The astrometric analysis of the Gaia observations can only be performed to the highest accuracy if one knows where the spacecraft was at the time of observation. Requirements are very stringent for the orbit tracking and highly challenging for the teams involved at ESOC. At the very end one expects the position of Gaia to be known within one hundred metres uncertainty and its velocity at the mm/s level. We know today that this challenge will probably be met thanks to the combination of the radio, Doppler and optical tracking of the spacecraft.
But the data processing is on-going and orbital data are available at any time, based on a preliminary solution using the tracking up to the date close to the orbit release (typically the orbit is updated every week), and extended with a predicted orbit covering the whole mission. The requirement was such that the true orbit should depart from the planned orbit determined a few days after launch by no more than 7000 km. This limit was set by the design of the relativity experiment requiring the apparent direction of Jupiter as seen from Gaia could be ascertained at launch time to be within 1/10th of the planet radius.
The delivered orbit file comprises three time-segments: (i) a finally reconstructed orbit fitted on tracking data and ending typically one week before the release; (ii) a provisionally reconstructed orbit covering about the last week of tracking and likely to move into the first segment at the next delivery; (iii) a predicted orbit extending from the delivery date to the nominal mission end, which will change from delivery to delivery. The first segment is final and will not be updated in the subsequent deliveries, until late into the mission when all the optical observations processed with the Gaia catalogue can be integrated. In the following discussion the first two segments are merged into a 'reconstructed orbit' for the sake of simplicity.
The plot shows the positional difference between two orbit releases. One from 12 May 2015 and a more recent one delivered on 1 September 2015. The first one represents the reconstructed solution until around 12 May (and a finally reconstructed until 5 May) and a predicted orbit afterwards, and similarly for the September-solution. Therefore they have a common part, until 12 May, followed by an interval until 1 September, during which we have a predicted orbit and a true orbit fitting. Finally beyond this date, we have two realisations of the predicted orbit, which have no reason to be identical. These intervals are identified at the top of the diagram along with the kind of orbits which are compared. The orbits are given with the origin at the barycenter of the Solar System in equatorial coordinates. The coordinate differences are plotted in km.
During the overlap of the fitted periods, the two orbits are fully identical (and almost identical in the last week, the difference being not appreciable in the plot). This was expected and seeing it in real is more than reassuring.
In the intermediate period one sees that the (nearly) final orbit reconstruction departs gradually from the predicted orbit by 300 km at most, a very nice achievement compared to the requirements. One should notice that the numbers should not be understood into 'how well one can follow a targeted orbit'. This is not the role of the predicted orbit.
Finally one sees in the last interval that the two predicted orbits differ from each other by less than 600 km, and the separation remains in this range over a longer timespan.
A previous comparison with a much earlier initial orbit (October 2014) shows that the difference between the predicted and the final orbit has dropped by a factor 3, down to 400 km from 1500 beforehand. Although it can go differently in the future, it is nice that we can rely on a predicted orbit so close to the final orbit for the assessment of observations of solar system bodies and to make observation predictions long in advance with a sufficient reliability. Unlike the stars, the apparent direction of nearby minor planets is very sensitive to the exact location of the observing platform. Likewise events like the passage of the Moon on the solar disk can be predicted more safely with a reliable predicted orbit.
Credits: ESA/ESOC/DPAC, F. Mignard (Observatoire de la Côte d'Azur), S. Klioner, A. Butkevich, (Lohrmann Observatory, Dresden), F. Budnik (ESOC) and the Orbit Determination Team