| Summary: | Length-of-day (LOD) is used to model variations in Earth's rotation. In recent years, prediction techniques have been improved as predictions of highly variable LOD are slightly less accurate than observations even for a few days in the future. LOD is linked with changes in the climate. The variability in the internal dynamics of the ocean and atmosphere is an essential factor that determines the Earth’s climate and is represented by the climate indices. But their relationship with LOD is complex and not yet fully understood. In general, LOD predictions are facilitated by effective angular momentum (EAM) functions based on ECMWF data. In this work, we use an evolutionary machine learning approach to predict LOD in the short term (up to 10 days), medium term (up to 30 days), and long term (up to 365 days). First, we use LOD from IERS with EAM provided by GFZ as predictors. The latter results are compared with that of the second approach, where we use both EAM and climate indices as predictors in an effort to boost our prediction. In order to investigate the relationship between LOD and climate indices, we employ multivariate empirical mode decomposition to identify the various frequency components of the auxiliary data. The climate indices examined are North-Atlantic Oscillation, El-Nino Southern Oscillation, Pacific Decadal Oscillation, Indian Ocean Dipole, and Southern Annular Mode. This helps to clarify the relevance of climate indices on LOD prediction.
|
|---|