The author seems to treat the atmosphere as a, well, sphere. It isnt. It is not of equal depth at all areas. As starlink (and every other sat) only travels in part of the atmosphere each day they cannot measure the entire atmosphere. So you measure a thousand starlinks to get an average. Ok, but a significant bulge on one side of the planet will slow all sats, giving the false impression of a homegenous trend. So you have to track very specific sats in very specific survey orbits... but also update that list as the orbital shell rotates around the earth. It is a very complex problem.
Unfortunately the data doesn't have the required time or space resolution. The drag term is averaged over many orbits, and you can't even use perigee/apogee drift to localize it since the atmosphere rotates so it gets averaged over all longitudes.
You can do some rudimentary localization by latitude, however, by comparing satellites in different inclinations. You also get data by altitude, of course.
That's cool.
Reminds me that scientists can use GPS to track earthquakes indirectly as it interacts with the ionosphere: https://link.springer.com/article/10.1186/s40623-025-02211-y
Link to article: https://www.spaceweather.com/starlink/starlink_drag_explaine...
The author seems to treat the atmosphere as a, well, sphere. It isnt. It is not of equal depth at all areas. As starlink (and every other sat) only travels in part of the atmosphere each day they cannot measure the entire atmosphere. So you measure a thousand starlinks to get an average. Ok, but a significant bulge on one side of the planet will slow all sats, giving the false impression of a homegenous trend. So you have to track very specific sats in very specific survey orbits... but also update that list as the orbital shell rotates around the earth. It is a very complex problem.
I suspect the author understands the concept of a spheroid, though they don't mention the term as such.
There is a fair bit of statistical rigour described in the original article at : https://www.spaceweather.com/starlink/starlink_drag_explaine...
Unfortunately the data doesn't have the required time or space resolution. The drag term is averaged over many orbits, and you can't even use perigee/apogee drift to localize it since the atmosphere rotates so it gets averaged over all longitudes.
You can do some rudimentary localization by latitude, however, by comparing satellites in different inclinations. You also get data by altitude, of course.