Methodology and sources
How we compute Moon phases and the eclipse, which sources we use and how accurate they are.
What we calculate
For every city we compute the Moon phase, illuminated fraction, Moon age, distance from Earth, rise, set and transit times, the exact moments of the principal phases, and supermoons. Everything is computed ahead of time and published as HTML: pages show correct data even without JavaScript.
Engine
We use astronomy-engine 2.1.19 (MIT licence), an open-source ephemeris library. The difference between Terrestrial Time and Universal Time (ΔT) follows the Espenak–Meeus model (≈ 76 s in 2027).
How we verify
Principal phases: we compare all 149 phases of 2026–2028 with Fred Espenak’s table (NASA GSFC). All agree within ±1 minute.
Rise, set and transit: we compare 132 days (11 cities from Madrid and Istanbul to Buenos Aires and Tromsø) with the US Naval Observatory (USNO). All times agree within ±2 minutes, including days when the Moon does not rise, does not set, or stays above or below the horizon all day.
Illumination: at most 1 percentage point from USNO. These checks run automatically every time the site is published.
Conventions
Rise and set refer to the upper limb of the Moon, with standard atmospheric refraction, at sea level — the same convention as USNO. Mountains, buildings or elevation can shift what you see by a few minutes.
Times are shown in each city’s time zone (IANA database) with its abbreviation, and in UTC on a second line. Daily illumination, age and distance are computed at local noon.
A day gets the name of a principal phase (new moon, first quarter, full moon, last quarter) only if the exact moment of that phase falls on that day in the city’s local time. That is why the same full moon can fall on different dates in Madrid and in Mexico City.
Supermoons and full moon names
Supermoon: a full moon whose distance from Earth is within the closest 10% of its orbit, i.e. distance ≤ perigee + 10% × (apogee − perigee), using the nearest perigee and apogee. This is Richard Nolle’s definition, the one most used by the media.
Traditional names (Wolf Moon, Harvest Moon…) come from the northern-hemisphere folk calendar. The Harvest Moon is the full moon closest to the September equinox; the next one is the Hunter’s Moon.
City data
Coordinates, population, time zones and city names come from GeoNames (CC BY 4.0), reviewed by hand. The orientation of the Moon disc accounts for the hemisphere: in the south the Moon is drawn rotated.
The 2 August 2027 eclipse
We compute the eclipse from Fred Espenak’s Besselian elements (NASA GSFC) using the standard method of the Explanatory Supplement to the Astronomical Almanac: times of the four contacts and maximum, fraction of the Sun covered, totality duration, Sun altitude and path limits. astronomy-engine serves as a second, independent check.
We verify against the tables of Spain’s National Geographic Institute (IGN) for 8124 municipalities: the median contact difference is 0.4 s and 99.8% are within ±10 s; the few exceptions come from wrong municipal coordinates, not from the calculation. Against NASA, the instant of greatest eclipse agrees within 0.1 s and the central duration (6 min 22.6 s) within 0.1 s. The full report is published with every version of the site.
ΔT (the difference between Terrestrial and Universal Time): we use 69.2 s, the current value. IGN uses an older forecast of about 76.5 s, so our times are about 9 s later than theirs; totality durations barely change. We will review ΔT before the eclipse.
Near the edge of the path, duration changes a lot over a few hundred metres and depends on the profile of the lunar limb, which we do not model. So within 3 km of the edge we show a range and a warning instead of a single number, and in Spain we never show a longer totality than IGN does.