspected non-meteorological Doppler velocities, by removing the reflection of the main lobe from the sea surface, as well as data dominated by side-lobes and data speckles. Then the data are de-aliased.

A several-step process is used to effectively de-alias the Doppler data. First only data that can be simply de-aliased radially back to the aircraft, without radial gaps, are used to produce a coarse analysis that is the sum of a mean flow at each vertical level, plus a field of radial, tangential and vertical wind that only varies in height and radius. That field is then used to de-alias all available data to produce a second, finer, analysis with wavenumber 0 and 1 azimuthal Fourier components for each height and radius. The second analysis is used to do the final de-alias of the data, which are then interpolated to a three-dimensional grid to produce a Cartesian analysis. That analysis is a least-squares fit to the Doppler radial velocities and the three-dimensional anelastic continuity equation. In addition to the three-dimensional wind field, vertical profiles of wind speed, radial wind, and vertical wind are computed without the continuity equation by assuming no azimuthal variation near the flight track. The absence of a continuity assumption permits finer detail in these vertical profiles.

As soon as the last quality control of Doppler radials is complete, scripts developed by RSS and AOC radially average bins of the quality-controlled Doppler radial velocity and transmit them to the NWS gateway for assimilation into HWRF. One the analyses are finished a separate script developed at AOC transmits the wind analyses and superobs (for HRD and its research partners) to AOC.

Occasionally in weak, highly-sheared systems, the de-aliasing scheme will still fail in portions of the analysis. This is, however, now the exception rather than the rule. HRD continues to work to reduce these errors. Such de-aliasing errors occurred in at least one analysis of Hurricane Kate (2015), and we hope to report on changes to the QC procedure that eliminated these.

One further feature of the process we would like to accomplish before the next season is to automatically post graphics of Doppler wind analyses on the AOML web site. At the conference we will present examples of the job-file java application, the quality-controlled data, analyses, as well as examples on the HRD web site of analyses produced and posted using real-time scripts. The automation permits the analysis of the large historical airborne-Doppler data set to produce composite and statistical analyses. We will also discuss the quality of the research analyses.