CLUTTER DETECTION AND MITIGATION FOR DUAL-POLARIZATION WEATHER RADAR

Abstract

Ground clutter in weather radar observations causes degradation of data quality and can lead to misinterpretation of radar echoes. It is important to detect clutter and mitigate its effects to obtain accurate weather measurements. The focus of this study is to improve the performance of clutter detection algorithms by presenting different discriminant functions. A Bayesian classifier is used to make an optimal decision based on discriminant functions to detect clutter mixed with weather echoes. The conditional probability density functions for clutter and weather signals may change and may need to be updated due to changing weather conditions, clutter, and radar parameters. Therefore, to make it more efficient, a multivariate Gaussian mixture model is presented to parametrize discriminant functions and reduce the complexity of detection algorithms. The model parameters are estimated based on the maximum likelihood, using the Expectation-Maximization (ML-EM) method. A dual-polarization clutter filtering algorithm is also presented to mitigate ground clutter effects on weather radar measurements. A multivariate Gaussian model is introduced to parametrize clutter and weather power spectrums, and the Maximum A Posterior (MAP) method is used to estimate weather components. Instead of using a random phase, the phase of the retrieved weather spectrum is estimated based on the statistical properties of dual-polarization weather signals. The performance of the clutter detection and filtering algorithms are shown by applying them to the radar data collected by the national WSR-88D (KOUN) polarimetric radar and are compared to existing detection and filtering algorithms. It is shown that the proposed algorithms can effectively mitigate clutter effects and substantially improve polarimetric weather radar data quality.