Even with current elements, pass predictions with Celestia will never be as accurate as with SatTrak or some other specialized program. It's because the NORAD TLEs are really designed for and meant to be used with a specific algorithm, the 'SGP4 model', that most satellite prediction programs use but that Celestia doesn't. When you convert TLEs to Celestia format, you basically ignore the first entire line of information, which has data about the drag and perturbations that the satellite experiences. Celestia just uses the 'mean elements', which means that it only models the 'average' elliptical orbit, without the alterations to that orbit caused by the sun, moon, non-sphericity of the earth, atmosphere, etc.
Because of this, Celestia will be most accurate for mid-altitude, highly circular orbits. LEO, GEO, and highly elliptical orbits will be signifigantly (though I dunno how much) less accurate.
For each new set of TLEs that NORAD generates, though, the perturbations are averaged out, and thus Celestia's model should be pretty accurate for the actual orbit of the prediction, and decreasingly accurate in either direction.
It would be extremely cool if Celestia allowed something like...
in an SSC.
FYI, I'm currently in the process of making a /large/ addon SSC that includes all 9500+ objects in orbit currently tracked by NORAD. It'll be several megs, but I'm setting it up so I'll be able to regenerate it occasionally with updated TLEs. My current version is pretty rough, but it shows the 'geosynchronous belt' quite nicely.
Unfortunately, no one location on the web that I can locate provides a unified listing of all 'official' TLEs anymore (apparently NORAD complained), so I'm having to use Orbitel to concatenate a large number of shorter lists. It's kindof a pain.
)
