A hybrid algorithm for ship clutter identification in pulse compression polarimetric radar observations

Abstract. With the rapid development of active-phased arrays and solid-state transmitters, pulse compression technology has become increasingly important. Currently, pulse compression waveforms with peak sidelobe levels better than −50 dB have been developed, enabling the broader application of pulse compression technology in weather radar systems. However, existing sidelobe suppression levels are still insufficient to ensure that radar data quality is unaffected by range sidelobes for ship clutter, which have a high echo intensity and cannot be removed by conventional quality control methods. In this study, we introduce a hybrid ship clutter identification (HSCI) algorithm to address this issue in pulse compression polarimetric radar observations. The HSCI algorithm comprises two parts: mainlobe and sidelobe identification (including the range and antenna sidelobes). Mainlobe identification uses a random forest model that integrates multiple features to identify the mainlobe of ship clutter. Sidelobe identification uses a series of heuristic criteria derived from the statistical characteristics of ship clutter to distinguish them from precipitation echoes. The analysis results of two typical cases indicate that after implementing the HSCI algorithm, the impact of ship clutter on radar data is visually imperceptible. The statistical results show that the HSCI algorithm achieves a promising performance in ship clutter mainlobe identification on a test dataset comprising 400 ship clutter gates and 2500 range gates of precipitation echoes, with the precision, recall, and F1-score all exceeding 97 %. Application of this algorithm to the University of Helsinki C-band dual-polarization Doppler weather radar data successfully reproduced ship tracks in the Gulf of Finland.