The Precision Assessment And Statistical Analysis Of The Post-fit Residuals In Data Reduction Of Space Geodesy

The solution of weighted least-square problem obtained with Givens-Gentlemanorthogonal transformation method is shown to be equivalent to that in statistical orbitdetermination theory in chapter 2. Using this method, covariance analysis quantities ofstatistical significance such as considered covariance matrix and sensitivity matrix arederived along with their propagation from initial epoch. It turns out that using thismethod, covariance analysis enjoys more flexibility in parameter combination to achieveoptimal solution, and better numerical stability.Based on the Givens-Gentleman algorithm for covariance analysis, a program todeal with real data, the CAPS, is coded. When applying to the GM solution which wasobtained in previous work, the errors of the station coordinates are considered. Theconsider sigamas are proven to be larger than the computed sigmas, which meansdegradation of accuracy. As another application of the CAPS, the achievable precisionof the solution of the sub-daily term (SDT) of Earth's rotation with SLR is assessed.The error of the Earth gravitational coefficient GM on SDT solution is considered. Also, todetermine the solvability of SDT with SLR, the case of treating SDT as considerparameters is analyzed.The latest progress in the space VLBI orbit determination and itsastrogeodynamic applications are reviewed in chapter 4. The Space VLBI observablesinvolve simultaneously radio celestial reference frame, dynamical reference frame andterrestrial reference frame. Therefore it is the only existing technique that is able toconnect the above three independent reference frames directly. As new, preciseobservable, space VLBI delay and delay rate can also be used to improve the accuracyof orbit determination. Theory of Covariance analysis to assess the possible precision ofconnecting the reference frames and stochastic perturbation model to treat the solarradiation pressure are introduced.Dfferent methods Of muIti-technique data combination and the precisionachievabIe are discussed in chapt6r 5. Independent of the RMS, the Pearson xz and themoments of the residuaIs as a sampIe Of Gaussian noise are used to assess thecombination soIutions. Based on these statisticaI data, the d6termination Of weightf8Ctors is addressed. The influence of the errors of consider parameters on combinationsoIutions and the relation betW6en consider covariance and reI8tive weighting factorsare aIso anaIyZed.A new statisticaI test of the postilit residuals is proposed in this paper. UnIike theusual residuaI anaIysis, The residuaIs are treated as a sample of a random variabIe,which specifically is a white gaussian noise in this paper. The higher oFder statisticaImoments and Pearson X' are computed to determine how far away a sampIe is fromwhite gaussian distribution. AppIying to the SLR data processing in chapter 6, we candiscrimate soIutions with the same leveI of RMS but significantly diff6renrt orbits andparameters. Furthermore, a criterion for choosing proper soIve-fOr parameters isdiscussed.The difficuIties to program the numerical method Of the computation Of statetransition matrix are anaIyZed in chapter 7. ln view of the praCtice of precise orbitdetermination and parameter solution, the difference algorithm tO compute statetransition matrix is proposed. A good approximation to the numericaI method, thediff6rence algorithm is easy to program, but the accuracy might be Iost near the initiaIepoch. The results of both methods are compared and the way to assure the accuraCy ofthe difference algorithm is discussed.To end this dissertation, aIl the resuIts obtained are summaried in chapter 8.Based on what have deveIoped, further work is aIso breiny disscused in this chapter.In summary, the fOllOwing three points are exploited in detail in this dissertation.. Based on the theory of consider covariance analysis and the aIgorithm developed inthis dissertation, the precision assessment...