An Enhancing Biological Echo Classification Method with Polarimetric Weather Radar Using Random Forest Classifiers

Polarimetric weather radar can enhance precipitation and biological scatter monitoring by measuring the polarization moments of targets, thus providing valuable shape information. This study employed five Random Forest classifiers to classify birds, insects, and precipitations at the five lowest elevation angles. The training and testing sets were artificially generated using typical cases of birds, insects, and precipitations captured by the polarimetric weather radar. A two-dimensional median filter was employed to reduce the volatility of polarization moments. To mitigate the impact of differential phase jumps, the centre of the probability distribution function for the differential phase in each volume scan was fixed at 180 degrees. The classification accuracy for each elevation angle and sample type exceeded 91%. The performance of the classifiers was assessed using two representative cases.