Abstract
Abstract
We demonstrate that a single photograph of a total solar eclipse, together with its capture timestamp, suffices to geolocate the photographer to a few tens of kilometers. The method matches the observed pattern of Baily's beads and the bright emergence arc against a forward model that combines lunar limb topography (LOLA LDEM_128), the instantaneous lunar orientation (DE421 librations), and the Moon-Sun contact geometry seen from each candidate site.
On all 6,973 measurable, unselected crowdsourced photographs of the 2017 August 21 eclipse from the Eclipse Megamovie archive - unknown camera roll, consumer EXIF clocks - the pipeline attains a median error of 66 km (90th percentile 103 km; photographer-blocked bootstrap 95\% confidence interval on the median [64, 68] km), with not one frame in error by more than 1000 km; collapsing to the 285 distinct observing sites gives the same accuracy (median 66 km). A single well-exposed bead-phase frame from the 2024 April 8 eclipse, searched over the eclipse's entire continental track (Mexico to Newfoundland, 2,453 candidate sites) with no positional prior, inverts to 16.8~km, with the arc-width channel confining the observer to a thin along-track segment of the path. We further show, by direct computation, that the lunar limb-profile shape carries no usable position information at photographic precision (< 0.01’’ rms change per 100~km of observer displacement, demeaned), while the Moon-Sun offset changes by ~54’’ per 100~km: all single-image position signal resides in the contact geometry, with the topography determining where along the contact arc the beads appear.
On a 190-frame stratified sample from the independent geometry of the 2024 April 8 eclipse (Eclipse Megamovie 2024 release; no disk-fit priors, 100 m GPS truth) the same pipeline, searched over the full Mexico-to- Newfoundland track, achieves a median of 125 km (p90 202 km, no failure beyond 1000 km), confirming generality across eclipses.
Probes at the true sites place the method's local floor at 23 km; the gap to the realized 66 km is the shallow along-track score valley of roll-free width matching, indicating a clear path below 30 km through multi-frame sequences or roll-resolved matching. Applications include GPS-denied positioning, forensic verification of historical eclipse imagery, and a quantitative case for the scientific value of citizen eclipse photography.