I am reaching out to inquire about the handling of Earth Orientation Parameters (EOP) in Orekit, particularly for times that extend beyond the predictions in the EOP data file, such as the year 2026. As a new user of Orekit, I have been examining the code and noticed that the software adheres to Technical Note 36, specifically chapter 5.5. This chapter details time-function tables and equations for correction analysis, which Orekit seems to incorporate.
Could you please confirm if my understanding is correct? Are these equations utilized once the EOP data file’s predictions are surpassed by the mission’s timeline? Additionally, I am curious about the accuracy of these predictions compared to an approach where all corrections are set to zero for any propagation outside the EOP file’s span.
Your clarification on this matter would greatly assist me in ensuring the precision of my mission’s trajectory predictions.
There is however a dedicated feature for predicting EOP with class PredictedEOPHistory. Beware however that this feature is far inferior to what IERS provides! It was developed for short terms prediction, and errors grow up significantly as time passes. If you look at the testAccuracyShortTerm and testAccuracyLongTerm in the tests, you will see that when using fitting parameters optimized for short term prediction, the error on Earth surface (angular error multiplied by Earth equatorial radius) is roughly at meter level for the first week but will grow up to 40 kilometers for 3 months prediction and that when using fitting parameters optimized for long term prediction, the error on Earth surface is above meter level for the first week but remains contained to less than 30m for 3 months prediction. Of course, your mileage will vary, you can run the tests by yourself with different fitting parameters and different EOP input. I guess with the current weird behavior of Earth rotation (it has been accelerating for the last few months), we could expect almost anything, so I would not trust the prediction, its use case is clearly not for long term prediction.