Abstract:
The common mode error (CME) is one of the major error sources in the regional Global Navigation Satellite System (GNSS) network. Aiming at the problem that GNSS time series is subject to no-Gaussian distribution, and the principal component analysis (PCA) with second-order is inaccurately employed to separate the CME. In this paper, the independent component analysis (ICA) introduces high-order statistics to extract the CME. The effectiveness of the method is validated by processing the data of GNSS stations from 2011 to 2018 in Xinjiang, China, and then compared and verified the filtering effect of PCA. We analysis the influence of the CME for GNSS coordinate time series and the yearly signal of the CME. The results show that the CME mainly consists of the 6th independent components and can be attributed to satellite orbit, surface mass loading, and clock errors. After the ICA filtering, the reduction of mean RMS is 31.83%, 32.29%, 35.49% for the north (N), east (E), and up (U) components, respectively. The reduction of velocity uncertainty can achieve 44.14%, 38.49% and 35.49% in three components. In addition, the yearly amplitude of each GNSS station is more consistent that before spatiotemporal filtering, indicating that the ICA can effectively extract the CME and further improve the accuracy of coordinate time series.