| Abstrakt: |
Abstract: The constantly growing needs of permanent stations’ velocities users cause their stability level to increase. To this research we included more than 150 stations located across Europe operating within the EUREF Permanent Network (EPN) with weekly changes in the ITRF2005 reference frame. The obvious long-range dependencies in the stochastic part of GPS time series were proven by Ljung-Box test. Moreover, the results of Detrended Fluctuation Analysis (DFA) comprised in 90 % in the long-time range dependency, indicated an autocorrelation of the signal with a possible coloured noise, similar to the flicker one. The noise in the GPS time series was analysed here with the Maximum Likelihood Estimation (MLE). The amplitudes of white noise for the white, flicker and random-walk combination vary between 0.5 to 3 mm for horizontal components and between 1.5 up to 9 mm for the vertical component. The amplitudes of white and flicker noise show a clear geographical latitude dependence with a correlation with its minimum centred at approximately 50°N. The assumption of white plus flicker plus randomwalk noise led to stations’ velocity errors from 0 to 3 mm/year for the flicker noise and from 0 up to 7 mm/year for the random-walk. The spectral indices for an approach assuming the occurrence of the white and power-law noise model in weekly time series varied between -1.7 to -0.2 and -1.4 to -0.2 for horizontal and vertical changes. The performed spatial filtering (weighted stacking method) resulted in decreasing the power-law amplitudes by averagely 1-2 mm/year-κ/4 and moving the spectral indices closer to zero, with amplitudes and indices median decreased by 5 %, compared to the time series before stacking. |
