Comprehensive Analysis of the Effects of Common Mode Error on the Position Time Series of a Regional GPS Network

Abstract

This study uses the daily horizontal and vertical Global Positioning System (GPS) position time series of 41 regionally distributed GPS stations in Southern California covering the period of January 2011 to January 2017 to extract the common mode error (CME) in east-north-up (ENU) components with the principal component analysis (PCA) method. We assume the first component with common spatial responses is CME. Firstly, we focus on the effects of CME on the root mean square (RMS) of the GPS time series. The results indicate that after being filtered, CMEs can reduce the average RMS values by 42.9, 39.1 and 28.8% for ENU components, respectively. Secondly, we investigate the changes in noise type before and after spatio-temporal filtration. The outcome reveals that noise model of the GPS position time series can be mainly described by power law noise (PL) plus white noise (WH). After CME correction, 32, 12 and 13% of the stations’ noise types are altered for ENU components, respectively. Thirdly, we confirm that stations whose noise properties can be best described by PL + WH, PL + WH and FL + WH or RW (Random walk noise) + FL + WH show no obvious differences in velocity and velocity uncertainty. Stations whose noise properties are described by RW + FL + WH, PL + WH and RW + FL + WH show significant differences in velocity and velocity uncertainty. Therefore, for all GPS stations, excluding the ones whose noise properties are described by the RW + FL + WH model, we estimate the noise parameters and velocity before and after spatio-temporal filtration by fitting a PL + WH model. According to our findings, the PCA method does not appear to be efficient for horizontal components. The CME is superimposed on the stochastic process realization for the vertical components. The impact of CME on vertical velocity estimation cannot be ignored, and eliminating CME enhances the credibility of the determined velocity field. Finally, we discuss the correlation between CME and mass loadings. The results of the Pearson correlation analysis and wavelet coherence analysis indicate that CME cannot be interpreted by mass loadings in a regional GPS network and that it may be related to GPS technique errors.