Abstract：We apply PCA method on Taiwan GPS coordinate time series to extract the CME and perform spatial filtering. The results show that the first principal component exhibits strong consistency in space while the second to fifth principal components exhibit local response characteristics. The average reduction rates of RMS in the N, E and U directions of the residual time series after filtering are 33%, 31% and 19%, respectively. The result of filtering in a horizontal direction is better than in a vertical direction, and there is a difference in the filtering effect between the eastern and western parts of Taiwan, which may be related to the terrain. Meanwhile, the amplitude of power law noise decreases by 36%, 43% and 23% in N, E and U direction, respectively.
