Abstract
An effectiveness of a new prediction method based on a characterization of dynamics of a non-stational time series data is evaluated. Through a comparison of the proposed method with the conventional prediction method based on a linear estimation of the parameter of the stational time series of data of interest, it is confirmed that the proposed method is superior to the conventional method for the non-stational time series of the simulation data derived from the Auto Regressive model of which the parameters of the model are determined from the SSM/I (Special Sensor of Microwave / Imager) of 19.3 GHz vertical polarization of brightness temperature data in all the cases in concern.
