EN ▾
Why two sources give different eclipse times
Compare any two eclipse predictions far enough ahead and they disagree by seconds, and by kilometres on the ground. Almost none of that is error — it is how fast the Earth will be turning on the day.
Take an eclipse thirty years out and compare the times two reputable sources give for it. They will differ — by a second or two, sometimes more — and the ground track will differ with them, by a kilometre or so. Neither is wrong. The disagreement is almost entirely one quantity, and it is not knowable in advance.
Two clocks, and only one of them is steady
Astronomical positions are computed on a uniform time scale, one that ticks at a constant rate. The Earth's rotation is not uniform: it is slowed by tidal friction over centuries and wanders unpredictably from year to year as mass moves between the core, the oceans and the atmosphere.
ΔT is the difference between the two — between uniform time and the time the Earth's rotation actually keeps. Today it is about 70 seconds. In 1900 it was close to zero. Two thousand years ago it was around three hours.
The geometry of an eclipse — where the Moon is relative to the Sun, and how long totality lasts — is known to a small fraction of a second. Where that geometry lands on the ground depends on how far the Earth has turned, and that is where ΔT enters.
Why a second matters on the ground
The Earth turns 15° of longitude an hour, which at the equator is about 465 metres per second of clock time. So a one-second disagreement about ΔT moves the whole path of totality nearly half a kilometre east or west.
That is small compared to a 200-kilometre-wide path and irrelevant to any traveller. It is not irrelevant to a published table, which is why two of them can look like they disagree when they are simply using different ΔT models.
What is knowable and what is not
ΔT is measured, not predicted. Up to the present it is known precisely, from decades of very long baseline interferometry and satellite laser ranging. Beyond the present it can only be extrapolated, and the extrapolation diverges:
| How far ahead | Uncertainty in ΔT | On the ground |
|---|---|---|
| Today | under 0.1 s | tens of metres |
| 10 years | ~1 s | ~0.5 km |
| 50 years | several seconds | a few km |
| 500 years | minutes | hundreds of km |
None of that touches the date, the duration, or whether a place is inside the path — those are geometry. It touches the clock time, and only by seconds within any horizon a reader is planning around.
What this site does
Times here are computed from observed ΔT where it exists and from the standard polynomial extrapolation beyond it, and the observed table is refreshed against the published source. The site's own accuracy note says the times are good to about a minute and the path to a couple of hundred kilometres, which is a deliberately conservative statement covering the truncation of the series as well as ΔT.
For an observing campaign, use the published canon and its own ΔT. For deciding where to stand and when to be outside, the difference has never mattered.
Related
Updated 25 August 2026.