Analysis of periodic patterns, noise characteristics, and predictive modeling of polar motion and length of day

Abstract

Earth orientation parameters (EOP) are critical for applications in orbit determination, astronomy, space geodesy, and geophysics. Accurate predictions of EOP rely on the identification of both deterministic periodic patterns and noise characteristics. This study addresses these requirements by analyzing the Polar Motion (PM) and Length Of Day (LOD) time series to determine its stochastic model structure, estimated using least squares variance component estimation (LS-VCE). With this model, deterministic periodic patterns were extracted through least squares harmonic estimation (LS-HE) and validated against colored noise components to identify significant signals. Using the IERS 14 C04 data from January 1, 2000, to December 31, 2019, the study identified the noise as power-law with a spectral index of −1.5, suggesting non-stationary fractional Brownian characteristics. LS-HE detected dominant frequencies in PM–notably the Chandler and annual signals–and in LOD, with annual, semiannual, 14-day, and 9-day signals. Building on these findings, short-, mid-, and long-term prediction model were developed for the PM and LOD time series from September 2021 to December 2022. The predictive model combines the LS-HE-extracted signals and the noise model to generate forecasts. These predictions were compared with other models from the Second Earth Orientation Parameter Prediction Comparison Campaign, demonstrating competitive accuracy, particularly for the initial forecast days. The results validate that combining LS-HE with a realistic noise model provides an effective approach for short-term of the PM and LOD forecasting, meeting the accuracy goals of geodetic and geophysical applications.