Investigating Mass Loading Contributors of Seasonal Oscillations in GPS Observations Using Wavelet Analysis

We adopt the cross wavelet transform (XWT) to examine the potential geophysical contributors of seasonal oscillations in GPS observations. Daily vertical GPS position time series and mass loadings [atmospheric, oceanic, and hydrological loading (AOH)] of 30 globally distributed GPS sites, spanning from January 2002 to December 2014, are used to quantify the performance. First, we examine the spectra of GPS time series and AOH. The results confirm the anomalous sub-seasonal peaks in GPS spectra are seen to have not an obvious geophysical explanation. The Akaike information criteria is then used to quantify how well the noise models fit the two series. The Generalized Gauss Markov plus white noise (GGM + WH) model is in most cases the preferred noise model for GPS, and the fifth order autoregressive plus white noise (AR(5) + WH) model is the preferred noise model for AOH. Second, we test the significance of periodic oscillations in GPS residuals and AOH. We find both series have significantly high power located near one cycle per year frequency band, whereas harmonic signals at higher draconitic frequency are identified as non-white process. Finally, we adopt XWT to examine the relative phasing between the two series, and find the annual variations in two series are physically related for most sites. The time variable phase asynchrony obtained using the XWT-based semblance analysis confirms that the annual variations in GPS observations are resulting from a combination of geophysical signals and systematic errors. The weighted least squares fitting method where the covariance matrix follows a specific stochastic noise model is also performed for comparison.