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

Combined BiLSTM and ARIMA models in middle- and long-term polar motion prediction

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

Precise prediction of polar motion using sliding multilayer perceptron method combining singular spectrum analysis and autoregressive moving average model

The CNN-LSTM-attention Model for short term prediction of the Polar Motion

Medium-short-term prediction of polar motion combining the differencing between series with the differencing within series

A new medium-long term polar motion prediction method based on sliding average within difference series

Polar Motion Ultra-Short-Term Prediction of Least-Squares+Multivariate Autoregressive Hybrid Method by Using the Kalman Filter

Prediction of Satellite Time Series Data Based on Long Short Term Memory-Autoregressive Integrated Moving Average Model (LSTM-ARIMA)

A Hybrid Wind Speed Prediction Approach Based on Ensemble Empirical Mode Decomposition and BO-LSTM Neural Networks for Digital Twin

Analysis of watershed terrestrial water storage anomalies by Bi-LSTM with X-11 time series prediction combined model

A new approach to improve the Earth's polar motion prediction: on the deconvolution and convolution methods

A hybrid model for TEC prediction using BiLSTM and PSO-LSSVM

Monthly Load Forecasting Based on an ARIMA-BP Model: A Case Study on Shaoxing City

The Short-Term Prediction of Low-Latitude Ionospheric Irregularities Leveraging a Hybrid Ensemble Model

Multistep-Ahead Prediction of Ocean SSTA Based on Hybrid Empirical Mode Decomposition and Gated Recurrent Unit Model

Application of hybrid model based on empirical mode decomposition, novel recurrent neural networks and the ARIMA to wind speed prediction

Short-term prediction of celestial pole offsets with interpretable machine learning

A BiLSTM Hybrid Model for Ship Roll Multi-Step Forecasting Based on Decomposition and Hyperparameter Optimization

Prospects of Predicting the Polar Motion Based on the Results of the Second Earth Orientation Parameters Prediction Comparison Campaign

Land subsidence analysis along high-speed railway based on EEMD-Prophet method

Short-term motion prediction of a semi-submersible by combining LSTM neural network and different signal decomposition methods