A long arc approach to GPS satellite orbit improvement

Thumbnail Image


Journal Title

Journal ISSN

Volume Title



The object of this thesis was to design a model, implement software, and test the model and the software for improving the accuracy of the orbits of the satellites of the Navstar Global Positioning System (GPS) using double difference GPS phase observation. A dynamic long arc approach is used and satellite orbits are integrated continuously over multiple days. The model includes: The coordinate system transformation between the geocentric inertial coordinate system and the Earth-fixed coordinate system; the modelling of the forces acting on the GPS satellites; the integration methods for the solution of the equations of motion for the GPS satellites, and the partial derivatives of the satellite position vectors with respect to initial state vectors and dynamical parameters; the designation of unknown parameters which are solved for in data processing; the handling of biases in GPS observations; and the adjustment and computational algorithm. A software package associated with the above model has been developed for the Macintosh computer family. The software development is partially based on the GPS Differential Positioning Program (DIPOP) package of the Department of Surveying Engineering of the University of New Brunswick, and the new software package is called DIPOP-E (Enhanced version of DIPOP). The main-processor of DIPOOP-E differs the new version, although DIPOP-E inherits most of the features of the main-processor in DIPOP; the pre-processor has been modified only to accommodate new features in the main-processor, and to improve operational efficiency for processing large data sets. A new special utility tool has been developed for the cycle slip detection and correction. The tool uses window, menu, button, mouse, and graphic display features of the Macintosh computer, which greatly enhances the operational efficiency of the package. Other auxiliary programs have been developed to facilitate the use of the main-processor. The model and software development have been tested thoroughly with the Standard GPS Data Set of the International Association of Geodesy Special Study Group 1.104. All test show that the most accurate results were for the latitude components, followed by the baseline lengths, the longitude components, and finally height components. This [phenomenon is due to the sky distribution of the GPS satellites. The optimistic results of formal uncertainties show that there are some remaining systematic biases in the model. The daily repeatabilities show that the latitude component and baseline length determination were mostly better than 0.1 ppm. For the regional and continental stations, 0.05 ppm level repeatabilities were achieved for the lengths of baselines. The worse results for the short baselines (in terms of ppm) and for the height components may be due to the residual tropospheric effects as no observed meteorological data have been used in data processing.