Abstract: Sea clutter is a common non-precipitation echo in coastal areas that can adversely affect precipitation estimation. Using the LightGBM (Light Gradient Boosting Machine) algorithm and based on the S-band weather radar observation data from Wanning and Taizhou, three sea clutter identification models were developed for dual-polarization radar and Doppler radar: a model based on the combination of polarization parameters, a model based on a single polarization parameter, and a nonpolarization model applicable to conventional Doppler radar. Testing results show that the three models achieved Critical Success Index (CSI) values of 0.9876, 0.9660, and 0.7605, respectively. All models effectively identified extensive sea clutter in Wanning while preserving precipitation echoes. Similar performance was achieved when testing with radar data from Zhoushan and Fuzhou, which were not included in the training dataset, demonstrating the models' strong generalization capability.