ALTAI

How this works

Reentry data comes from Space-Track.org (US Space Force), checked about once an hour:

• TIP messages (Tracking and Impact Prediction) give a predicted reentry time, a ± window, and the point on the ground under the object at that time. They're issued for some objects in their final days, and updated as reentry gets closer.
• Decay records confirm that an object has come down (usually to the day), and give rough predictions for others weeks ahead.

The map shows the final 30 minutes of every reentry in the chosen period that has a known location. Select one for its full track: the last one and a half orbits, coloured by altitude, with the uncertainty window dashed. Upcoming reentries in the last day are kept on the Upcoming side until Space-Track confirms them.

Visibility. A reentry is brightest around breakup, near 80 km, and its fragments keep burning a little longer. Altai counts the towns (15,000+ people) where, from 3 minutes before the reentry time to 1 minute after, the fireball rose at least 10° above the horizon while the Sun was at least 6° below it. "Likely seen" means at least 50,000 people were in that position. When the time is uncertain (a wide TIP window, or only the day is known) the same is worked out for times spread across the window, and the percentage is the share of them that would be seen, assuming any time in the window is equally likely. When only the day is known, that comes out at around 25–35% for almost everything.

• It assumes clear skies and a fireball bright enough to notice; it doesn't know the weather or how big the object is.
• It assumes an uncontrolled reentry. Some spacecraft, such as cargo ships, are steered down over empty ocean on purpose; their chances here are overstated.
• People outside larger towns, at sea or in aircraft aren't counted.

Mass, and the metal it leaves behind. Space-Track gives only a radar size, so each object's mass comes from Jonathan McDowell's GCAT catalogue, matched on its NORAD number: its dry mass (the best guess at what's left at the end of a mission, without propellant), or its launch mass when GCAT has no dry mass. Fragmentation debris almost never has a mass there, and objects from the most recent launches aren't in GCAT yet; they're counted as unknown mass, shown next to every total, and left out of the kilograms rather than guessed. So the totals are a floor.

Most of that mass doesn't reach the ground: it melts, ablates and vaporizes high in the atmosphere, leaving metal vapour that condenses into fine particles, mostly aluminium oxide from aluminium structures. How much burns up depends on the object, so Altai uses one stated assumption per kind of object (below): the share demised (burnt up rather than surviving to the ground) and the share of that which is aluminium. Studies of past reentries find 10–40% of a large object's mass reaching the ground; small satellites and Starlinks burn up entirely. Estimates of spacecraft aluminium content range from about 30% to 77%; the low end is used here. All demised and aluminium figures are estimates, good to a factor of two or so, not measurements.

For past reentries with a fitted drag model, the details also show a drag-area check: the fitted ballistic coefficient (mass ÷ drag area) and GCAT's mass give the drag area it implies, which is compared with the object's size from GCAT. A value far outside what its size allows suggests the mass or the fit is off.

Ground tracks are reconstructed, not measured. Altai propagates the object's last published orbit with SGP4 through the reentry time.

• The track's shape is reliable, but the timing along it is not: the orbit model's drag is crude below about 150 km, and the last orbit is often hours old.
• For the same reason the orbit model's altitude stays far too high near the end (often 140–200 km at reentry). So for past reentries with a TIP time, the altitude comes from a separate drag model (gravity plus drag through a standard atmosphere), whose drag is fitted so it comes down through 80 km, the usual reentry altitude, at the TIP time. The track then ends there. Upcoming reentries still show the orbit model's altitude.
• When there's a TIP point, the track is slid along its orbit (by up to 20 minutes) until it passes through that point at the TIP time. That corrects the timing error above; the details show how far it was moved. A track that never comes within 100 km of the TIP point is left as propagated and flagged.
• Without a TIP message, the reentry time is SGP4's own estimate ("orbit model"), or only the day is known ("date only"). Date-only reentries aren't drawn on the overview map; selecting one shows its last tracked orbit (or, if it's still up, its current orbit): where it was, not where it comes down.