An automatic identification of clutter and anomalous propagation in polarization-diversity weather radar data using neural networks

Radar polarization measurements have mostly been used to improve rainfall estimation and hydrometeor characterization. The authors extend the use of such measurements to the problem of ground clutter recognition, including the case when this problem is associated with anomalous propagation of the electromagnetic wave. They present a methodology used for recognizing both clutter and meteorological targets. The methodology is based on the knowledge of the scattering properties of the targets, as provided by the polarization measurements and the use of the neural network approach that performs the classification. The results show that if circular polarization is used, the circular depolarization ratio and the degree of polarization are good discriminators of clutter and nonclutter. They have used data from the Alberta polarization diversity radar to build an automatic decision process using a feedforward neural network. After they trained the neural network, they tested the classifier for two common clutter situations: when there is an electromagnetic wave anomalous propagation and when targets from rain are mixed with the clutter close to the radar.