Empirical modelling of site-specific errors in GPS observations

Abstract

GPS is an essential element of the global geospatial information infrastructure, it is free, open and dependable. Precise positioning and navigation enabled by GPS has led to the development of hundreds of applications affecting every aspect of modern life and is now found in everything from mobile phones to bulldozers.

Underpinning the day-to-day operation of GPS is the International Terrestrial Reference Frame (ITRF). Without an accurate earth-centred, earth-fixed reference frame, such as ITRF, it would not be possible to accurately determine station location and position as a function of time. To achieve an accurate reference frame precise models of all aspects of the GPS system are required, including; the satellites, their orbits, the signal propagation medium, the ground receivers and antennas, and the orientation and motion of the Earth's crust.

For more than two decades GPS observations have been integral to the determination of the ITRF. GPS is the critical technique that provides the connection, through collocation, between other terrestrial observation systems, SLR, and VLBI necessary to define accurately the origin, orientation and scale of the ITRF. GPS solutions provide the most precise and accurate estimates of polar motion and is the geodetic technique most commonly used to access the ITRF. The main weaknesses of GPS observations today are due to unmodelled site-specific errors, particularly at collocated stations, orbit mismodelling errors (such as solar radiation pressure), errors in the conventional model for diurnal and semi-diurnal variations in Earth orientation due to ocean tides (griffiths2013), and an under-determined TRF scale due to uncalibrated satellite antenna phase centre offsets

Analysis and modelling techniques have continuously been refined and improved. Despite these advances, there has been little progress on addressing site-specific biases in GPS processing. In this thesis, we are mainly concerned with site-specific biases due to reflections of the incoming GPS signal, as well as errors in the antenna model. These site-specific errors can alias into the GPS station position time series producing time-correlated errors which do not average out over time. The result is a GPS time series which will have unmodelled biases that can affect the interpretation of geophysical signals. This is particularly a problem for reference frames if there are site-specific biases at GPS stations used to collocate the different observation techniques.

This thesis presents a methodology that can account for site-specific errors at the observational level, which is applicable to historic and future data sets. The technique relies on using carrier phase residuals obtained from the processing of a large network of GPS stations. These residuals are then used to model the errors at individual stations, and those associated with individual satellites. We have investigated the applicability of carrier phase residuals to model site-specific biases, through the use of simulations. The technique has then been tested and verified by applying the models to short-baseline kinematic solutions for 3 different collocation stations. We also investigate the impact of applying the model to large global solutions, in particular, we investigate the impact upon coordinate and velocity estimates as well as orbit and clock products, key products used to access and determine the reference frame. Show more