ABSTRACT. Though pure mathematical approximations such as regression models and neural networks show good results in Earth rotation forecasting, dynamical modeling remains the only base for the physically meaningful prediction. That assumes the knowledge of cause-eect relationships and physical model of the rotating Earth. Excitation reconstruction from the observed Earth orientation parameters (EOP) is a crucial stage, needed for comparison with known causes, such as tidal forcing, atmospheric (AAM), oceanic (OAM) angular momentum changes, and uncovering unknown ones. We demonstrate dierent approaches, which can be used to avoid ill-conditionality and amplication of noises during the inversion. We present amplitude and phase studies of the model and reconstructed excitations of Chandler wobble. We found out, that modulation of Chandler excitation is synchronous with 18-yr tidal eects in the Earth's rotation rate changes. The results of the study can be used for excitation and EOP forecast. The key issues of the EOP prediction are discussed.Keywords: Earth's rotation, Chandler wobble, excitation functions, inverse problems.
INTRODUCTIONThe results of the EOP PCC have shown, that various mathematical methods of EOP prediction compete with each other at approximately equal level, but the best results are given by the methods based on the physical modeling of the Earth's rotation. They use dynamical model and involve a number of additional data about the processes that cause variations in the rotation velocity and orientation of the poles of the planet. Mathematical approximations are based on the selection of dependencies, which approximates observations well, perhaps with a large number of parameters. For long-term forecast such models can be unsuitable. Physical models are based on the identied cause-eect relationships, they also have mathematical representation, possibly with fewer parameters. They should explain observations and have a predictive power. The following key issues related to the EOP prediction by means of physical modeling can be raised