A Dynamic Short-range Animal Migration Forecast Model Based on Weather Radar Network

Predicting aerial animal migration has great implications for biological research and conservation, and agricultural production. Animal migration is deeply coupled with spatio-temporal and meteorological factors, however, the existing forecast model is only based on the spatio-temporal features of animal migration and the impact of meteorological factors on migration is not considered. In this paper, we propose a dynamic forecast model to forecast airborne migration from the perspective of the influence of wind field data on migration and makes prediction with the help of neural network. Specifically, we calculate the transition probability matrix using the wind data and the migration spatial field to represent the migration dynamic model, and integrated this model into deep neural network. We use the data of China's weather radar network to verify that our model outperforms the competing methods. In addition, the cause of prediction error at peak value is analysed by visualization results. In future applications, more meteorological factors should be considered to improve the representation of dynamic models.