Extensive analysis of IGS REPRO1 coordinate time series

The work describes the analysis conducted on the IGS REPRO1 coordinate time series, in order to detect GNSS permanent stations periodic behavior. Frequency analysis requires cyclostationary time series, while observed coordinates time series are not cyclostationary because of discontinuities of different kind and origin and of long term linear or non linear trend. For this reason time series offsets and trends must be estimated and eliminated, prior to conduct the harmonic analysis. Discontinuities are usually documented by IGS, but undocumented discontinuities also exists and need to be detected. The long term component of the signal is generally modeled as a linear trend, but the linear model is often inadequate to obtain cyclostationary residuals. An alternative model based on a discrete time Markov process will be adopted. The study has been conducted on the up component of the REPRO1 raw coordinates time series. No correction for the atmospheric pressure loading has been applied. Harmonic analysis has been performed using the non linear least square algorithm implemented by F. Mignard in the Frequency Analysis Mapping On Unusual Sampling software (Mignard, FAMOUS, Frequency Analysis Mapping on Unusual Sampling, (OCA Cassiopee), 2003). We obtained a complete statistic on the vertical component period, amplitude and phase. Signals at from 1 to 7 cycle per solar and draconitic year can be observed in most stations as expected, but also other signals have been detected that can be attributed to tidal model errors. Some interpretation will be given referring to recent literature.