Study on Weather Radar Echo Data Generation Based on DCGAN

Abstract

Doppler weather radar can detect the changes in precipitation clouds for short-term forecasting. In the process of development of Doppler weather radar and weather identification algorithms, some typical Doppler weather radar base data corresponding to different weather phenomena are necessary for signal processing unit test and algorithm verification. However, the existing real weather radar base data with high quality can't meet the requirement in amount. In this paper, an algorithm based on Deep Convolutional Generative Adversarial Networks (DCGAN) to generate typical weather radar base data is proposed. And in the test signal simulation step, the power spectrum algorithm is improved. The results show that the data produced by the DCGAN have the same characteristics with the real weather radar base data without obvious non-meteorological noise. Moreover, the improved power spectrum algorithm performs better in terms of accuracy rate of the simulation echo signal.