Assessing Variable Importance in a Radar QC Algorithm

Recently, a radar data quality control algorithm has been devised to discriminate between weather echoes and echoes due to non-meteorological phenomena such as bioscatter, instrument artifacts and ground clutter using the values of polarimetric moments at and around a range gate. Because the algorithm was created by optimizing its weights over a large reference data set, statistical methods can be employed to examine the importance of the different variables in the context of discriminating between weather and no-weather echoes. Among the variables studied for their impact on the ability to identify and censor non-meteorological artifacts from weather radar data, the method of successive permutations ranks the variance of Zdr, the reflectivity structure of the virtual volume scan and the range derivative of PhiDP (Kdp) as the most important.