Accuracy enhancement of GPS time series using principal component analysis and block spatial filtering

Abstract This paper focuses on performance analysis and accuracy enhancement of long-term position time series of a regional network of GPS stations with two near sub-blocks, one block of 8 stations in Cascadia region and another block of 14 stations in Southern California. We have analyzed the seasonal variations of the 22 IGS site positions between 2004 and 2011. The Green’s function is used to calculate the station-site displacements induced by the environmental loading due to atmospheric pressure, soil moisture, snow depth and nontidal ocean. The analysis has revealed that these loading factors can result in position shift of centimeter level, the displacement time series exhibit a periodic pattern, which can explain about 12.70–21.78% of the seasonal amplitude on vertical GPS time series, and the loading effect is significantly different among the two nearby geographical regions. After the loading effect is corrected, the principal component analysis (PCA)-based block spatial filtering is proposed to filter out the remaining common mode error (CME) of the GPS time series. The results show that the PCA-based block spatial filtering can extract the CME more accurately and effectively than the conventional overall filtering method, reducing more of the uncertainty. With the loading correction and block spatial filtering, about 68.34–73.20% of the vertical GPS seasonal power can be separated and removed, improving the reliability of the GPS time series and hence enabling better deformation analysis and higher precision geodetic applications.