Noise Analysis Of GNSS Timing Data In Northwest Terrestrial Network Considering Common Mode Error

High precision GNSS coordinate time series can obtain a large amount of geoscience information,but in addition to structural signals,there are also non structural signals in the GNSS coordinate time series.These factors will affect the accuracy and reliability of the GNSS coordinate time series.The estimation of characteristic parameters such as data preprocessing,nonlinear signal extraction,and noise analysis for GNSS coordinate time series will help to establish high-precision coordinate time series models,providing a solid data foundation for the construction and maintenance of reference frameworks.This paper takes the coordinate time series of continuous stations in the land state network in northwest China as the research object,and mainly studies the preprocessing method of GNSS coordinate time series,the signal noise separation method of GNSS coordinate time series,and the influence of common mode error on the characteristic parameters of coordinate time series.This provides reliable support data for subsequent research on geophysical phenomena,the variation law of nonlinear crustal motion,and the monitoring of plate motion speed.The research work done in this article mainly includes the following aspects:(1)The research progress of time series data preprocessing at home and abroad,the research progress of GNSS coordinate time series at home and abroad,and the existing problems in signal noise research are studied.Based on this,the research content and technical route of this paper are proposed.(2)Aiming at the problem of discontinuous and uneven coordinate time series caused by gross errors and missing values in GNSS coordinate time series,the quartile distance method and dynamic interpolation method based on Kriging Kalman filter model were used to process them respectively.The results show that the above methods can effectively eliminate gross errors and obtain coordinate time series with more uniform and complete intervals;The commonly used principal component analysis(PCA)and regional superposition filtering are used to extract common mode errors,respectively.The results show that PCA has a good filtering effect,and the RMS mean values of N,E,and U components after filtering have significantly decreased.However,regional superposition filtering cannot extract common mode errors well on larger spatial scales,and the filtering effect is poor.It provides a reliable data base for analyzing coordinate time series.(3)Aiming at the problem that useful signals and noises in GNSS coordinate time series cannot be accurately separated,this paper proposes a combined noise reduction method based on combined weighted wavelet and set empirical mode decomposition.Using root mean square error,signal-to-noise ratio,flicker noise amplitude,and velocity uncertainty as comprehensive evaluation indicators,the feasibility of this method is verified.The results show that the denoising effect of this method is lower than that of wavelet denoising and empirical mode decomposition,with lower root mean square error and higher signal-to-noise ratio.The uncertainty of flicker noise amplitude and velocity is also significantly reduced,but some stations have less denoising effect than wavelet denoising.(4)Aiming at the problem of colored noise modeling in GNSS coordinate time series,the optimal noise model was determined based on Bayesian information criteria.A comparative analysis of the changes in the noise models of all station coordinate time series in the study area before and after PCA filtering was conducted.The results showed that the optimal noise model before filtering was WN+FN,and some station noise models changed after filtering,but the combined noise model of WN+FN still dominates.On the basis of considering the optimal noise model,a continuous station velocity field model for the northwest land state network under the framework of ITRF2014 was established,and the changes in velocity estimation and uncertainty before and after PCA filtering were compared and analyzed.The results show that the accuracy of the filtered horizontal and vertical velocity field models is higher than that before filtering.It is necessary to eliminate common mode errors when analyzing coordinate time series.