Abstract UT1-UTC predictions especially short-term predictions are essential in various fields linked to reference systems such as space navigation and precise orbit determinations of artificial Earth satellites. In this paper,… Click to show full abstract
Abstract UT1-UTC predictions especially short-term predictions are essential in various fields linked to reference systems such as space navigation and precise orbit determinations of artificial Earth satellites. In this paper, an integrated model combining the grey model GM(1, 1) and neural networks (NN) are proposed for predicting UT1-UTC. In this approach, the effects of the Solid Earth tides and ocean tides together with leap seconds are first removed from observed UT1-UTC data to derive UT1R-TAI. Next the derived UT1R-TAI time-series are de-trended using the GM(1, 1) and then residuals are obtained. Then the residuals are used to train a network. The subsequently predicted residuals are added to the GM(1, 1) to obtain the UT1R-TAI predictions. Finally, the predicted UT1R-TAI are corrected for the tides together with leap seconds to obtain UT1-UTC predictions. The daily values of UT1-UTC between January 7, 2010 and August 6, 2016 from the International Earth Rotation and Reference Systems Service (IERS) 08 C04 series are used for modeling and validation of the proposed model. The results of the predictions up to 30 days in the future are analyzed and compared with those by the GM(1, 1)-only model and combination of the least-squares (LS) extrapolation of the harmonic model including the linear part, annual and semi-annual oscillations and NN. It is found that the proposed model outperforms the other two solutions. In addition, the predictions are compared with those from the Earth Orientation Parameters Prediction Comparison Campaign (EOP PCC) lasting from October 1, 2005 to February 28, 2008. The results show that the prediction accuracy is inferior to that of those methods taking into account atmospheric angular momentum (AAM), i.e., Kalman filter and adaptive transform from AAM to LODR, but noticeably better that of the other existing methods and techniques, e.g., autoregressive filtering and least-squares collocation.
               
Click one of the above tabs to view related content.