The Sensitivity Analysis of the Fuzzy Logic Classification Algorithm of Precipitation Particles for X-band Dualpolarization Radar

The polarization radar precipitation particle classification algorithm based on fuzzy logic is widely concerned and applied because of its stability and ease of use. However, due to the characteristics of the algorithm itself, the data quality of the input parameters (radar polarimetric variables: ZH, ZDR, KDP and ρHV) of the algorithm has a great impact on the classification results, which can be quantitatively evaluated through simulation experiment to analyze the sensitivity of precipitation particle fuzzy logic classification algorithm to the deviation of input variables. In this paper, we assume the measurement data of radar polarimetric variables as the true value, then introduced bias of ± 3 dBZ (in an interval 0.25 dBZ) for ZH, ± 1 dB (in an interval 0.25 dB) for ZDR, $\pm 1.25^{\circ} / \mathrm{km}$ (in an interval $0.25^{\circ} / \mathrm{km}$) for KDP, and -0.1 (in an interval 0.01) for ρHV respectively. By studying a case of typical stratiform cloud precipitation, the sensitivity of the precipitation particle classification results to the deviation of polarimetric variables has been analyzed using the method of the control variable test in a quantitative way, which can be evaluated by the percentage of identified precipitation particle distribution. The results show that, for ZH and KDP, the proportional distribution change of the output type of the algorithm is not obvious, however, for ZDR, distribution in the classification of the liquid precipitation particles has relatively obvious change. As for ρHV, the classification algorithm is very sensitive to the change of the deviation of ρHV.