Scope and coverage
- The question is how eclipse computational products are computed well enough to build one, from the list of eclipses in a period, through world maps of the path, to the second at which totality begins for one observer standing on one hillside.
- The reference standard is NASA SVS's 2024 work, published as Wright and Young 2024: a lunar limb from LRO and SELENE topography, terrain under every observer, and a JPL ephemeris, traced through to the shape of the umbra on the ground.
- All eleven topics are covered. Each rests on primary documents and public code, is recorded in raw notes with graded citations, and is carried into the reports.
- The gaps are named artefacts: the SVS code, Occult's help file, Meeus's Elements, the 2013 Supplement's eclipse chapter, and the unpublished 2024 edge reductions.
The question
Eclipse predictions are published by a handful of people and programs: Fred Espenak's NASA and EclipseWise tables, Xavier Jubier's interactive maps, Dave Herald's Occult, Ernie Wright's NASA SVS visualisations, and commercial sites such as timeanddate. They agree to within a few hundred metres at the path edge and a few seconds at a site, and the places where they disagree are the places where the interesting modelling decisions live: the radius of the Sun, the mountains on the Moon's limb, the height of the observer, and the unpredictable rotation of the Earth.
The subject is what a developer needs to know to compute every one of these products, and to know how good the result is, organised by product from coarse to fine.
The eleven topics
| Product | Question | Topic | Covered |
|---|---|---|---|
| Besselian elements | What is the fundamental plane and how are the elements derived from an ephemeris? | Foundations | Yes; formulas quoted from the 1961 and 1992 Supplements |
| The list of eclipses in a period | How are eclipses found and classified? | Catalogues and types | Yes; Meeus's method tested against 221 catalogued eclipses |
| World maps | How are global circumstances computed from the elements? | Global circumstances | Yes; every classic curve with its formula |
| Site tables | How are local circumstances computed? | Local circumstances | Yes; checked against NASA's GPL JavaScript |
| Limb-corrected contacts, beads, true umbra shape | How does the lunar limb profile enter? | Lunar limb profile | Yes; the DEM-to-profile algorithm and datasets |
| Path width and duration at the edge | Which solar radius, and what does the choice cost? | Solar radius | Yes; every value traced to its measurement |
| Every product | Which Earth figure, which time scale, how much ΔT uncertainty? | Earth model and time scales | Yes; effects table with sizes |
| Every product | Which ephemeris, which lunar figure, which libration model? | Lunar and solar model | Yes; kernel names and frames |
| Any product | What software and repositories exist, and how complete are they? | Software and repositories | Yes; thirty repositories inventoried |
| Any product | How well do predictions match observation? | Validation and error budget | Yes; the budget assembled and a five-level protocol |
| The reference standard | What exactly did SVS do for 2017, 2023 and 2024? | NASA SVS and Ernie Wright | Yes; the paper in full, with the datasets inventoried |
The standard of detail
Each raw note is written for someone who will implement from it. That sets the bar: formulas quoted rather than described, constants named with their provenance, datasets named with their product identifiers and download locations, and code cited by file when the program is public.
What is quoted rather than reproduced
- Only the enumeration is independently reproduced. Meeus's series and Kluepfel's Saros formula are implemented from their quoted constants and checked against the NASA catalogue. No Besselian, limb or terrain result is reproduced independently; each such number is quoted from its source. The reading list names the reference cases.
- Two standard texts are cited only at second hand: Meeus's Elements of Solar Eclipses 1951-2200 and chapter 11 of the 2013 Explanatory Supplement, both through implementations that transcribe them.
- Three sources are unavailable: Occult's help file, distributed only inside the Windows installer; Solar Eclipse Maestro's statement of its ephemeris and ΔT, which its help pages omit; and timeanddate's ephemeris, and ΔT source, which its accuracy page does not name.
- NASA SVS pages are cited from Wayback Machine snapshots, the host being unreachable. Snapshot dates are in the reference notes.
What is deliberately out of scope
- Lunar eclipses, transits and occultations, except where the same machinery is reused and the source says so.
- Eclipse weather, photography, safety and travel.
- Historical eclipse identification as a discipline, except where it validates a ΔT model.
What would change the conclusions
Ranked in open questions. The first three: the SVS source code or a statement of its 2024 constants; a published reduction of the 2024 edge observations, which would fix the default solar radius; and a measured SVS-against-Irwin limit offset.