Prediction of polar motion and UT1-UTC based on the hybrid EEMD_LSTM model

Accurate and real-time Earth rotation parameters (ERPs) are crucial for various scientific fields, including astronomy, geoscience, and oceanography. Although there are various space geodetic techniques, such as VLBI, GNSS, DORIS, and SLR, which can determine the ERPs with high accuracy, these metho...

Full description

Saved in:
Bibliographic Details
Main Authors: Chenxiang Wang, Pengfei Zhang, Wenbin Shen, Jiayao Wang
Format: Article
Language:English
Published: Taylor & Francis Group 2025-06-01
Series:Geo-spatial Information Science
Subjects:
Online Access:https://www.tandfonline.com/doi/10.1080/10095020.2025.2522154
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:Accurate and real-time Earth rotation parameters (ERPs) are crucial for various scientific fields, including astronomy, geoscience, and oceanography. Although there are various space geodetic techniques, such as VLBI, GNSS, DORIS, and SLR, which can determine the ERPs with high accuracy, these methods have a certain delay and are difficult to meet the needs of real-time applications. Therefore, the ERPs prediction with high accuracy is very significant. Conventional approaches to predicting ERPs are mainly based on the linear models. However, it is not always optimal for the non-linear effects of ERPs. In this paper, we propose a model that combines Ensemble Empirical Mode Decomposition (EEMD) with Long Short-Term Memory (LSTM), denoted as EEMD_LSTM. We make the prediction experiments by using the EEMD_LSTM model, all experiments use the IERS EOP 20C04 as the basic series, the sliding window set as one week, and the prediction accuracies of polar motion (PM) and UT1-UTC from 2011 to 2021 are analyzed. Then, comparing the prediction values of EEMD_LSTM with the least square and autoregressive (LS_AR) model, and International Earth Rotation and Reference Systems Service (IERS) Bulletin A. Results indicate that the EEMD_LSTM model can improve the prediction accuracy of PMX and PMY up to 35.79% and 32.30%, respectively. Additionally, the accuracy of UT1-UTC predicted by using the EEMD_LSTM model demonstrates an improvement of 41.93% in the mid- and long-term (120–365 days).
ISSN:1009-5020
1993-5153