A GNSS Time Series Denoising Method with Mixed Use of Cross Validation and CEEMD-WT

Objectives The coordinate time series derived from global navigation satellite system (GNSS) usually contains noise, which may lead to misinterpretation of some geophysical phenomena. This paper aims to reduce noise in GNSS datasets by combining cross validation (CV), complementary ensemble empirical mode decomposition (CEEMD) and wavelet transform (WT).

Methods First, the original GNSS time series is decomposed into several intrinsic mode function (IMF) components by CEEMD. The CV strategy is then used to classify these IMFs into signal-dominant components and noise-dominant components. After removing the noise-dominant IMFs, WT-based denoising method is applied to the remaining signal-dominant IMFs to obtain final GNSS time series.

Results Both simulated signals and real dataset collected from 117 GNSS stations are used to evaluate the performance of the proposed method. The experiment results show that the proposed method can effectively attenuate noise in the original time series. Compared with the residuals of the original time series, the residuals of denoising results derived by the proposed method are reduced by 43%, 43%, and 46% in north, east, and up directions, respectively.

Conclusions The proposed method shows superior noise reduction in GNSS time series compared to standalone WT, CV-based CEEMD, and correlation-coefficient-based CEEMD methods.