Predictability of the Earth’s polar motion

The numerical prediction of the Earth’s polar motion is of both theoretical and practical interest. The present paper is aimed at a comprehensive, experimental study of the predictability of polar motion using a homogeneous BIH (Bureau International de l’Heure) data set for the period 1967–1983. Based on our knowledge of the physics of the annual and the Chandler wobbles, we build the numerical model for the polar motion by allowing the wobble period to vary. Using an optimum base length of six years for prediction, this “floating-period” model, equipped with a nonlinear least-squares estimator, is found to yield polar motion predictions accurate to within 0″.012 to 0″.024 depending on the prediction length up to one year, corresponding to a predictability of 89–82%. This represents a considerable improvement over the conventional fixed-period predictor, which, by its nature, does not respond to variations in the apparent wobble periods (in particular, a dramatic decrease in the periods of both the annual and the Chandler wobbles after the year 1980). The superiority of the floating-period predictor to other predictors based on critically different numerical models is also demonstrated.