INVESTIGATION OF TIME-CHANGEABLE SEASONAL COMPONENTS IN GPS HEIGHT TIME SERIES: A CASE STUDY FOR CENTRAL EUROPE

摘要

Nowadays, beyond the dispute we should take into account the time-varying parameters of seasonals in the GPS-derived position time series. Either real geophysical effects or system specified artefacts can introduce non-sinusoidal changes. For this study, we used 18 daily position time series from Central European stations provided by the Jet Propulsion Laboratory (JPL) processed in the GIPSY-OASIS software in a Precise Point Positioning (PPP) mode. We tested two different approaches to subtract the seasonal signals: Least-Squares Estimation (LSE) and Singular Spectrum Analysis (SSA). The SSA approach is suitable for all (stationary and non stationary) time series, without prior knowledge about the data characteristics, which is an undisputable advantage. We extracted periodicities from GPS position time series, and demonstrated the usefulness of the SSA approach on the example of the vertical component. We showed, that the reassembled signal, containing only the first four Reconstructed Components, has a larger correlation coefficient than LSE-extracted signals, with respect to the original time series. Moreover, using the Alcaike Information Criterion and Fisher-Snedecor test we tested optimum length of the sliding window and significance of the obtained RCs, respectively.