Prediction of the Length of Day From Atmospheric Angular Momentum With Lstar Model

Abstract

Adopting the time series of atmospheric angular momentum (AAM) from the National Meteorological Center of USA, the study of the prediction of the length of day (LOD) has been made by the Leap-Step Threshold AutoRegressive (LSTAR) model. The LSTAR model presented by the author is a sort of models for nonlinear time series analysis such as $$ {Z_n} = TS{M^{(j)}} + {E_n}^{(j)} {Z_n}\varepsilon Dj j = 1,2,L,L$$ where Dj is the j-th leap-step domain of the data series Zn, and if the sample number N=LxM, then Zj+(LxK)ɛDK=0,1,…,M-1. En (j) denotes the white noise of data in the j-th leap-step domain. TSM denotes a class of models in time series analysis and the nonlinear threshold autoregressive model is used here.