Sources · Methodology

How we calculate

Every time and duration on this site comes from one dataset that we compute ourselves. This page explains how, what we check it against, and where it can be wrong.

The eclipse times and durations

We compute the eclipse for each place from the positions of the Sun and the Moon in the JPL DE440 ephemeris, the planetary and lunar ephemeris published by NASA’s Jet Propulsion Laboratory. The calculation runs in Python with the Skyfield library. For each town we take the apparent positions of the Sun and the Moon as seen from that exact point on the WGS-84 Earth ellipsoid, at sea level. Then we search, to a fraction of a second, for the four contacts, the moment of maximum eclipse, the Sun’s altitude and azimuth, and the share of the Sun’s disc that is covered.

The constants behind every number on the site.
EphemerisJPL DE440 (file de440s)
SoftwareSkyfield (Python)
Lunar radiusk = 0.2722810 Earth radii
Solar radius696,000 km
ΔT (TT − UT1)69.1 s
Earth modelWGS-84 ellipsoid, sea level
Lunar limb profileNot included (see below)

Why k = 0.2722810

This is the lunar radius that NASA (Fred Espenak) and most eclipse calculators use for the umbra. It describes the Moon’s mean edge, below its highest mountains. A slightly larger value would make every total eclipse a little shorter and the path a little narrower.

ΔT: the clock of the Earth

The Moon and the Sun move on a uniform time scale. Clocks follow the Earth’s rotation, which is slightly irregular. The difference is called ΔT. We use 69.1 seconds, from the IERS data built into Skyfield. The latest IERS Bulletin A points to about 69.3 seconds for 2 August 2027. The gap moves every time by about 0.2 seconds, less than our rounding. NASA’s eclipse pages for 2027 were prepared years ago with 71.7 seconds. That is why NASA’s clock times are about four to five seconds earlier than ours.

How we checked it

  • We ran the same code on the total eclipses of 2017 and 2024. It reproduces NASA’s published greatest durations to within 0.3 seconds.
  • For 2027 we compared 13 places from Spain to Saudi Arabia with the calculator of Xavier Jubier, the reference used by many eclipse chasers. Durations agree within half a second and contact times within one second.
  • For Spain we also compared with the Instituto Geográfico Nacional. Durations agree within about two seconds. IGN’s clock times are about nine seconds earlier than ours, a constant shift that points to a different ΔT.
Duration of totality in seconds from four sources. Jubier’s second column adds his lunar limb correction. NASA: interactive map with ΔT 71.7 s.
PlaceThis siteJubier Jubier, limbNASAIGN
Tarifa278.5278.5279.0278.7278
Cádiz173.9173.9167.9174.7174
Ceuta288.3288.3289.3288.4288
Luxor380.5380.6379.3380.2
Kom Ombo222.3222.3215.8217.7
El Quseir138.8139.2141.2147.1

The limits of the calculation

  • The Moon is treated as a smooth sphere. Its real edge has mountains and valleys. They move second and third contact by one to three seconds in most places, and by up to six or seven seconds close to the edge of the path. If you plan to stand within a few kilometres of a limit, use a limb-corrected prediction such as Jubier’s.
  • Times are for sea level and for the coordinates shown on each page. A hill or a move of a few kilometres changes the seconds, most of all near the edges.
  • Magnitude is the fraction of the Sun’s diameter covered at maximum. Obscuration is the fraction of its area.

Click anywhere on the interactive map and your browser computes that point with NASA’s Besselian elements and the same ΔT as our dataset. It agrees with the town pages to about a second, and to three or four seconds within a few kilometres of an edge. Town markers on the map show the town page numbers exactly.

The cloud figures

The cloud odds on the weather page and on each town page come from NASA POWER. The underlying data are the CERES SYN1deg satellite product: the daily mean cloud amount in a grid cell of 1° by 1°, about 100 km across.

  • Period: 26 July to 9 August, every year from 2001 to 2024. That is 15 days around the eclipse date over 24 years, 360 days per place.
  • Clear days: the share of those days with a daily mean cloud amount under 20%.
  • Cloudy days: the share with 50% or more.
  • Median: the typical daily cloud amount.
  • The bar chart on each town page shows 2 August of each year.

What these numbers cannot tell you

  • They are daily means, not the cloud at the hour of the eclipse. Morning sea fog that burns off by noon still counts.
  • A 1° cell mixes coast, sea and inland ground. It cannot see a fog bank on one beach.
  • Dust and haze do not count as cloud, but they can dim the corona.
  • This is climatology, not a forecast. For the days before the eclipse, use the national weather service.

Drive times and distances

Drive times in our trip planning tools come from OSRM, a routing engine that runs on OpenStreetMap road data. They are estimates for normal traffic. On eclipse morning the roads near the path will be much busier, so allow far more time. Distances to the central line and to the path limits on the town pages are straight lines, measured to our own DE440 path.

Official and reference sources

The eclipse

Weather

Roads

Found a number that looks wrong?

Please tell us the place, the number you expected and where it came from. We check every report against the sources above.

Last updated: · Sources and methodology