Estimation of raindrop size distribution from polarimetric radar measurements at C-band

Abstract

Many studies have proposed methods to estimate the raindrop size distribution (DSD) parameters from polarimetric radar data as a part of rain attenuation correction and/or rainfall rate estimation algorithms, in which a modified gamma distribution model has been often used to characterize the natural variation of DSD. The parameters that determine the modified gamma DSD are a shape parameter μ, median volume diameter D0 or slope parameter Λ, and a number concentration N0 or its normalized version NW. While D0 (or Λ) and NW can be retrieved relatively straightforward from the polarimetric radar measurements, estimation of shape parameter is not an easy task. Instead, empirical relations including μ-Λ and/or μ-D0 relations derived from surface measurements of DSD are widely used to estimate μ implicitly.   Adachi et al. (2015) proposed a method to estimate the three parameters of the DSD from polarimetric radar data without any assumptions of relationship among the parameters. In that method, they assumed a constant shape parameter in a range profile. However, this assumption may not be satisfied if the radar is sampling mixed convective/stratiform echoes that simultaneously exist in a single profile. Theoretically, on the other hand, a shape parameter can be estimated from a correlation coefficient ρHV at each range gate (e.g., Thurai et al. 2008). However, to estimate shape parameters with a method of this kind, it is necessary to obtain a correlation coefficient with quite high accuracy, for which a very long sampling time is needed to apply it to radars to satisfy this condition for most radar measurements. The MRI C-band polarimetric radar is equipped with solid-state transmitters, which enable the radar to make observations of correlation coefficient with high accuracy in a relatively short time, especially in high-SNR regions. Thus, we have developed an algorithm to estimate a shape parameter at each range gate both from correlation coefficient and differential reflectivity along with D0 and NW, and compared it with surface measurements.