Starlink drag is now mapping the upper atmosphere's hardest-to-see layer

Kyoto University researchers have used publicly available Starlink orbital data to produce a tomographic reconstruction of thermospheric density near 500 km altitude, according to an initial report in Earth, Planets and Space published on 4 August 2026 and re-covered by ScienceDaily on 12 August.

The thermosphere is the electrically neutral layer of the upper atmosphere, sitting between roughly 100 and 1,000 km up, and it is more than 99% neutral gas. The ionosphere, the thin ionized layer just above it, is the part radio astronomers and ground stations can see, because plasma bends radio signals in measurable ways. The thermosphere, being neutral, doesn't, which is why its density has been the hard part to measure. With low Earth orbit filling up, drag from that invisible layer decides which satellites fall, when, and how accurately ground stations can predict their paths.

The Kyoto team, led by corresponding author Mamoru Yamamoto of the Research Institute for Sustainable Humanosphere, used publicly published ephemeris data for roughly 1,200 Starlink satellites and applied the same inversion mathematics that underlies medical CT scans. In a CT scanner, an image is built from many X-ray projections taken at different angles. Here, each satellite's drag history is a different projection through the same atmospheric volume. By combining the drag histories of many satellites passing through overlapping volumes, the team reconstructed a three-dimensional picture of thermospheric density along their flight paths.

The technique inverts the signal everyone else has been treating as noise. A satellite in low Earth orbit doesn't fly a perfect Keplerian ellipse. The atmosphere is constantly tugging on it, and the small extra decay is usually written off as a disturbance to be averaged over. The Kyoto team treated that disturbance as the data.

The operational payoff is sharper forecasting for both satellites and debris. Collision-risk models use thermospheric density to predict orbital decay over hours and days, and a denser-than-modeled layer can turn a nominal close approach into a real conjunction. The reconstructed density field, if validated, would feed directly into those same models.

"This is a multidisciplinary study between space science and space engineering," Yamamoto said, describing the work as a deliberate bridge between the two camps. The paper, "Tomography of thermospheric density from Starlink Ephemeris: initial report," appears in Earth, Planets and Space (DOI 10.1186/s40623-026-02509-5), with a corresponding record in Kyoto's KURENAI institutional repository, and was carried by a Kyoto University press release and the EurekAlert distribution.

The result is an unintended dividend of commercial megaconstellation growth. Starlink wasn't built as a science instrument, and SpaceX isn't a partner in the study, but the same crowding that raises collision risk in LEO is what makes the drag data rich enough to invert. Each new satellite is another drag recorder the team didn't have to launch, because the LEO congestion that makes traffic harder to manage is what keeps the inversion's input growing.

Yamamoto's group describes the paper as an initial report, and the next step the team flags is extending the reconstruction across more altitude bands and validating the density values against ground-based instruments over a longer window.