A high resolution hydrometeor phase classifier based on analysis of cloud radar doppler spectra

A recently developed method of hydrometeor phase classification based on automatic analysis of millimeter wavelength cloud radar (MMCR) Doppler spectra has demonstrated great potential as a reliable tool for revealing cloud phase structure with high spatial and temporal resolution. A limited cloud phase classification dataset derived from co-located high spectral resolution lidar (HSRL), microwave radiometer (MWR), and MMCR measurements was used for training the classifier. Thereafter, the only input to the classifier is MMCR spectra, and the output, hydrometeor phase (liquid, solid, mixed). The method accomplishes class separation by extracting morphological information from Doppler spectra in greater detail than that provided by low order moments, and analyzing it with a neural network. The Mixed-Phase Arctic Cloud Experiment (M-PACE) conducted at the Atmospheric Radiation Measurement (ARM) program's North Slope of Alaska (NSA) ARM Climate Research Facility (ACRF) brought together the HSRL, MWR, and MMCR instruments that generated the month-long dataset used for training and evaluating our classifier. We show examples of coherent hydrometeor phase structures in arctic clouds as decoded by this method into liquid, solid, and mixed-phase classes at a resolution of 5 seconds and 45 vertical meters. We also provide our best quantitative estimates of the method's accuracy.