Space environment effects on geostationary spacecraft: Analysis and prediction

Abstract

We investigate how the two geostationary spacecraft, Meteosat-3 and Tele-X, are affected by the space environment, as characterized by Dst, Kp, and energetic electron flux (> 2 MeV, EEF). Neural network models are developed to predict spacecraft anomalies 1 day ahead. It is found from superposed epoch analysis that the spacecraft anomalies frequently occurred during the recovery phase of geomagnetic storms and that the space environment during the last 4–6 days preceding an anomaly contributes statistically the most to the anomaly occurrence. We find from network prediction (1) that when training is only on Meteosat-3, Kp, Dst, and the EEF, respectively, give the total prediction rate of 79%, 73%, and 62% for Meteosat-3 (both anomalies and non-anomalies) and the total prediction rate of 64%, 66%, and 67% for Tele-X; (2) that when training is only on Tele-X, Kp, Dst, and the EEF, respectively, give the total prediction rate of 57%, 58%, and 51% for Meteosat-3 and the total prediction rate of 70%, 71%, and 74% for Tele-X; (3) that Kp is the optimal environment parameter to correlate with the anomalies on both Meteosat-3 and Tele-X. It is revealed from the anomaly local time distribution, superposed epoch analysis, and network prediction that Meteosat-3 was more subject to surface charging than to internal dielectric charging, and that Tele-X was more subject to internal dielectric charging than to surface charging. The developed neural network models can be used to predict times with higher risks for anomalies in real-time from ACE data.