A simple-but-realistic fuzzy logic hydrometeor classification scheme for the French X, C and S-band polarimetric radars

The ability to perform hydrometeor classification at high space-time resolution (5 minutes – 1 km²) in precipitating systems is one of the main benefits brought by polarimetry. In this paper, we revisit the fuzzy logic hydrometeor classification with the aim to make it more realistic (with respect to actual radar measurements), simpler and more efficient compared to previous approaches. We start – and this first step is completely independent from the wavelength at which the algorithm will be run - by defining a limited number of hydrometeor types that the algorithm will try to discriminate. This leads to: rain, melting snow, dry snow / crystals, small hail (diameter < 5 mm), medium hail (diameter between 5 – 20 mm) and large hail (diameter > 20 mm). Adding more hydrometeor types at this stage would not be realistically compatible with the known discriminating power of the polarimetric variables collected operationally. The second step consists in establishing empirical 2D membership functions (2DMBF) for (Z_H, Z_DR), (Z_H, K_DP) and (Z_H, ρ_HV) in so-called “ideal conditions”. The membership functions are segregated by hydrometeor type and by wavelength (S, C and X band). A large number of radar data collected by the French polarimetric radars (9 C-band, 1 S-band, 1 X-band) are used in that step and areas of homogeneous hydrometeor types are determined subjectively by an expert. This constitutes the calibration and validation dataset. “Ideal conditions” refers here to low attenuation (Φ_DP < 10°), high Signal-to-Clutter Ratio (SCR > 15 dB), high Signal-to-Noise Ratio (SNR > 15 dB), low Partial Beam Blocking (PBB < 10%) and short distances (d < 60 km). The empirical 2DMBF are then modelled by using two half-Gaussians (an upper one and a lower one) per class of Z_H (5 dBZ increment). In a next step, the dependency of the modelled 2DMBF with the actual measurements conditions (Φ_DP, SCR, SNR, PBB, distance, …) are analysed and parametrized in a simple way. The fuzzy logic algorithm is then applied as follows: for each polar pixel, the actual measurement conditions are taken into account to generate modified 2DMBF from the ideal 2DMBF. The resulting modified 2DMBF values for (Z_H, Z_DR), (Z_H, K_DP) and (Z_H, ρ_HV) are calculated and added for each hydrometeor type. The result is multiplied by a 1D Z_H-dependent MBF and a 1D temperature-dependent MBF. The hydrometeor with the highest score is considered as the dominant one. The second highest and its score are also considered as a measure of the quality of the identification. The validation of the algorithm is done using an independent dataset. The influence of temperature is shown to be extremely important in the case of frontal cases.