Starlink Collision-Avoidance Files Yield First Real-Data Map of Thermospheric Density at 482 Kilometres
A researcher has turned one week of SpaceX's publicly posted Starlink orbit files into a map of atmospheric density at 482 kilometres altitude, drawn by latitude and longitude. Mamoru Yamamoto of Kyoto University's Research Institute for Sustainable Humanosphere believes it is the first time the technique, known as satellite orbit tomography, has been run on real data rather than synthetic numbers. The work was published as an Express Letter in Earth, Planets and Space.
Every eight hours, SpaceX refreshes a set of files predicting where each Starlink satellite will be for the next 72 hours, minute by minute, with velocity and a covariance matrix. The files exist for collision avoidance, keeping the constellation clear of other satellites and spent rocket stages, and reach the public through the Space-Track website. SpaceX has mirrored them on its own server since July 2025. An earlier paper had proposed reconstructing atmospheric density from the collective slowing of many satellites but had verified the method only on synthetic data.
The measurement is made in energy. A satellite in a bound orbit holds a fixed specific mechanical energy, and for the satellite worked through in detail in the paper that figure averaged minus 29.1 megajoules per kilogram. Drag subtracts from it and thruster burns add back. Each satellite is propagated forward with GMAT, NASA's open-source mission analysis software, once under a no-drag condition, and that frictionless twin is compared against the ephemeris prediction. Station-keeping burns appear as sudden steps and are removed as discrete jumps. In the first dataset, manoeuvres were detected in 80 percent of the data.
The analysis used data from 1 to 7 September 2025, drawn from a single Starlink shell of about 1,200 satellites at 482 kilometres and 53 degrees inclination, producing 19 separate reconstructions. The number of satellites feeding each ranged from 850 to 1,100. The map is a spherical harmonic expansion truncated at degree one, with density assumed to fall exponentially at a fixed scale height of 60 kilometres. Pushing the expansion past degree one fails, throwing up false peaks at high latitudes, because a shell inclined at 53 degrees leaves the fit with no data above that latitude.
The European Space Agency's SWARM satellites, which derive density along their own tracks and publish openly, supplied an independent check. Individual reconstructed density values ran from 0.5 to 1.5 times the SWARM numbers, and the average across all 19 case means was 0.95. Near the peaks of variation, the reconstruction, SWARM and the existing NRLMSIS 2.1 model all agreed. Near the minima, the reconstruction tracked SWARM more closely than the model did.
Yamamoto describes the results as preliminary and low-resolution, and states the claim is procedural: a method that works on public data refreshed every eight hours. He notes that SpaceX's tracking system likely serves as a filter, blending real observations with model predictions, so biases from the filtering could carry into the density estimate, particularly when density is changing sharply. Density sets how fast dead satellites and debris come down, and how confidently anyone can predict where a spacecraft will be, which is what collision avoidance rests on.
The condition that makes those numbers most valuable is the one the reconstruction has least to say about. Density swings hardest during geomagnetic storms, and that is precisely when the filtering behind the input data most needs watching. The paper reports no test of the method through a storm.