Abstract:The noise of GPS site time series is difficult to model and separate from signal. In this paper, the noise of GPS time series for 9 IGS stations is reduced by empirical mode decomposition (EMD), an adaptive signal analysis method. First, we decompose the GPS time series by EMD and get a series of intrinsic mode components and its trends. Then we use the correlation coefficient to distinguish the intrinsic mode function (IMF) between noise with signal and restructure the intrinsic mode function. Finally, we use the correlation coefficient, signal-to-noise ratio, and percentage of energy, to evaluate the effectiveness of the EMD method for noise reduction. The results show that the EMD method could separate signal and noise in sequence reasonably and it effectively weakens the impact of noise in GPS time series. Furthermore, it could further improve the accuracy of the GPS site time series.