Realization And Evaluation Of GNSS Terrestrial Reference Frame Following The Function Model And Environmental Loading Correction

On the one hand,Terrestrial Reference Frame(TRF)is fundamental for locating or tracking objects around the Earth and for the earth science research.On the other hand,TRF provides the datum for wide range of applications which involve economic development and national defense building.The recent resolution adopted on 26 February 2015 by the General Assembly of the United Nations for Sustainable Development recognized that the adoption of the accurate and reliable Terrestrial Reference Frame is of critical importance for science and society.Therefore,one of the main tasks for geodesy is updating and refining the current TRF.Nowadays,the GNSS contribution to the TRF is fundamental in several respects.Firstly,GNSS provide the densest monitoring network of grounded based receivers for the realization of the TRF.Secondly,GNSS are highly available via which a vast majority of users access the TRF products.Thirdly,GNSS play a primordial role in the elaboration of the TRF,as the link between the three other techniques.Fourthly,the definition of the TRF orientation and its continuity between the successive ITRF realizations is also ensured by GNSS.Fifthly,GNSS are the major contributing technique to the TRF input data.However,there are still some problems in the investigations of realizing and evaluating TRF using GNSS: what is the current state of GNSS TRF origin and scale after the IGS second reprocessing campaign;the traditional GNSS data processing strategy,that is dividing the whole network into several sub-networks and then combining them on the normal equation level,is not rigorous in theory;how to select the optimal parameterization(mathematical models)to realize a more accurate TRF;how best to evaluate different TRFs.In view of the above,this paper mainly completed the followings: obtaining geocenter motion and scale offsets of different IGS analysis centers(AC),and then analyzing the current GNSS TRF origin and scale to provide a reference for the TRF realization;implementing reprocessing of a global GNSS network using the latest GNSS data processing models and an integrated strategy to obtain accurate and reliable TRF input data;realizing GPS TRFs with different parameterizations,and then proposing a new parameterization for the TRF realization;studying the impact of different noise models on the TRF and proposing a new method for evaluating different TRFs.Main research contents and contributions are as follows.(1)This paper elaborates the theory and methods of the realization of GNSS TRF systematically.The GNSS data processing methodology using the undifferenced observable model is introduced,and the latest models and strategies of GNSS data processing is also studied.The mathematical descriptions of several common noise models used for GNSS coordinate time series analysis,including the variance covariance matrix and the power spectral density function,are given in detail.The realization of ITRF2014 is mainly elaborated and deduced,including the processing and analysis of the original observations of the contributing techniques,the intra-technique combination,the long-term stacking of the combined solutions,inter-technique combination,the definition of the datum and the additional constraints.(2)This paper investigates geocenter motion derived from the combined solutions and 6 individual Analysis Centers solutions of the IGS second reprocessing campaign using the network shift approach,assesses these GNSS geocenter estimates by comparison with independent estimates from SLR and SLR-GRACE,and then analyzes the current status of the GNSS TRF origin.The results show that the offsets at the reference epoch and the rates of the GNSS geocenter motion are within the estimated uncertainties of the origin and origin rate of ITRF2014,but the offset in the Z direction reaches 4mm.GNSS annual geocenter estimates are in reasonable agreement with SLR estimates in the X and Y directions.In the Z direction,however,the annual signals derived from the IGS solutions using empirical solar radiation pressure models disagree with SLR estimates(CODE,GFZ,MIT),except for three particular ACs using a priori solar radiation pressure models(EMR,ESA,JPL).This suggests that a priori solar radiation pressure models used by these ACs may constitute an improvement over the conventional strategy employed by the other ACs.The background noise in GNSS and SLR geocenter time series finally appears to be correlated(0.3?0.4),suggesting that it might partly reflect real,aperiodic geocenter motion.Considering the significant difference in the Z direction between GNSS geoceneter motion and SLR results in terms of annual signals,GNSS alone could thus not suffice to completely define an accurate TRF origin.(3)Considering the latest satellite antenna phase center correction model(from igs08.atx to igs14.atx),this paper analyzes the scale offsets calculated from the combined solutions and 7 individual ACs solutions of the IGS second reprocessing campaign with respect to the IGS14 using different satellite antenna phase center offsets(z-direction),and investigates the impact of different satellite antenna z-PCOs on noise characteristics,the long-term rate and seasonal signals of scale offsets.The results show that GNSS scale offsets exhibit correlated noise which is better represented by a white plus power law noise model.The rates estimated from the scale offsets fluctuate between-0.22 and-0.13mm/y depending on the different ACs.The annual signals are significant,and semi-annual signals are weak.The change of satellite antenna z-PCOs has a significant effect on the mean scale offset with the maximum difference reaching 0.5ppb,but there are no obvious variations of noise characteristics,the long-term rate and seasonal signals.(4)This paper assesses the impact empirically by injecting progressively larger numbers of artificial offsets and solving for a series of long-term secular GNSS frames in terms of the TRF stability,velocities and seasonal signals quantitatively.The results show that the impact of the offsets existing in 340 IGS stations on the TRF stability,horizontal velocities and annual amplitudes is negligible,with significant effects at the formal error level.On the other hand,the vertical velocity estimates are more seriously affected(0.27mm/yr).As the mean vertical rate uncertainty is 0.18 mm/yr for all the IGS stations for the current set of position offsets.(5)This paper designs a complete strategy for the initial data processing and the analysis of GPS TRF,implements the reprocessing of a global GNSS network including the CMONOC(Crustal Movement Observation Network of China)stations using the latest GNSS data processing models and an integrated strategy,and then obtains accurate and reliable input data for the realization of GPS TRF.The results show that for the GPS daily solutions,the degree of freedom of the unconstrained normal equation matrix is 3,in accordance with the theoretical result that GPS could not define the TRF orientation;the standard deviations of the origin,scale and orientation contained in the daily solutions are about 1mm,0.1mm and 1m,respectively;The WRMS values of the differences between the pole coordinates(in the X and Y directions)and their rates,and the length of day(LOD)derived from this paper and IGS combination results are 0.034 mas,0.038 mas,0.130 mas/day,0.197 mas/day and 0.034 ms/day,respectively;the WRMS values of the differences between these results and the IERS C04 are 0.088 mas,0.118 mas and 0.082 ms/day(including pole coordinates and LOD),indicating the reliability of the Earth Rotation Parameters derived from this paper;in the E,N,and U directions,the average WRMS values of the GPS solutions in this paper with respect to IGS14 results are 2.86 mm,3.37 mm and 7.27 mm,respectively.The average WRMS values with respect to the IGS combined solutions are 2.52 mm,3.10 mm and 5.60 mm,respectively,indicating that the GPS reprocessing results are stable and in good agreement with IGS combined solutions.(6)A new TRF realization method of combining the function model and environmental loading correction is proposed,and GPS TRFs with different parameterizations are realized,respectively.The post-seismic deformation(PSD)models of some stations are established and used as the input data for the GPS TRF.The results show that PSD models can effectively correct the slow nonlinear deformation of the station after the epoch when the earthquake occurs.FREQ2016 based on the function model,and GFZ2016 and EOST2016 considering the environmental loading correction are realized.The results show that at the reference epoch 2010.0,these three GPS TRFs datums are consistent with the IGS14 datum;the WRMS averages of the postfit residuals of FREQ2016 are smaller than those of the GFZ2016 and of the EOST2016;the horizontal velocity field differences of these three are almost negligible(< 0.08mm/yr),and the large vertical velocity differences are for some stations and may reach 1mm/yr;the reduction of FREQ2016 velocity formal errors is about 0.01mm/yr.This paper suggests that FREQ2016 is better than GFZ2016 or EOST2016.In addition,GFZ2016?1 and EOST2016?1 combining the above two parameterizations are realized.Compared with FREQ2016,the WRMS averages of the postfit residuals are similar in the horizontal components and decrease by 0.2 mm in the vertical components;the horizontal velocity differences are not significant(< 0.06mm/yr),but the vertical velocity differences of isolated stations are relatively large(< 1.1mm/yr);the velocity form errors are nearly unchanged.Only from the WRMS perspective,GFZ2016?1 and EOST2016?1 are better than FREQ2016,and in terms of the velocity,the difference between these three is almost negligible.(7)A new method of the TRF evaluation,namely the analysis and evaluation of the TRF prediction capability,is proposed and the feasibility of this method is verified.This new method is able to answer two questions: First,the TRF with which parameterization can provide more accurate station prediction coordinates in the future;secondly,the predicted coordinate uncertainty computed from which kind of stochastic models can describe the predicted coordinate errors more realistically.The results show that the prediction effect of GPS TRFs realized without considering the GPS draconitic signal is better than that with estimating the GPS draconitic signal;with two parameterizations,namely combining the function model and the environment loading correction,the GPS TRF prediction coordinates accuracy is better.The prediction coordinate uncertainties derived from the flicker noise model are the closest to the prediction errors,indicating that the flicker noise model is better than other colored noise models selected in this paper,and the optimal noise model determined by the AIC/BIC method is also the flicker noise.At present,the GPS TRF with above two parameterizations can provide more accurate prediction coordinates;the prediction coordinate uncertainties derived from the white noise plus flicker noise model can describe the prediction errors more realistically.In the future,the GPS TRF needs to consider the impact of the flicker noise.