Segmented Polar Motion Prediction Based on Varying Effective Angular Momentum Forecast Horizons

Polar motion (PM), a key component of Earth orientation parameters (EOPs), is essential for high-precision satellite orbit determination and deep-space navigation. However, delays in data acquisition and processing limit its availability for real-time applications, necessitating the development of prediction models based on historical observations. Common approaches include least squares extrapolation (LS), autoregressive (AR) models, and their combination (LS + AR), often enhanced by effective angular momentum (EAM) from Earth’s fluid components. This study examines an EAM + LS + AR method for PM prediction, systematically evaluating how different EAM forecast horizons (1–10 days) affect 90-day prediction accuracy for both PM X and Y components. A segmented optimization strategy is proposed and validated against International Earth Rotation and Reference Systems Service (IERS) official predictions using the IERS EOP 14 C04 product. Key findings include: (a) Adjusting the EAM horizon substantially reduces prediction errors. Segmented prediction improves PM X accuracy by 20–30% (1–60 days) and 10–20% (61–90 days) relative to IERS rapid products, while PM Y short-term accuracy improves by 20–40% (1–15 days). (b) The influence of EAM horizon on long-term PM Y prediction gradually weakens, with errors converging to approximately 8 mas by day 90. (c) For 1–10-day forecasts, optimal horizons follow a systematic pattern: day m predictions achieve the highest accuracy using an (m−1)-day EAM horizon, while a 10-day horizon is optimal for long-term forecasts. (d) The proposed method shows clear advantages over IERS forecasts, with 83.5% of PM X predictions (1–90 days) and 50.78% of PM Y predictions (1–15 days), outperforming IERS daily products during the 2024 test period.