Thursday, 17 September 2015
Oklahoma F (Embassy Suites Hotel and Conference Center )
During the Dynamics of the Madden-Julian Oscillation (DYNAMO) field campaign in 2011-2012, the NASA Tropical Oceans Global Atmosphere (TOGA) C-band Doppler radar was integrated onto the R/V Roger Revelle for a period of 4 months. During this time, the radar obtained quality Doppler velocity measurements within a variety of tropical convection. During mesoscale precipitation events, significant velocity aliasing occurred due to the strong low-level jets that sustained these systems combined with the relatively low Nyquist velocity (13.4 m s-1) of the TOGA radar. In order to prep the TOGA data for assimilation into a limited-domain model, there was a need to automate the dealiasing of the Doppler winds due to the imposing size of the dataset. The Four-Dimensional Dealiasing (4DD) algorithm, originally developed at the University of Washington and currently implemented within the Python Atmospheric Radiation Measurement (ARM) Radar Toolkit (Py-ART) software, was first used. However, after many attempts the results were not entirely satisfactory and thus significant effort was put toward hand-dealiasing TOGA radar volumes, one volume every 30 minutes for 3 separate ~3-week cruises. This large dataset fortuitously proved to be a useful validation tool for future automated dealiasing algorithms. Two such algorithms, Region-Based and Phase-Unwrap, have since been implemented in Py-ART. In this study we report on a statistical analysis of the performance of these three distinct automated methodologies against the large hand-dealiased dataset. Metrics such as probability of detection (i.e., correctly unfolded a folded gate), false-alarm rate (i.e., unfolded a gate that was not folded to begin with), and critical skill index are being determined for various implementations of these three algorithms. In addition, the effects of dealiasing errors on applications such as radar data assimilation and single-Doppler retrievals of low-level 2D winds will be assessed. On the basis of this, the relative strengths and weaknesses of each of the dealiasing algorithms will be compared, and recommendations for their usage on other radar datasets will be made.
- Indicates paper has been withdrawn from meeting
- Indicates an Award Winner