The moon that may not have been visible

This is the single most-cited piece of physical evidence for dating the crucifixion to 3 April AD 33 — and the single most-contested. The argument is genuinely clever. The objections to it are genuinely serious. Most popular retellings give you the first and not the second, which is why this page exists.

The claim

In 1983, Colin Humphreys and Graeme Waddington published a short paper in Nature arguing that the crucifixion can be dated astronomically. Their calendrical work gives two candidate Fridays — 7 April AD 30 and 3 April AD 33 — which is the same result this site's engine reproduces from first principles.

To break the tie, they reached for something else. On the evening of 3 April AD 33, a partial lunar eclipse was in progress as the moon rose over Jerusalem, around 6:20 pm local time. A moon rising already in the earth's shadow appears reddened. And in Acts 2:20, Peter quotes Joel: the moon turned to blood before the day of the Lord. Humphreys and Waddington proposed these are the same event — that Peter, weeks later, was appealing to something his audience had actually seen.

It was, they noted, the only lunar eclipse visible from Jerusalem at Passover across the whole of Pilate's prefecture. If the reference is to a real eclipse, only one year fits.

The objection: it would have been washed out

Seven years later, the astronomer Bradley Schaefer — a specialist in exactly this problem, the visibility of astronomical phenomena to the naked eye — published a detailed rebuttal in the Quarterly Journal of the Royal Astronomical Society.

His objection is not about whether the eclipse happened. It did. It is about whether any human being in Jerusalem could plausibly have noticed it. Two things work against it:

• It was never deep. At most about 59% of the disc was in shadow, and by moonrise the eclipse was already ending. A partial eclipse leaves the rest of the disc in full brilliance.
• Light leakage swamps the effect. The unshadowed portion is overwhelmingly brighter than the shadowed part. Against a bright twilight sky, low on a dusty horizon, the reddening that makes a "blood moon" striking simply does not register.

Schaefer's conclusion was that the event would have been, at best, barely noticeable — not the kind of sight that lodges in a city's memory and can be invoked from a podium weeks later.

Humphreys and Waddington did not concede. Their 1992 Tyndale Bulletin treatment defends the visibility of the eclipse's final phase against precisely this line of attack. The exchange has never been cleanly resolved — which is itself worth knowing, because popular accounts tend to report the 1983 paper as though the following decade never happened.

A newer objection: the weather

In 2022 the astronomer Guillermo Gonzalez revisited the calculation in Bibliotheca Sacra. His finding is more interesting than a simple yes or no, and it cuts in both directions.

On the pure astronomy, Gonzalez is somewhat kinder to Humphreys and Waddington than Schaefer was: under ordinary atmospheric conditions, he concludes, the closing phase of the eclipse would indeed have been visible. But he then argues that conditions that evening were probably not ordinary. The gospel accounts of darkness are consistent with a khamsin — the hot, dust-laden wind that periodically buries the region in haze. Under a khamsin, a low moon at twilight is not merely unimpressive; it is invisible.

His assessment is blunt: this removes one of the principal arguments for the AD 33 date.

The objection nobody needs a telescope for

Set aside the physics entirely and a second problem remains, which is about how texts work rather than how light does.

Acts 2:20 is a quotation. Peter is citing Joel 2:31, a piece of prophetic literature written centuries earlier, in which cosmic disturbance signals the day of the Lord. Darkened sun and bloodied moon are a standing idiom of apocalyptic writing, deployed repeatedly across the Hebrew Bible and Second Temple literature without anyone supposing a specific eclipse lies behind each instance.

So the inference has a gap in it. Even granting that a visible eclipse occurred, one still has to establish that a speaker quoting a stock prophetic image intended a report of last month's weather. That step is interpretive, not astronomical, and no ephemeris can settle it.

Why this site shows you both

The calendrical engine on this site does not need the eclipse and does not use it. It reconstructs Nisan from lunar phase and hands you the weekdays. The eclipse enters only when someone wants to choose between AD 30 and AD 33 — and at exactly that point, the evidence gets weaker, not stronger.

That is the honest shape of the problem: the astronomy is the solid part, and it gives you two answers. The tie-breaker is the disputed part. A tool that hid the dispute in order to hand you one date would be lying about which half of its own reasoning is load-bearing.

Next: why even the two-candidate result rests on a contested cycle →