Prediction of the Polar Motion and Length of Day Time Series

In this paper, a statistical study of the properties of forecasts of the Earth's rotation parameters (EOP): the coordinates of the pole, the length of day LOD and the Earth’s rotation velocity represented by the discrepancy between the Universal time UT scale and the Universal time coordinated UTC atomic time scale. The daily forecasts were carried out in the framework of the international EOPC PPP (Earth Orientation Parameters of Prediction Pilot Project) competition, held in the USNO (U.S. Naval Observatory) since 2011 year. The initial data for the forecasts was taken from the EOP C04 bulletin containing combined solutions for the Earth rotation parameters. We used a special version of the bulletin, containing information up to the current date. The forecasts were built for 90 days in the future. The statistics were based on comparison of the forecasts with the C04 series, obtained later. The predictions were performed using autoregressive (AR), least-squares collocation (LSC), neural networks (NNET). Shanghai Observatory (SHAO) predictions were based on the least squares and AR methods. The combination (COMB) of all four series demonstrated the best quality. The prediction methods have been developed and described using C ++ language. The processing of forecasts has been implemented using a program written in the MATLAB script language. The statistics have been represented in the graphical form by plots and color maps.