On the precision and accuracy of IGS orbits

In order to explore the precision and accuracy of International GNSS Service (IGS) orbits, we difference geocentric satellite positions midway between successive daily Final orbits for the period starting 5 November 2006, when the IGS switched its method of antenna calibration, through 31 December 2007. This yields a time series of orbit repeatabilities analogous to the classical geodetic test for position determinations. If we compare our average positional discontinuities to the official IGS accuracy codes, root-sum-squared (RSS) for each pair of days, we find the discontinuities are not well correlated with the predicted performance values. If instead the IGS weighted root-mean-square (WRMS) values from the Final combination long-arc analyses are taken as the measure of IGS accuracy, we find the position differences and long-arc values are correlated, but the long-arc values are exaggerated, particularly around eclipses, despite the fact that our day-boundary position differences apply to a single epoch each day and the long-arc analyses consider variations over a week. Our method is not well suited to probe the extent to which systematic effects dominate over random orbit errors, as indicated by satellite laser ranging residuals, but eclipsing satellites often display the most problematic behavior. A better metric than the current IGS orbit accuracy codes would probably be one based on the orbit discontinuities between successive days.