Detecting Solar Eclipse Effects on Neutral Density Using Starlink Ephemerides

This study examines whether the effects of the total solar eclipse of 12 August 2026 can be detected in publicly available orbital data for Starlink satellites. Two sources were used: Two-Line Element (TLE) sets, used to build the initial list of satellites expected to cross the shadow, and the ephemeris files published by SpaceX, which forecast each satellite's position at one-minute intervals. Because SpaceX does not archive past ephemerides, these files were downloaded before, during, and after the eclipse. Reduced solar heating, whether at night or within an eclipse shadow, causes the upper atmosphere to contract. This should lower neutral density, and therefore aerodynamic drag, for satellites such as Starlink that orbit within this region, the thermosphere. The question addressed here is whether lower-precision public data, aggregated across an entire constellation, can resolve this effect. Satellites were first classified by the region of the shadow they crossed (umbra, penumbra, or none); this classification was then replaced by a single continuous value per satellite, the shadow exposure expressed in equivalent umbra-seconds and treated as a shadow dose. The position of each satellite in an ephemeris forecast issued before the eclipse was compared with its position in a forecast issued after the eclipse. The later forecast incorporated more recent tracking and would therefore reflect any effect of the eclipse. Before analysis, satellites at low altitude, probable thruster firings, and outliers in the forecast gap relative to the rest of the fleet were removed. The 701 satellites that passed through the shadow ended about 80 m further back along their path than the 161 that did not. Because of the drag paradox, a position further back indicates less drag, not more. A Mann-Whitney test gave an effect size of A = 0.746 with p ≤ 0.0002. In a placebo check, the same test applied to the day before and the day after the eclipse showed a difference only on the eclipse day. These results are consistent with Starlink satellites encountering lower neutral density as they crossed the eclipse shadow.