Assessment of length-of-day and universal time predictions based on the results of the Second Earth Orientation Parameters Prediction Comparison Campaign

Predicting Earth Orientation Parameters (EOP) is crucial for precise positioning and navigation both on the Earth's surface and in space. In recent years, many approaches have been developed to forecast EOP, incorporating observed EOP as well as information on the effective angular momentum (EAM) derived from numerical models of atmosphere, oceans, and land-surface dynamics. The primary objective of these efforts has been to improve the capabilities of EOP forecasting, especially for ultra-short-term prediction horizons (i.e., up to 10 days into the future), which are vital for real-time applications. The Second Earth Orientation Parameters Prediction Comparison Campaign (2nd EOP PCC) aimed to comprehensively evaluate EOP forecasts from many international participants, quantify the current forecasting capabilities, and identify the most promising prediction methodologies. This paper presents the validation results of predictions for universal time (UT1-UTC) and length-of-day (LOD) variations submitted during the 2nd EOP PCC, providing an assessment of their accuracy and reliability. After providing a statistical summary of the submitted files, we conduct a detailed evaluation of all valid forecasts using the 14 C04 solution provided by the International Earth Rotation and Reference Systems Service (IERS) as a reference and mean absolute error (MAE) as the quality measure. Our analysis demonstrates that approaches based on machine learning (ML) or the combination of least squares and autoregression (LS+AR), with the use of EAM information as an additional input, provide the highest prediction accuracy for both investigated parameters. Utilizing precise EAM data and forecasts emerges as a pivotal factor in enhancing forecasting accuracy. We also examine the impact of the considered period on the forecast accuracy and show that the majority of predictions exhibit time-varying MAE values. Although several methods show some potential to outperform the IERS forecasts, the current standard predictions processed at U.S. Naval Observatory are highly reliable and can be fully recommended for operational purposes.