A method based on MTLS and ILSP for GNSS coordinate time series analysis with missing data

For GNSS coordinate time series analysis with missing data, we propose an improved comprehensive scheme without any pre-interpolation of missing data, and we perform a simulation experiment to evaluate the validity of this method. The coordinate series of more than 260 reference stations in the CMONOC (Crustal Movement Observation Network of China) from 2011 to 2018 are employed to verify this method further. First, the moving median method is applied to detect outliers and determine the occurrence times and amplitudes of all offsets. The results of the offsets indicate that the discontinuities caused by large earthquakes can be located and adjusted accurately. Second, a method based on MTLS (Multivariate Total Least Squares) and ILSP (Improved Lomb–Scargle Periodogram) is applied in a weighted iterative procedure. Using the ILSP-based algorithm, not only can the influence of both observation noise and the mutual modulation between different frequency signals be eliminated, but the periods of all actual periodic signals can also be accurately determined even if there are false spectral peaks. Analysis of the velocity results shows that the crustal movement in the eastern region of mainland China is weaker than that in the west, where the most intense crustal activities usually happened (such as Tibet, Sichuan, and Yunnan). For the periodic signals in the coordinate series of each station, the annual (365.25 d) and semi-annual (182.26 d) signals are dominant, followed by other quad-annual (91.3 d) signals, and the amplitude of each periodic signal is associated with specific geographical or hydrological causes. The reliability of the MTLS + ILSP method is further verified by a comparative analysis of three sets of velocity field products, where the RMS (Root Mean Square) of the residual velocity is reduced by more than 70% on average, and high precision velocity field is conducive to realize and maintain regional reference frame.