Characteristics of aggregate snowflakes, derived from triple-wavelength Doppler radar measurements

Radar is a powerful tool to probe the microphysics of ice clouds and snowfall, but the information which can be retrieved using a single radar wavelength is limited. This is because the particle size spectrum is described by at least two parameters, whereas only one measurement is available to constrain it. In theory, dual-wavelength measurements can overcome this problem, provided that the particles are large enough. However in practice one must still make assumptions about the density of the particle (mass-size relationship), and the scattering model, neither of which are known a-priori.

Recently, a handful of studies (Kneifel et al 2011, Leinonen et al 2012, Kulie et al 2014) have suggested that adding a measurement at a third wavelength may allow these assumptions to be better constrained. However very little triple-wavelength data is available to test these ideas, and the data that is available suffers from imperfect colocation, poor sensitivity at the longest wavelength, and uncertain calibration/attenuation biases.

Results are presented from a new field experiment at the Chilbolton Observatory in the UK in which high-resolution measurements were collected simultaneously from three colocated radars at 3.2mm, 8.6mm and 9.75cm in fifteen deep ice clouds. The sensitivity of all three radars is high (approx -40dBZ at 1km), and the ground based operation allows for longer integrations and very accurate Doppler. Full Doppler spectra were recorded at all 3 wavelengths, and we demonstrate how uncertainties in attenuation or calibration at the shorter wavelength can be overcome using the Doppler spectra, by matching the reflectivity from the (slow-falling) Rayleigh-scattering particles to the data collected at the longest wavelength, which is absolutely (and independently) calibrated to within 0.5dB.

Initial analysis shows that the observations are inconsistent with recent discrete-dipole scattering models for pristine crystals such as bullet rosettes, columns or dendritic crystals (Liu 2008), but instead support a fractal model of aggregate snowflakes (Westbrook et al 2006, Hogan and Westbrook 2014). Meanwhile, measurements of the spectrally-decomposed dual-wavelength ratio as a function of fall speed, when combined with recent improvements in our understanding of ice particle terminal velocity (Heymsfield and Westbrook 2010), allow us to infer the density of the aggregates as a function of their size, and these results are consistent with aggregate snowflakes which decrease in density as they grow larger in size. We also show that the "saturation" of dual-wavelength ratio observed by Leinonen et al (and in our study) for 3.2/8.6mm wavelengths when the snowflakes are large contains information about the fractal dimension of the aggregates. Finally, off-zenith dwells are used to test whether the aggregates are preferentially oriented along their long dimension, based on measurements of the differential reflectivity and copolar correlation coefficient.