Predicting Avian Frequency Based on Weather Radar and Citizen Science

Predicting aerial animal migration holds significant importance for the study and conservation of wildlife, particularly in guiding protective measures for various bird orders. Existing large-scale prediction models utilize weather radar data. However, weather radar cannot accurately distinguish the detected animals, and thus cannot target specific bird orders. This paper first analyses the correlation between weather radar data and citizen science (eBird) data, then proposes a deep learning model to predict eBird avian frequency based on weather radar data. The results indicate that in the Bohai Bay area across three provinces, the correlation between avian migration traffic estimated by weather radar and eBird frequency reached as high as 0.837, demonstrating the feasibility of using radar data to predict eBird frequency. In the prediction experiments, the proposed model achieved a correlation of 0.782 between the predicted results and the actual values, validating the effectiveness of our model. Future applications will focus on enhancing the performance of the prediction method and incorporating additional data sources, such as flight call record, to supplement the limitations of manual bird report.