Fuzzy-wavelet based prediction of Earth rotation parameters

Abstract

Prediction of Earth rotation parameters (ERPs) is of importance especially for near real-time applications including navigation, remote sensing, and hazard monitoring. Therefore, prediction of ERPs at least over a few days in the future is necessary. Fuzzy-inference systems (FIS) are increasingly popular and have advantage over classical FFT that lacks stochastic stability due to non-stationarity, multiscaling, and persistent autocorrelations. Wavelet filtering can be used to handle such phenomenon. A FIS rule-base created from ERP time series, where the volatilities (returns) of the preprocessed series are used, and high frequency signals removed, is summarized. The performance of this system, trained using the fuzzy-wavelet method, is compared with that of a conventional FIS, trained on raw time series. The results show that the predictions by the fuzzy-wavelet method are superior to the FIS-only model for short-term predictions (up to 10 days in future). The improvement of prediction accuracy is found to be about 30% in terms of RMS error.