A Study on Insect and Bird Echo Recognition Using Machine Learning-Based X-Band Dual-Polarization Weather Radar

Weather radars have shown great potential in the field of aerial ecological monitoring. Leveraging a network of weather radars allows for the analysis of insect and bird migration patterns on a continental scale, thus better protecting bird habitats and preventing pest outbreaks. However, as the radar beam propagates, the observed altitude increases with distance, resulting in a decrease in low-altitude monitoring coverage by the weather radar network. To address this issue, China's new generation weather radar network (CINRAD) has extensively deployed X-band dual-polarization radars. However, the application of operational X-band radars for ecological monitoring is unprecedented, and the impact of wavelength on the scattering characteristics of insects and birds is unknown. To solve this problem, this study collects X-band radar echo datasets of insects and birds through joint observations with S-band and X-band radars, and further develops a machine learning-based algorithm to identify insect and bird echoes. By training the model to converge, the final recognition accuracy reached 83.11%. The algorithm was applied to two typical cases, confirming its robustness.