Noise Analysis Of GNSS Coordinate Time Series Considering Common Mode Error

Global positioning technology is entering a period of rapid development in application.Thousands of reference stations distributed around the world provide reliable data for plate displacement research,seismic research,slip fault research and so on.The coordinate time series is defined as the coordinate of the station arranged in time order,which not only includes the displacement caused by tectonic movement,but also includes nonlinear changes.In order to obtain the accurate linear velocity of the station,the nonlinear changes in the time series should be accurately separated.The current mainstream research direction is to study the induction mechanism of nonlinear changes in time series and conduct corresponding modeling,so as to accurately separate the linear and nonlinear movements in time series and obtain the accurate position and velocity changes of the station,which is not only conducive to the reasonable interpretation of plate movements,but also conducive to the geodynamic research such as post-ice rebound.It has important theoretical significance and application value.In this paper,GAMIT/GLOBK10.71 was used to calculate the high-precision original observation data of 7 CORS stations in Hong Kong to obtain the original time series.As the research object,the superposition filtering method was used to extract common-mode errors.Considering that the traditional superposition filtering method is only suitable for the observation network with approximately uniform distribution of common-mode errors,Moreover,the extraction effect is greatly affected by the number and spatial distribution of measuring stations,so the correlation between measuring stations is fully considered.Based on the superposition filtering,the correlation coefficient,the spherical distance,the relative distance between measuring station and the center of the observation network and other factors are taken into account in this study,which is used as the weight to propose an improved superposition filtering method considering various weight factors.The applicability of this method in Hong Kong is analyzed to verify the effect of common mode error extraction.In addition,we analyze the time series before and after the removal of common-mode errors,and explore the influence of common-mode errors on the noise and velocity estimation of time series.This paper’s main work and conclusions are the following:(1)GAMIT/GLOBK(V10.71),was used to calculate the original observation data of 7CORS stations in Hong Kong,China,and the obtained high-precision station coordinate time series was the main data of this study.(2)The original time series was detected by the quarter-distance method,and the missing data was interpolated by the Kalman filter software-GMIS based on MATLAB.The trend item and period item were extracted from the time series data by the least square fitting,and the obtained time series was used as the basic data for subsequent extraction of common mode errors.(3)Time series analysis software Hector was used to carry out noise analysis of time series data of stations in Hong Kong with different noise models.The results showed that for most stations in the selected research area,white noise + flicker noise was a better noise model.For a few stations in the research area,for example,the combination of white noise + power law noise,white noise + random walk noise is a better noise model.(4)In order to extract common-mode errors more accurately,a new filtering method is proposed.Based on the traditional superposition filtering method,this method fully considers the spatial response of each station,including single-day solution accuracy,correlation coefficient between stations,spherical distance between stations,and relative distance between stations and the center of the observation network,and proposes an improved superposition filtering method.The results show that the improved method can achieve better common-mode error extraction effect.(5)By studying the variation of common-mode errors in different spatial scales,the spatial characteristics of common-mode errors are explored.It is found that the common-mode errors of all stations are approximately evenly distributed in small scale regions;With the increase of the scale of the observation network and the distance between stations,the common-mode errors no longer show uniform distribution,and their spatial distribution is related to the magnitude of the correlation between stations.In addition,the change of noise component and velocity uncertainty of the time series before and after the extraction of the common-mode errors are also studied to analyze the influence of the common-mode errors on the time series.