Track satellite debris with Google

By Phil Plait | February 26, 2009 12:17 pm

Rob at Orbiting Frog is too cool: he has software that allows you to track the debris from the recent satellite collision using Google Earth. You can follow the debris in real time! That is so excellent.

Debris tracked by Google Earth

This makes me wonder why the guys at NORAD and other trackers didn’t know this collision was going to happen before it occurred. There are a lot of satellites, and the orbits change, but it seems simple enough to write code that allow you to see if any two satellites come within, say, a kilometer of each other at any given time. You could run that every time you update the orbital parameters of the satellites. So I’m scratching my head about that. Anyone out there in BABlogland have any inside knowledge?

CATEGORIZED UNDER: Astronomy, Cool stuff

Comments (52)

Links to this Post

  1. Track space debris on Google Earth « The Cosmic Web | February 27, 2009
  1. Wayne

    No inside knowledge (or even sources) but I remember reading that the error bars for satellite positions are so much bigger than the spacecraft themselves that it’s nearly impossible to determine if a collision will actually occur and that this conjunction wasn’t even a particularly close one according to predictions.

  2. cid kilroy

    phil, satellite orbits are well-known, but we cannot waver in the face of alien terrorism. the rampant destruction of satellites and theft of the ozone layer has gone on long enough, i say!

  3. I could tell you, but then I’d have to kill you… Not worth it.

  4. dziban

    Who’s to say that NORAD *didn’t* know about the impending collision?

  5. Damir Tubin

    Much too many pieces of junk, the orbits are not well enough known for automatic analysis, and not enough manpower and funds to analyze it all. A much better explanation is on a Space review article:

  6. NASA Guy

    But Phil, who’s responsible for this? Is it NASA? Well, then money has to be allocated to pay for this. But why should NASA be responsible for commercial or foreign payloads … they should do it themselves. Then we have to get around the idea that there’s a public NORAD database of objects but you’d need access to the full database that includes “secret” payloads. That’s not easy to get. Then you have to get past the idea (noted above) that the orbit data has uncertainty in position and velocity. Fine, once we determine that there could be a possible problem between a commercial payload and a foreign payload (like what happened) NASA would have no authority to tell either entity to move their satellite (if that’s even possible). Furthermore, what if the collision prediction is between pieces of space junk who can’t maneuver out each others way. Then what? This problem is much, much bigger than writing some code to look for potential intercepts.

  7. Interesting take, dziban, who’s to say that NORAD didn’t really want to be rid of that Russian satellite. The sacrifice of one Iridum bird was a small price to pay. ūüėČ

  8. Phil, such software exists, even in the free domain actually (Rob Matson’s COLA and ALLCOLA software).

    The problem is that the orbits published by USSTRATCOM are not accurate enough to pinpoint a collision (they don’t publish their orbits at full accuracy). As these things move 7.5 km/s even small uncertainties in the orbit lead to a hit or a miss…. So there always is an element of “assesment” of the true risk involved. Compare it with assessing whether Apophis will hit or not.

  9. Actually, based on the published orbits with an epoch closest to the collision, ALLCOLA predicted them to pass at 800 meters distant to each other, with 100 meter height difference between the orbits. In reality, they hit each other as we now know. This gives you some indication of the uncertainties involved.

  10. Chris A.

    What worries me is that, even if we could see that a piece of the debris was headed toward Hubble, there’d be precious little we could do about it, being that Hubble has no on board thrust capability. How many other high-value satellites, I wonder, are similarly vulnerable?

  11. There are two reasons, as far as I know.

    One is that, as stated before, the error bars are much larger than the objects being tracked. It’s nice to see the things on Google Earth, but a single pixel in an image that shows the entire Earth represents a couple dozen square kilometers. The location data is good enough for looking at on a screen but not that good for tracking.

    The second is that comparing every combination of object A with object B requires the square of the number of objects being tracked comparisons (so for 10,000 tracked objects that’s 100 million comparisons), and each comparison is computationally expensive (you have to chart out where it’s going to be at every instant, including propagating error margins, for the next several days). NORAD doesn’t have the resources to do that for every object, so they concentrate on calculating whether really important stuff like the ISS and the Shuttle are going to be impacted, rather than bothering with trying to protect every privately-owned satellite in orbit.

  12. djcinsb

    In the wild rumor category, I have heard that Iridium did a station keeping maneuver shortly (within a day) before the collision. Given the error bars on the orbits, even a small perturbation could turn a close approach (e.g. 800 m) into a collision.

  13. IVAN3MAN

    Larian, it looks like you’re gonna have to kill Flooey!

  14. Great discussion here by a former orbital analyst who supervised people doing this sort of thing:

  15. The error bars for the tracked satellite elements (see; you need to register) are quite larger than 1 km…

  16. QUASAR

    WOW! The next cool thing would have to be tracking satellites with Google Earth!

  17. Charles


    You really cannot be sure that USSPACECOM didn’t know of a high potential for collision. They simply may not have said anything publicly. That’s far different than not knowing.

    I foresee a profitable private space enterprise going and cleaning up junk left in orbit. With over 19,000 pieces of present or future junk circling the globe, artificial meteors will be commonplace.

  18. rob

    i saw a satellite last night that was VERY bright (brighter than Sirius.) it faded slowly, then got bright again about a minute later, then faded like i have seen other satellites do as they rotate.

    it was 18:36 pm central time in saint paul MN when i saw it, so it wasn’t even very dark yet. it was travelling from about east, towards north, and passed right through the cup of the big dipper.

    i looked a little on line to see if there was software that could tell you what satellite you saw given time, lat/long and approximate path, but did not have any success. anyone have any suggestions to try to figure out what i saw?

    (i thought at first it might have been the ISS cuz it was so bright, but a check of possible ISS sightings for saint paul yesterday was about 12 hours before i saw the object.)

    any suggestions anyone?

  19. Check Heavens Above (add necessary prefix):

    You probably saw an Iridium Flare

  20. ccpetersen

    From what I heard from folks who DO track things, the error bars are of sufficient uncertainty that they may not have known the two were going to collide until it was too late. And even then, there wasn’t much they could do. The RussSat was dead; therefore unable to be moved. Not sure if the Iridium could have been moved quickly enough. Space is big… and so are the error bars.

  21. It is possible with sufficiently accurate position data and sufficient computational capacity. It might be better, for a visualization, to simply put a circle around each satellite and notice when they intersect.

  22. Togusa

    Some enterprising chap cooked up a presentation that uses the Google Earth API and browser plug-in to make this real-time visualization of some of the collision debris: (link opens a full-page presentation)

    The browser plug-in, which works on Windows and Mac OS X, can be found here:

  23. It was mentioned above that NORAD doesn’t have the computational power to devote to tracking all the satellites and bits of debris in order to look for collisions. I wonder if it would be possible to do distributed computing client, along the lines of SETI@Home or the many others out there, to track potential collisions? Errors in positioning being a known variable, couldn’t the results be plugged into a probability matrix that could account for the positional sloppiness through statistical analysis?

    It seems mathematically and computationally possible to come up with something moderately useable.

  24. QUASAR

    Hey, man! Check out Randi’s latest video on his YouTube channel!

  25. John Keller


    As Wayne pointed out, the error bars for any orbiting vehicles are much bigger than the vehicle itself. There are many factors, like drag, out gassing, off gassing and solar isolation which affect the orbit, but are difficult to accurately calculate. It’s almost an art to get it right.


  26. robodude

    the word on the street is that NORAD does in fact have collision software to report and provide warning for just such an occurrence. I have heard reports that the Russian satellite actually performed an orbital maneuver to intercept the Iridium satellite. It seems that the Russians may have falsely claimed that the satellite had failed.

  27. rob

    thanks Tom Hill! i will check out the site.

  28. Ray Wagner

    There is an article in this week’s “The Space Review” that analyzes the situation. There are two parts:

  29. Recent article at Cocktail Party Physics implies that sunspots are at least partly to blame:

    I don’t feel qualified enough to judge the accuracy of these statements. I’ll leave that to people smarter than me.

  30. “I have heard reports that the Russian satellite actually performed an orbital maneuver to intercept the Iridium satellite. It seems that the Russians may have falsely claimed that the satellite had failed.”

    Now, this sounds like a paranoid rant…Occam’s razor anyone?

  31. ccpetersen

    I’m with Changcho here — would robodude like to give us some more details on this “some guy says” report? Otherwise I’m thinking “trollish” stuff and wondering why the Russians would go after just ONE Iridium? This makes no sense. So, I’m going with my original inclination – this is a trollish statement without any literary or factual merit.

  32. Avi

    It seems to me that it may be an issue of computer power. We’re talking about 10s of 1000s of pieces being tracked. To perform the analysis you’re talking about, you’d have to compare 10s of 1000s of trackfiles against 10s of 1000s of other trackfiles – resulting in over 100 million comparisons. A super computer could probably do it, but its no easy task.

    If you combine that with the precision issues mentioned in previous posts, you get a system that’s full of holes.

  33. Michelle

    What I wonder is, is any of these pieces big enough to be seen from down here? If we could see the toolbox of some astronaut lady (Not to the naked eye but you know…) we can probably spot a piece of this thing, right???

  34. Interesting twist here is that the Iridium satellite in question was launched from Russia, and it’s my understanding that the space treaty makes the launching nation responsible for the satellite. Granted, that was written in the days when international agreements to launch commercial communications satellites were not even dreamed of, but it’ll give the lawyers something to chew on.

  35. It doesn’t seem to be that hard, except that LEO is so littered with debris that the alarms would be going off constantly.

    There are unpredictable forces at work, the upper atmosphere drags orbits by a small but unpredictable amount. Then there’s the satellite’s own stationkeeping which was active on the Iridium bird, which also alters the orbits by an unpredictable (to NORAD) amount automatically.

    Add the above two to that NORAD does not track everything in realtime (last I heard was that they tracked each object in rotation over the course of several months to get several locations and update their parameters) and you have error bars of several hundred metres at the very least, no good when you’re dealing with objects between the size of a briefcase and a bus.

  36. A friend of mine and I were talking about this. He made a good point, the velocity vectors of the satellites were almost perpendicular at impact. If either satellite had arrived even a fraction of a second later, there would have been no collision. This really was one in a million.

    Suppose I stand at home plate on a baseball field and fire a bullet towards 2nd base and you stand a 1st base and fire a bullet at 3rd base. The odds that the bullets collide are infinitesimally small.

    I noted that the computational complexity of orbital collision calculations is reasonably hard. It is an O(n^3) algorithm. This means that the number of calculations required to predict a collision scales with the cube of the number of satellites. For any single satellite, it is O(n^2).

    Compound this by the fact that a lot of orbital software runs on old, slow computers. At my work, even though we generally do computations on computers that are new or only a couple of years old, we normally have to dust off the early 90s sun workstation to do our orbital calculations. So sadly, it isn’t as easy as one might hope.

    An easy solution to the computational problem, is to prune calculations that involve orbits with high relative inclination, because these are unlikely to collide. This brings us back to the first point. This technique leaves you with a small possibility of collision with another object at high relative inclination.

  37. MHS

    I think I read on that USSTRATCOM has a list of potential collisions and that the Kosmos-Iridium encounter was not even in their top-150. That makes you wonder what’s the point of it all…

  38. Bakafish

    I am not a physics major, but it would seem to me that you would only have to do calculations on objects that have intersecting planes and overlapping orbital distances. I’d venture to guess that there are a lot of simple optimizations you could make to the algorithm that would allow you to eliminate most of the comparisons. It’s only O(n^3) if they are all in intersecting plains and altitudes, which they most certainly are not. We also can do very rough calculations first (margin of error ~5 KM) to discover the possible collisions, and then dedicate the horsepower to actually find out if they really get close. You might as well remove all the possible collisions that can’t move (i.e. two objects that are out of hydrazine) since there isn’t anything we could do anyway :-)

  39. BigBob

    Satellite tracking software has been available for a long time now. AGI’s Satellite Tool Kit is used for visualisation of these birds and is very sophisticated.

  40. @ robodude:

    “See, the commies are at it again!”
    Dude, don’t forget your tin-foil hat when you go out…

  41. There is a cool slide show at the Wall Street Journal (of all sites!) that gives a nice description of the space debris problem. The slideshow is embedded in the following article:

  42. dre

    I’m wondering… on the overlay provided in the link, the orbits of the debris from each individual satellite still exhibit narrow chokes or bottlenecks, which I presume are ghosts of the original impact – the fragments of a given satellite scattered on impact but reconvene each orbit to some degree bacause of still-nonintuitive-to-me orbital dynamics.

    What I’m unsure of is why the two bottlenecks representing impact have drifted away from each other. What orbital mechanism leads to that? I know that Earth continues to rotate, and so I didn’t expect to see bottlenecks over the same spot in Siberia. Why don’t the Iridium and Cosmos bottlenecks still overlap? That’s what I’m getting at.

    Dang, it’s hard to ask the question when I don’t know the terminology.

  43. Anytime you apply a change in velocity to an object in orbit, that object will return to the same location repeatedly. In this case, some pieces were put into lower orbits, some into higher orbits, and other pieces had their inclination changed. They all come back around to the same point in inertial space, though.

  44. gwiz

    Mainly because the earth isn’t a perfect sphere, but also because of lunar and solar perturbations, satellite orbits are not fixed in space. The plane of the orbit rotates about the earth’s axis and the orbit rotates within its plane.

  45. Mach25

    Rob – Judging from the time and location you specified, the most likely candidate for your sighting is SAFIR 2 – a German communications satellite. You can use the following website to do your own analysis:

    Once you select your ground location, go to the SkySearch tab to specify a time slightly before your sighting. It will take several minutes to go through the entire catalog, but SAFIR 2 comes up. You can select it on the SkyLog tab, press “More Info”, and then “SkyTrack”. It will bring up a graphic of the sighting – complete with SAFIR 2 passing through the “cup” of the big dipper.

    As for the collision, USSTRATCOM would have been happy to provide Iridium with a conjunction assessment…if they had bothered to ask. It would also have helped if Iridium had provided information about any maneuvers it may or may not have performed in the hours prior to the collision. You can be sure that USSTRATCOM has certainly looked (after the fact) at what the conjuction would have been given that information.

  46. PopsicleMud

    I sort of wondered about the ability to predict collisions, but it sounds like given the uncertainty involved, by the time you’re able to make an accurate prediction, you don’t have enough time before the collision to do anything more useful than yell, “look out!”

  47. Calli Arcale

    There’s an additional problem with predicting collisions. The error bars are high enough that not only do you make a lot of false-positive reports, but there are so many false-positive reports that if they were to respond to all of them, they’d expend all their available propellant in short order. (And that’s even assuming they *have* onboard propulsion. Not all active spacecraft do. Hubble, for instance.)

    Plus, of course, each avoidance maneuver screws up what is known about the spacecraft’s orbit, increasing the error bars on the *next* conjunction . . . .

    In the end, at the current rate of collions (1 ever), it’s more cost effective to cross your fingers and hope than to try to respond to every possible collision.

  48. Calli Arcale

    Note: I should clarify that the current known rate of collisions is 1 collision between two intact spacecraft ever, excluding deliberate collisions and screwed up docking attempts (like Progress versus Mir). There have been a number of smaller collisions.

  49. @Bakafish

    Because all orbital planes must be centered on the center of mass of the planetary body, all orbital planes intersect. So you can’t really ignore any orbital planes, but orbital planes that have a large relative inclination have a low but non-zero probability of intersection.

    This was exactly this sort of collision improbable but not impossible collision.

    Of course, independent of the algorithmic discussion, I tend to agree with what other have been saying that the uncertainty in the equation was high enough that they couldn’t have predicted it anyway. The high relative inclination, only increases that uncertainty.

  50. Buzz Parsec

    I get “Forbidden” messages (never seen *that* before!) when I click on any of the orbitingfrog links. Are they over loaded or have the black helicopters stolen their tin foil hats?


Discover's Newsletter

Sign up to get the latest science news delivered weekly right to your inbox!


See More

Collapse bottom bar