Exploring various approaches to combine Earth Orientation Parameter (EOP) predictions gathered during the Second EOP Prediction Comparison Campaign (2nd EOP PCC)

The Second Earth Orientation Parameters Prediction Comparison Campaign (2nd EOP PCC) aimed to evaluate and compare various methods of Earth Orientation Parameters (EOP) predictions. One of the goals of the 2nd EOP PCC was to prepare a combination of the predictions to obtain one robust and accurate solution for forecasts of individual parameters. This presentation focuses on identifying the most reliable and accurate combination of predictions for polar motion (PMx, PMy), universal time variations (UT1-UTC), and length of day (LOD) among the methods tested during the 2nd EOP PCC.

Two types of experiments were designed for this study: "operational" combinations tailored to real-time comparisons and practical application and "final" combinations designed for comprehensive analysis. Boths approaches incorporated six methods for handling outlier predictions, ranging from no filtration to progressively stricter criteria using the σ+β method (with α values ranging from 5 to 1). All experiments cover the period of 2nd EOP PCC (from September 1, 2021, to December 31, 2022), and each approach includes 70 10-day predictions.

The results show that combining various submissions generally enhances stability and accuracy of EOP forecasts. The σ+β criterion with α = 1 achieved the smallest Mean Absolute Prediction Error, indicating high accuracy of prediction. However, this method of eliminating outliers forecasts is the most restrictive, as it excludes a significant number of predictions. In contrast, operational combinations without filtering proved more practical for real-time applications, albeit with slightly higher errors.

The findings underscore the importance of tailoring combination strategies to specific goals—whether prioritizing maximum accuracy or practical applicability. This research highlights the benefits of prediction combination methods in improving EOP forecasts, offering a foundation for further development of operational strategies and expanding their use in geophysical and astronomical applications.