Nanjing University of Information Science and Technology
In recent years, deep learning models has been increasingly used in solving nowcasting problems that have a large impact on disaster prevention and mitigation. In this paper, we took nowcasting as a spatio-temporal sequence prediction task, and use radar reflectivity factor as the test object. We use TAGAN deep learning model based on GAN frame to predict the radar echo image of the future 1h, and compared it with Rover optical flow Method and 3DUnet model based on convolutional neural network. The radar echo data set of the 2018 Global Weather AI Challenge is selected for training and testing. The test results show that the TAGAN model shows advances by multiple scores such as the hit rate (POD), false alarm rate (FAR), critical success index (CSI), and correlation coefficient. The TAGAN model performs well in the above test scores and increases with the prediction time compared to the traditional optical flow method and the comparative deep learning model. Compared with the traditional optical flow model, the improvement effect is more significant. The results may shed some light on expanding and improving the application of deep learning models in near-weather forecasting.