Abstract：To solve the problem of common mode error(CME) in regional GNSS coordinate time series, the PCA and ICEEMDAN methods are proposed to eliminate the spatial domain error, taking the GNSS coordinate time series in north China as an example.Firstly, we use the PCA filter to extract the CME. Then we use the ICEEMDAN method to decompose the CME to extract the useful IMF components, which reduces the loss of original information. Finally, we compare and analyze the regional filtering results of PCA, ICA and IC-PCA. The results show that the filtering effect of IC-PCA is better than ICA and PCA. The residual RMS in N, E and U directions before and after IC-PCA filtering decreases by 1.93%,1.92% and 7.60% on average, and the filtering results are more consistent with the change trend of the original time series, which effectively reduces the loss caused by PCA on the original time series.