Analysis of radar Doppler spectra shape—an insight into Cloud Microphysics

Recent advancements in radar digital receiver technology, fast computers and low-cost data storage allow for the continuous recording and storage of the radar Doppler spectra from a wide range of radars (e.g., cloud radars, profilers and weather radars). The recorded Doppler spectra are typically used for filtering spurious radar signal artifacts, identifying clutter, turbulence and microphysical retrievals. Typically, the three first moments of the Doppler spectrum (reflectivity, mean Doppler velocity and Doppler width) are used to describe its shape and used as input into retrieval algorithms. This is based on the assumption that the shape of the Doppler spectrum can be approximated as Gaussian. Examination of Doppler spectra skewness, kurtosis, multimodality, local gradient and curvature and other asymmetry parameters revealed that the Doppler spectra are often highly non-Gaussian and that these asymmetry parameters are highly correlated in space. These new parameters were used as input into a neural network that was trained with the help of a cloud phase (e.g., liquid, mixed, ice) training dataset from the Mixed – Phase Artic Cloud Experiment (MPACE). The neural network shows remarkable ability to generalize and identify cloud phase in the broader dataset. This suggests that important microphysical information is contained in the shape of the Doppler spectrum, often not captured by the standard Doppler moments. These new parameters that describe the shape of the Doppler spectrum and its deviation from the Gaussian shape could find application to other areas of atmospheric research with radars such as wind profilers and operational weather radars.