How INTEGRAL was 'lost' and became a space rock
15 December 2015
When INTEGRAL changed its orbit last January/February not everywhere was its new orbit info (Two-Line Element sets, TLEs) updated. At the end of November, Jonathan McDowell (Harvard-Smithsonian Center for Astrophysics) tweeted that asteroid-observing astronomers had been tracking an unidentified object, WJ2B5AC, with a 3831 min period. It had been found by the Catalina Sky Survey, while looking for asteroids. It was checked against known satellites using TLEs, but got no match. And thus it became a candidate space rock.
In June, however, it was noted that the orbit looked a lot like that of a satellite with COSPAR identification 2002-048A (or NORAD identification 27540) and C, i.e., the INTEGRAL spacecraft and booster. And, so, subsequently, it was realized by Jonathan McDowell that the TLEs were in fact not up-to-date, and that they were still calculated using the old orbit. In the end, he contacted ESOC to send the new orbital elements.
Below is Jonathan McDowell's account of the story.
On 12/12/15 10:01 PM, Jonathan McDowell wrote:
There are 17000 artificial objects currently being tracked in Earth orbit by the space surveillance network of the United States, some smaller than 10 centimeters in size. The orbital data are made public as 'Two Line Elements' (TLEs). It's an impressive catalog, and pretty much complete for objects in low orbits around the Earth. It's much less complete for objects in high orbits - anything further out than 15 Earth radii or so - and no-one is responsible for tracking space junk beyond Earth orbit entirely.
Meanwhile, astronomers search for natural Near Earth Objects (NEOs): asteroids that could pose an impact threat to our planet. In recent years these searches have become more comprehensive, with the most successful current effort being the Catalina Sky Survey (CSS) which monitors the northern sky from Arizona.
An asteroid in heliocentric orbit near the Earth has an apparent motion that's not too different from a piece of space junk in loosely bound Earth orbit, so sometimes CSS picks up geocentric artificial objects. New discoveries are posted to the NEO Confirmation Page at the Harvard-Smithsonian's Minor Planet Center with a temporary designation assigned by CSS. Once enough observations are in, the ambiguity can be resolved: it's in geocentric orbit, so it is thrown away and never given a proper asteroid designation. Objects at the Lagrange points would occasionally slip through and be given an actual 'Minor planet provisional designation'. This happened a couple of times to the WMAP spacecraft; the asteroid would cause consternation when it then proceeded to make a course correction. But there's less danger of that with a fully geocentric object like, for example, INTEGRAL.
Normally the artificial object's orbit is quickly matched with that of a known satellite in the US space object catalog, but occasionally one shows up with no known match. It's not given an asteroid designation, and retains its temporary 'unknown source' Catalina survey name.
This October, for example, Catalina object WT1190F was picked up and determined to be in an orbit with apogee beyond the Moon, but on a trajectory which would cause Earth impact a few weeks later - which it duly did, reentering over Sri Lanka. A search for earlier observations showed that WT1190F was first picked up in 2009 and has been making occasional close passes of the Moon which rendered its orbit essentially chaotic, preventing a definite ID. (I suspect it may have been the rocket stage from the Lunar Prospector mission). Soon after that case was closed, or at least shelved, a new mystery object was investigated. Catalina object WJ2B5AC was first spotted in May 2015 and then followed up by observers in the Canary Islands, California and by amateur Peter Birtwhistle in England. Bill Gray, at his web site projectpluto.com, coordinates followup for the artificial objects discarded by the asteroid astronomers. He contacted me, since I maintain a list (so far unpublished, but I'll get around to it) of known lost artificial objects.
WJ2B5AC puzzled us at first. It was in an orbit quite similar to that of the Chandra observatory - but I work on Chandra, and know exactly where it is. The handful of known 'lost spacecraft' with similar orbital altitude included 1964's Orbiting Geophysical Observatory 1 (OGO-1), but that didn't seem to be a good match. We quickly spotted that the orbit had some similarities to that of the INTEGRAL satellite and its now-reentered Proton Blok-D upper stage. Maybe something had fallen off INTEGRAL?
Jonathan to Bill, Nov 22:
"I plotted the period and inclination of Chandra, INTEGRAL, INTEGRAL R/B and OGO-1 over time. Marked your reconstruction of the object's orbit as red crosses. The orbital period agrees well with Chandra and OGO-1, the inclination mirrors INTEGRAL. For what it's worth."
A week later, something clicked in the messy filing cabinet of my brain. I recalled seeing a mention of a planned orbit change for Integral back in January; at the time, I had followed its TLEs expecting to see the change and been puzzled not to see any evidence of it - then forgot about the issue. When I plotted the TLEs for INTEGRAL versus time, I saw that the orbital period values became a little noisy after February. I was aware that for these high orbit objects, US tracking would sometimes lose them and issue TLEs based on orbit propagation rather than new observations, that could then be used by their analysts to recover the object. I was also aware that the TLE format is sadly lacking in important metadata - in particular, there is nothing whatsoever to tell you if an orbital solution is based only on a model and not on actual observations. Further, I knew that the lack of attention to high altitude objects meant that incorrect orbital data for months or years would not be unprecedented.
I tried downloading a recent INTEGRAL ephemeris file from the science data archive, but the state vectors were given versus a spacecraft clock time and the file's metadata didn't give an obvious way to map that to JD or calendar date. Probably could figure it out with more research, but that seemed like too much work. I mined the INTEGRAL web pages for details of the burns, which confirmed they had occurred as scheduled but didn't give accurate parameters. In the meantime, I messaged Bill again.
Jonathan to Bill, Nov 30:
"I had a thought. According to ESA, INTEGRAL lowered its orbit from a 4310 min period to approx 3840 min period in Jan-Feb 2015. I think the TLEs are tracking a fictional object and you have found the real INTEGRAL - is it bright enough for that to be the case?"
Bill quickly wrote back confirming my suspicion: assuming a 10 percent albedo, the object was about 10 meters in size; the perigee times of INTEGRAL's orbit could be found from ESA's SkyMap application and were within a minute of the times derived from Bill's orbit solution for WJ2B5AC. The JPL Horizons web site also had orbit data for INTEGRAL, but it turns out these were seeded with the TLE data, so were giving the same wrong positions.
We informed Eric Christensen at the Catalina Sky Survey that we'd figured out what his object was. Now I wanted to get the TLEs and Horizons corrected, so that future accidental observations of INTEGRAL could be automatically filtered out. Fortunately I had contacts at ESOC who were happy to help; Benjamin Bastida Virgili kindly agreed to forward a set of state vectors to the Joint Space Operations Command (JSPOC); I alerted Jon Giorgini at JPL who runs Horizons so that he could update his models. As of Dec 2, the US Space-Track.org site began sending out reliable TLEs for INTEGRAL. Case closed - but the episode reminds us that the deep Earth orbit where INTEGRAL ranges is still a Wild West frontier for space tracking!
- Jonathan McDowell