中国团队利用AI提升南山射电望远镜大气修正精度

Core Insights - The research addresses the significant issue of tropospheric delay in electromagnetic wave propagation due to variations in air density and water vapor content, which affects Very Long Baseline Interferometry (VLBI) and Global Navigation Satellite Systems (GNSS) positioning [1][3] - A hybrid deep learning model combining Gated Recurrent Unit (GRU) and Long Short-Term Memory (LSTM) networks has been developed to accurately predict zenith tropospheric delay (ZTD) [1][3] Research Findings - The team conducted spectral analysis on years of GNSS observations from the Nanshan station, revealing that ZTD changes exhibit clear annual and semi-annual cycles, with higher values in summer and lower in winter, closely related to temperature and water vapor content [3] - The hybrid neural network model effectively captures both short-term fluctuations and long-term trends in atmospheric delay, achieving a prediction error of approximately 8 millimeters and a correlation coefficient of 96%, outperforming traditional statistical models and single neural networks [3] Applications and Implications - High-precision predictions of tropospheric delay can significantly enhance the atmospheric phase correction accuracy in VLBI observations, improving radio source positioning and baseline calculation results [3] - The research demonstrates the potential of artificial intelligence in atmospheric correction for radio telescopes, laying a technical foundation for the future operation of the QTT 110-meter telescope and multi-station interferometric observations in high-frequency bands [3]