What a two-line element set actually is
A TLE is a fixed-column text record of 69 characters per line. It carries a catalog number, an epoch, inclination, right ascension of the ascending node, eccentricity, argument of perigee, mean anomaly, mean motion, and a drag-like term called BSTAR.
Three things about that are routinely misunderstood, and they matter.
- These are mean elements, not a state vector. They are the output of a fit in which SGP4 itself is the model. Propagating them with any other propagator, or converting them to an osculating state vector and integrating numerically, produces invalid results. As one standard reference puts it, using these elements in a different general perturbations propagator gives completely erroneous results. The elements and the propagator are a matched pair.
- BSTAR is not a physical drag coefficient. It is a fitted pseudo-drag term that absorbs whatever the model could not explain over the fit span. It has no predictive value for future atmospheric density.
- The format is at the end of its life. Five-digit catalog numbers ran out on 11 July 2026, and new objects now receive six-digit numbers the TLE format cannot carry. The successor is the CCSDS Orbit Mean-Elements Message. The objects on this layer are all long-catalogued and low-numbered, so TLEs still work for them, but this is not the modern format.
Why accuracy decays away from epoch
SGP4 models the secular and periodic effects of Earth's zonal harmonics, a simplified power-law atmospheric drag, and for deep-space objects lunisolar gravity and resonance terms. GNSS satellites fall in the deep-space branch; the ISS does not.
Because the element set is a fit over a span rather than a measurement at an instant, and because the model deliberately omits most of the real force environment, accuracy decays in both directions away from epoch. The familiar rule of thumb is around a kilometer at epoch growing by roughly one to three kilometers a day, and it should be quoted as a rule of thumb, because that is all it is. The underlying validation study against precise GPS ephemerides found:
- In-track error dominates, then cross-track, with radial error typically far smaller than either.
- In-track error carries a real bias, so a symmetric error bar is wrong in both directions at once.
- Error characteristics for satellites in similar orbits can differ considerably, so there is no single correct kilometers-per-day figure even within one constellation.
Two practical consequences. Maneuvers break a TLE instantly: GNSS satellites perform station keeping and slot changes, the ISS performs reboosts and debris avoidance, and until a new fit is published the old elements describe an orbit the object has left. SGP4 gives no warning when this happens. And low Earth orbit degrades faster than medium Earth orbit, because drag is the dominant unmodeled force for the station and BSTAR cannot anticipate a geomagnetic storm inflating the thermosphere, while MEO satellites feel essentially no drag at all.
The four constellations
| System | Nominal constellation | Altitude | Inclination | Period |
|---|---|---|---|---|
| GPS | 24-slot baseline, expandable | about 20,200 km | 55 deg | 11 h 58 min |
| GLONASS | 24 operational, 3 planes | 19,100 km | 64.8 deg | about 11 h 15 min |
| Galileo | 24 operational plus 6 spares | 23,222 km | 56 deg | about 14 h 05 min |
| BeiDou (BDS-3) | 27 MEO, 3 IGSO, 5 GEO | MEO about 21,500 km | MEO 55 deg | MEO about 12 h 53 min |
We deliberately do not print a live count of operational satellites, because those numbers go stale and a hard-coded "31 GPS satellites" on a web page is wrong within months. The design constellation is stable; the current status belongs on the operators' own status pages.
One aside worth making: satellite-based augmentation satellites (WAAS, EGNOS, MSAS) are geostationary, not MEO, and are a separate group. They are easy to lump in with the constellations and should not be.
Why the station races and GPS crawls
Watch this layer for a minute and the contrast is obvious. It is worth understanding numerically, because the intuition that "the ISS is faster" is only a small part of it.
| ISS (low Earth orbit) | GPS (medium Earth orbit) | |
|---|---|---|
| Altitude | about 400 to 420 km | about 20,200 km |
| Orbital speed | about 7.67 km/s | about 3.87 km/s |
| Period | about 93 min | 11 h 58 min |
| Apparent angular rate near zenith | about 1.1 deg/s | about 0.011 deg/s |
| Longest visible pass | about 10 to 11 min | hours |
Two effects compound. The station is only about twice as fast in orbit, but it is fifty times closer, and apparent angular rate scales roughly as speed divided by range. That alone is two orders of magnitude. On top of it, a GPS satellite's motion is partly cancelled by Earth's own rotation in the same direction: the satellite advances 360 degrees in 11.97 hours while an observer advances 360 degrees in 23.93 hours, leaving a relative drift of only about 15 degrees of Earth-centered angle per hour. That is why GPS satellites hold station in the sky for hours, and why the constellation repeats its sky pattern twice per sidereal day.
Why a surveillance site plots satellites
The link is direct and does not need inflating. ADS-B is GNSS-derived by construction. An ADS-B Out broadcast is an aircraft reporting a position it computed from its own satellite navigation receiver, and the message carries explicit quality indicators that degrade before positions visibly jump. That is the basis of the entire GNSS interference layer.
Geometry is what moves accuracy. Position error is approximately dilution of precision multiplied by the ranging error. Ranging error is small and tightly controlled: the US commits to a global average user range error of 7.8 m or better 95 percent of the time, and observed performance has been far better. So it is the spread of satellites above the horizon, not the quality of any one signal, that mostly determines how well a receiver can do. That is a reason worth plotting.
Multi-constellation coverage is part of the mitigation picture for the interference problem documented by EASA since 2022, which is why all four constellations are shown rather than GPS alone.
What this layer does not tell you
Every data source has an edge. These are the ones that matter for reading this layer correctly.
- Positions are propagated from fitted mean elements, not measured. Expect kilometers of error, dominated by the in-track component, growing with time since epoch and differing per satellite.
- A maneuver invalidates an element set immediately and silently. Station keeping, slot changes and ISS reboosts all do this, and the propagator gives no indication.
- This is accurate enough to show which part of the sky a satellite is in. It is not accurate enough for antenna pointing, occultation timing or precise pass prediction days from epoch.
- It cannot give you a real dilution of precision figure. That needs satellite health and usability flags, which satellites your receiver is actually tracking, elevation mask, airframe and terrain masking and multipath. A figure derived from orbits alone is a geometric best case for availability, not a receiver's DOP.
- Plotting satellites does not measure interference. This layer is context for the GNSS interference layer, not a substitute for it.
- The element sets are redistributed from US Space Force tracking. Their accuracy and refresh cadence are not ours to control.
AeroScope is an situational awareness and research tool. Nothing here is approved for air traffic control, navigation, flight planning or collision avoidance. For operational decisions, use the authoritative source named above.