Abstract : J.2
Evaluation of a polarization radar-based hydrometeor classification algorithm through the analysis of microphysical data in an orographic environment

David Plummer, Sabine GŲke, Scott Ellis, Jothiram Vivekanandan
Dept. of Atmospheric Sciences, University of Illinois at Urbana-Champaign

The polarization radar-based hydrometeor classification algorithm developed by Vivekanandan et al (1999) has been evaluated using radar and in situ data from the Mesoscale Alpine Program (MAP), a field project involving observations of orographic precipitation systems on the Mediterranean side of the Alps. The algorithm was implemented on the National Center for Atmospheric Research (NCAR) S-Pol dual-polarization radar, which operated during MAP. In addition, microphysical observations were made within the vicinity of the S-Polís scans by the NCAR Electra aircraft during a number of MAPís Intensive Observation Periods.

The goals of this study were to characterize the polarimetric signatures of cold cloud particles during MAP and to improve the hydrometeor classification algorithmís output through the development of a confidence field. For certain hydrometeors, the polarimetric quantities have considerable overlap. The algorithm was improved for these situations by the determination of confidence levels. These provide a measurement of the certainty of the classifications, providing the user with more information than the particle type alone.

The airborne in situ particle observations were compared to the radar data and hydrometeor classifications at locations closely matched in time and space. To accomplish this, an automatic algorithm was developed to locate closely matched data points between the radar and in situ datasets. This took into account the substantial differences in sampling methodology between these two types of observations as well as the time difference over which the in situ observations may be regarded as representative of the radar observations. The procedures and results of evaluating the particle classification algorithm, as well as considerations regarding the comparison of two significantly different datasets, will be presented at the conference.