Medium- and Long-Term Prediction of Polar Motion Using Weighted Least Squares Extrapolation and Vector Autoregressive Modeling

ABSTRACT This article presents the application of weighted least squares (WLS) extrapolation and vector autoregressive (VAR) modeling in polar motion prediction. A piecewise weighting function is developed for (LS) adjustment consideration effect intervals between observation prediction epochs on WLS extrapolation. Furthermore, VAR technique used to simultaneously model predict residuals x p , y pole coordinates misfit. The simultaneous predictions are subsequently computed by combination harmonic models linear trend, Chandler annual wobbles, stochastic (WLS+VAR). 365-day-ahead compared with those generated LS extrapolation+univariate AR extrapolation+VAR modeling. It shown that based WLS+VAR taking into both interval correlation outperform two others. accuracies 13.97 mas, 18.47 20.52 respectively 150-, 270-, 365-day horizon terms mean absolute error statistics, 36%, 24.8%, 33.5% higher than LS+AR, respectively. For predictions, 15.41 21.17 21.82 mas respectively, 27.4%, 11.9%, 21.8% LS+AR Moreover, differences observations smaller from LS+VAR which practically important practical scientific users, although improvement no more 10% relative LS+VAR. further comparison submitted 1st Earth Orientation Parameters Prediction Comparison Campaign (1st EOP PCC) shows while accuracy within 30 days comparable most accurate techniques including neural networks participating campaign coordinates, up 365 future better other except best PCC. therefore concluded medium- long-term can be improved together.