Study On The Noise Model Of GPS Coordinates Time Series

With the rapid development of space observation technology, GPS has become an important observational approaches in Geodesy and Geodynamics. The global distributed IGS reference stations have accumulated nearly twenty years of coordinate time series, which provide valuable basic data for the study of the geodynamics and global tectonics of earth's lithosphere and mantle. Moreover, GPS observable time series not only contain geophysical signals, but also unmodelled errors and other nuisance parameters, making the GPS coordinate time series present a nonlinear variation, which affects the performance in the estimation of site coordinates and velocity. Analyzing the coordinate time series, especially the physical mechanism of nonlinear movement, implementing various error model corrections, and making further systematic study on the origin and influence mechanism of GPS nonlinear variation, could not only help to implement various error model corrections, but also hold important theoretical meaning and practical value in better serving for geodynamics research. GPS has been seen tremendous advances in measurement precision and accuracy in the past two decades, however, there still remain many challenging problems which require further investigations.Following the international focus on GNSS time series analysis and application, we systematically study the research methodologies and basic theories of nonlinear variations in GPS coordinates time series and develop GPS time series analysis software tools. We focus on analyzing the optimal strategy of GPS daily-solution, investigating the influence mechanism of different environmental loading models on GPS coordinate time series in depth, studying the spatial and temporal variations of common mode error (CME) on GPS time series with a large-scale GPS network, and proposing an improved CME filtering method. Finally, we investigate the optimum noise model of IGS tracking stations coordinate time series, resulting in some significant observations.The major contributions of this study are summarized as follows:(1) Comprehensively introduce the basic principle and the methodology of GPS coordinate time series analysis on nonlinear variation, including GPS time series function model, spatial filtering method, establishment of optimal noise model of GPS time series. We summarize the current status of GPS time series analysis and related application, point out the insufficiency of the existing GPS time series analysis, and propose basic steps for analysis on nonlinear variation of GPS coordinates time series. To handle the problems existing in GPS time series analysis, we develop time series analysis tools to improve the efficiency of GPS data processing and analysis and provide convenient to explore the relationship between GPS time series and related geophysical phenomena.(2) Considering the differences between software and corresponding processing strategies, we perform an in-depth investigation on the strategies of GPS daily solutions. To improve the efficiency of data processing, we propose a GPS daily solution batch processing strategy, greatly reduce the processing complexity. Double-difference, un-difference and combined strategies using GAMIT, GIPSY and their combinationare implemented with observation data from 165 IGS stations. The results show that the combined solution is more accurate than the individual schemes and is able to eliminate errors caused by differences between software and processing strategies, producing more accurate GPS solutions such as for coordinates and velocity. Through the performance analysis with different weighting factors between GAMIT and GIPSY quasi-observation solutions, we obtain the optimal weights of the joint solutions, and verify the reasonableness and reliability of the empirical weight given by SOPAC.(3) Based on the significant difference in the scholars'opinions on the contribution of environmental loading on GPS site series, we propose reliability and regional analysis on loading effect. From the characteristic analysis (e.g. outlier analysis, PCA analysis, and noise analysis) on atmospheric pressure, soil moisture, non-tidal ocean and snow cover mass loading induced displacement time series, the results show that the accuracy of the loading induced time series is reliable. It is also proved that the annual and semiannual periods are clearly evident in all the NEU directions of the four different environmental loading induced displacement time series. From the analysis on the spatial variation of the surface loading effect with 206 globally distributed IGS stations, the contribution of environmental loading on GPS time series (coordinates) is quantified, together with the extremal characteristic analysis and the difference of WRMS before and after loading correction. The results indicate that loading effect behaves with regional difference in spatial scale, and this explains why there existed significant difference on the contribution of environmental loading on GPS time series from different scholars. Besides, the regional difference in spatial scale provides a reference to further study the error associated with the station and related geophysical phenomena.(4) Comprehensively investigate the physical origin of CME, showing that CME is of uniform in spatial scale. Principal component analysis (PCA) is used to separate CME on GPS time series with different scale, and the results show that the span of 400 to 500 kilometers is appropriate for effective filtering. As the scale increases and thus the distances between stations increase, the correlation weakens, making it hard to separate CME accurately. We observe that the traditional removal of seasonal terms followed by CME separation shows some limitations. Thus, we propose an improved general PCA CME separation method based on environmental factors. Taking correlation coefficient, distance, latitude and longitude, local effect, contribution rate of principal component, spatial response and other geographical factors into consideration provides a feasible method for estimating CME in large scale GPS networks. Better spatial filtering results can be obtained with the proposed method, which overcomes the limitations of traditional GPS time series model, shows more reliably thevariation of CMEs in space and periodity, and provides a basis for further improving the accuracy of GPS time series model.(5) Investigate GPS time series noise modeling and its impact on using global and regional IGS GPS observations. Results show that there is no significant correlation between the monument type of IGS station and noise model, but the noise model of IGS time series shows diversity and regional variation. The time span plays an important role in establishing an optimal noise model of IGS coordinate time series. The developed noise models are divergent and model uncertainty is large when the time span is small. As the time span increases, the noise model tends to be convergent and the velocity uncertainty becomes steady. It is thus proposed to use time series longer than 10 years to reduce the effect of noise on velocity and its uncertainty estimation. The proportion of random walk noise is proved to increase as time span increases, indicating that it is difficult to detect random walk noise with short time series, especially when the amplitude of random walk noise is small, while the flicker noise is dominant. Results shows that loading correction and CME separation have an obvious effect on noise model, indicating their regional changes in large spatial scale, and verifying the necessity of regional block filtering.