Over-sampling of radial velocity and 3dvar analysis of dual-doppler observations

For the detection of severe weather phenomena such as tornado, mesocyclone and strong wind shear, the azimuthal resolution of radial velocity measurements is most important. The typical azimuthal resolutions of one degree for the WSR-88D radars and of two degrees for the planned CASA (Center for Collaborative Adaptive Sensing of Atmosphere) radars are not sufficient for this purpose, especially at far ranges. Over-sampling is one strategy that can potentially provide more details about the azimuthal structures of flows, by collecting data at azimuthal increments smaller than the radar beam width. In the presence of dual-Doppler observations, variational method can be effectively used to recover sub-beam-width structures from over-sampled data, which, combined with the excellent radial resolutions, can provide high-resolution wind analyses that are valuable for, e.g., tornado detection. This idea is tested in this paper using simulated as well as re-processed Level-I data from a research WSR-88D radar located in Norman (KOUN).

For the simulated data tests, radial velocity data are sampled from a 25-m resolution simulation of an F4 to F5 intensity supercell tornado, using a realistic radial velocity emulator. It is shown that our variational scheme is indeed able to recover sub-beam-scale structures from the over-sampled dual-Doppler velocity data, and the tornado is much better resolved in the over-sampled analysis.

For real data we reprocessed the Level-I data from KOUN, for the May 10, 2003 central Oklahoma tornadic thunderstorm case. Over-sampling is achieved by processing the same number of pulse returns as for the regular non-oversampled data; the number of the pulse returns that differ between consecutive samples is equal to the number of returns received over the azimuthal over-sampling increment. Such a processing can be possible in real-time at the radar site without any change to the radar scanning strategy, and the procedure preserves the signal to noise ratio of the data. Azimuthal increments of 0.125, 0.25 and 0.5 degree are tested. The over-sampled data are analyzed together with regular radial-velocity data from the Oklahoma City WSR-88D radar (KTLX). Again, much better low-level wind analyses are obtained, and the analysis is the best with the highest over-sampling rate. Over-sampled data provide significant value added when the information is recovered through variational analysis. Future collection of over-sampled data using, e.g., CASA radars, is planned to further explore this method.