Research On Coordinate Time Series Analysis Of Regional CORS Reference Station

The analysis of the coordinate time series of regional CORS(Continuously Operating Reference System)reference stations has great research significance.It can not only obtain the precise position and speed of the reference station to establish and maintain a dynamic earth reference frame,but also help to reasonably explain the global plate tectonic movement and post-ice Geophysical phenomena at different spatiotemporal scales such as rebound and sea level changes,volcanoes and seismic deformation.This paper takes the coordinate time series of regional CORS reference station as the research object,and researches on the acquisition,preprocessing,common mode error elimination and seasonal signal extraction of time series.The main research contents and innovations are as follows:1.The theoretical model equations of non-difference precision single-point positioning and double-difference relative positioning are deduced.The configuration,processing flow,accuracy evaluation and other contents of the high-precision GNSS(Global Navigation Satellite System)data processing software GAMIT based on the double-difference network solution are introduced in detail.The time series acquisition process based on GAIMT software is studied with Zhengzhou CORS as an example.Based on the CORS data processing practice,a comprehensive CORS network data quality inspection and performance test method is established.This method comprehensively uses software such as GAMIT and TEQC to introduce data integrity,cycle slip ratio,multipath,static and dynamic positioning accuracy,system reliability,time availability,space availability and other performance indicators,respectively,the performance test of the server and user side.2.The gross error elimination aspect of GNSS coordinate time series.To solve the problem that the least squares-based gross error detection method is not robust and the residual time series used for gross error detection is not "real",a new method for gross error detection based on EMD(Empirical Mode Decomposition)is proposed.EMD is first used to adaptively decompose with the original time series to obtain several IMF(Intrinsic Mode Function)components and trend terms,and then the boundary between the signal and noise is judged based on the correlation coefficient,and then the demarcated IMF component is reconstructed into periodic terms of the time series The residual sequence of robustness is obtained as the periodic term and the trend term are deducted.Based on the residual sequences obtained by LS and EMD,the simulated and measured data were used to compare and analyze the gross error detection effects of LS-3 ?,LS-IQR,EMD-3 ?,and EMD-IQR.The detection rate of the difference detection method is higher,and it has obvious advantages for the gross error detection with a smaller level.3.The missing point interpolation aspect of GNSS coordinate time series.A data-driven Reg EM(Regularized Expectation Maximization)algorithm is introduced into the data interpolation of GNSS coordinate time series.The interpolation effect and performance of Reg EM and Lagrange method,cubic spline method,and orthogonal polynomial method are compared through continuous loss of different proportion.The simulated data and measured data contain missing data.The experimental results show that the Reg EM algorithm interpolation effect is better than the traditional method for interpolating data with different proportions continuously missing,and the effect is optimal when a large amount of data is continuously missing.For the measured data with missing data,the effect of maximizing the variance of the sequence obtained by Reg EM interpolation is the best.4.Common mode error elimination aspects of GNSS coordinate time series.The filtering effect of the regional stack filtering algorithm is affected by the number of regional network stations and the spatial scale.The correlation coefficient stack filtering algorithm also has the problem of selecting the spatial scale threshold.In order to weaken the impact of spatial scale on regional superposition filtering,the inverse distance factor and the correlation coefficient are combined,and the Spearman rank correlation coefficient is used to replace the original Pearson coefficient,and a regional stack filtering algorithm with different combinations is designed.The experiment is carried out by selecting the measured time series data.The results show that the regional stack filtering method combining the inverse distance factor and the correlation coefficient can better eliminate the common mode error.The filtering scheme based on Spearman rank correlation coefficient is equivalent to the original Pearson coefficient filtering algorithm.5.The GNSS reference station coordinate time series has obvious seasonal changes.The least squares method based on the harmonic function model can only solve the seasonal signal with fixed amplitude,but the amplitude of the real seasonal signal changes.When semi-parametric models are used for seasonal signal extraction,the optimal smoothing factor is difficult to determine and the iteration speed is slow.A semi-parametric model with relative weight ratio is proposed.The optimal smoothing factor is determined by a combination of a faster iterative golden section method and an improved efficiency method.The simulation data experiments verify the usability of the improved model.Experiments show that the calculation efficiency of the improved method is significantly improved.The results show that the improved method can effectively determine the optimal smoothing factor,and the calculation efficiency is significantly improved.The calculation accuracy is improved compared with the least square method and the semi-parametric method.There is no obvious seasonal signal in the residuals of the obtained model.The effects of the three methods on the extraction of seasonal signals from the measured data are compared and analyzed by the time series data of IGS stations provided by SOPAC(Scripps Orbit and Permanent Array Center).The results show that the seasonal signals extracted by the improved method are more in line with the actual movement trend of the time series.