Snowflake Size Spectra Retrieved from a UHF Vertical Profiler

Abstract

Abstract This paper develops a technique for retrieving snowflake size distributions (SSDs) from a vertically pointing 915-MHz vertical profiler. Drop size distributions (DSDs) have been retrieved from 915-MHz profilers for several years using least squares minimization to determine the best-fit DSD to the observed Doppler spectra. This same premise is used to attempt the retrieval of SSDs. A nonlinear search, the Levenberg–Marquardt (LM) method, is used to search the physically realistic solution space and arrive at a best-fit SSD from the Doppler spectra of the profiler. The best fit is assumed to be the minimum of the squared difference of the log of the observed and modeled spectrum power over the precipitation portion of the spectrum. A snowflake video imager (SVI) disdrometer was collocated with the profiler and provided surface estimates of the SSDs. The SVI also provided estimates of crystal type, which is critical in attempting to estimate the density–size relationship. A method to vary the density–size relationship during the event was developed as well. This was necessary to correctly scale the SVI SSDs for comparison to the profiler-estimated distributions. Five events were examined for this study, and good overall agreement was found between the profiler and SVI for the lowest profiler gate (225 m AGL). Vertical profiles of SSDs were also produced and appear to be physically reasonable. Uncertainty estimates using simulated Doppler spectra show that the retrieval uncertainties are larger than that for rainfall and can approach and exceed 100% for situations with large spectral broadening as a result of atmospheric turbulence. The larger uncertainties are attributed to the lack of unique Doppler spectra for quite different SSDs, resulting in a less well-behaved solution space than that of rainfall retrievals.