Swift Re-entry Challenge
NASA’s Neil Gehrels Swift Observatory was expected to re-enter the atmosphere on 4 November 2026; the time it did has not yet been confirmed. For the NASA Swift re-entry challenge, SWORD Track 1 predicted the re-entry every day from 5 October, from its own 1 to 30 day forecasts of the F10.7 solar radio flux and the Ap geomagnetic index. This page is the record of those predictions and is no longer updated. The F10.7 and Ap forecasts continue.
Each daily update started from Swift’s latest orbit from Space-Track and stepped it forward with NASA’s GMAT orbit software and the NRLMSISE-00 model of air density until it reached 80 km, the altitude counted here as re-entry. How strongly drag acts on Swift, its ballistic coefficient, was fitted in each update from the recent orbits of Swift and about 80 other satellites. Future F10.7 and Ap came from the SWORD Track 1 forecasts, and for comparison from every pair of the other groups’ forecasts: NOAA SWPC’s 27-day outlook, NJIT, ADAPT-SIFT, SWORD Track 3, ESA, the Schatten forecast, and the last observed day held flat.
Each forecast against what then happened, up to re-entry; before 1 October 2026 the SWORD Track 1 values are hindcasts. SWPC issues its 27-day outlook once a week, so the tables score both on those days only. Swift’s measured altitude is a one-orbit average from each new Space-Track orbit. The runs of 5 to 7 October were recalculated on 7 October with a corrected drag value.
For the last day the daily chain was replaced by a run built for it. The drag value was fitted to Swift’s own orbits over the previous week, in stretches of about 30 hours, and its trend carried forward. From Swift’s orbit of 3 November at 21:22 UTC, at 183 km, 300 GMAT runs combined 25 past SWORD Track 1 forecast error paths with 12 drag values. The middle run reached 80 km at 11:36 UTC on 4 November, with a 90% range of 10:20 to 12:59 UTC, less than two orbits wide.
The drag value. The ballistic coefficient fitted from about 80 reference satellites, most of them near 400 km, carried the density model’s error at those heights, not at Swift’s. Between 200 and 250 km the model ran 20 to 30% too dense in the final weeks, so from about 16 October every daily update came out early, and the gap grew as Swift fell. A drag value fitted to the satellite’s own recent decay, as in the final run, removes this.
F10.7 weeks ahead. In early October the SWORD Track 1 forecast put F10.7 at 115 to 120 for late October; it was about 90. The runs with NOAA SWPC’s 45-day forecast tracked the measured altitude at 27 days ahead, the SWORD Track 1 runs fell well below it. Weeks out, the level of F10.7 matters more than anything else in the calculation.
The ranges. The published 50% and 90% ranges came from the errors of past F10.7 and Ap forecasts around one drag value, so they were too narrow: the measured altitude sat above the whole range for much of the final two weeks. The spread in the drag fit itself belongs in them.

