Records of solar eclipses from a millennium and a half ago have allowed scientists to refine measurements of Earth's changing rotation.

A painstaking review of historical documents from the Byzantine Empire has given scientists timings and locations for five solar eclipses. The results, although consistent with previous findings, place new, tighter constraints on Earth's variable spin rate, giving us a better understanding of how our planet is changing over time.

The length of a day seems like a pretty reliable, unchanging metric. Twenty-four hours in a day: 86,400 seconds. That's what all our clocks count out, day after day after day. That's the beat to which we live our lives. But it's a bit of an illusion.

The rate at which our planet turns slows and accelerates in patterns influenced by a variety of factors both underfoot and overhead.

Consider the long-term trend in which our days are gradually stretching ever longer. Based on the fossil record, scientists have deduced that days were just 18 hours long 1.4 billion years ago, and half an hour shorter than they are today 70 million years ago. We seem to be gaining 1.8 milliseconds a century.

Then there's the strange six-year oscillations: scientists have figured out that Earth's days undergo time variations of plus or minus 0.2 seconds every six years or so.

A wobble in Earth's rotational axis seems to be able to produce anomalies, like a peculiarly short day recorded last year. Just for something different.

From core activity, to atmospheric drag, to the expanding orbit of the Moon, a number of factors can influence the actual length of Earth's days.

The discrepancy between the accepted length of a day which we all set our watches to (Universal Time, or UT) and a standardized metric precisely counted out by atomic clocks (Terrestrial Time, or TT) – the most accurate timekeeping devices we have – is a measurement known as ΔT (delta-T).

ΔT becomes really important when it comes to solar eclipses. That's because the positions of the Sun and the Moon are calculated and predicted using TT, but the Moon's shadow will be falling on a planet operating under UT. So you need to know the difference between the two times in order to predict from where on Earth the eclipse will be visible.

But, it also works in reverse! If you have the precise time and location of a solar eclipse, you can work out ΔT. Scientists have been able to work out ΔT from historical records from China, Europe and the Middle East.

Three scientists, Hisashi Hayakawa of Nagoya University, Koji Murata of the University of Tsukuba, and Mitsuru Sôma of the National Astronomical Observatory of Japan, have now pored through historical documents from and of the Byzantine Empire to do the same thing.

This is to fill in a significant gap: from the fourth to the seventh centuries CE, there is a scarcity of solar eclipse records. It's fiddly work. Often details that are pertinent to modern studies have not been included in the records, for instance. But the researchers were able to pinpoint five solar eclipses from records that hadn't previously been analyzed.

"Although original eyewitness accounts from this period have mostly been lost, quotations, translations, etc., recorded by later generations provide valuable information," Murata says.

"In addition to reliable location and timing information, we needed confirmation of eclipse totality: daytime darkness to the extent that stars appeared in the sky. We were able to identify the probable times and locations of five total solar eclipses from the 4th to 7th centuries in the Eastern Mediterranean region, in 346, 418, 484, 601, and 693 CE."

Largely, the values for ΔT that the team was able to derive from these results were consistent with previous estimates.

However, there were some surprises. From the account of the eclipse that took place on July 19, 418 CE, the researchers identified the site of observation for the eclipse totality as Constantinople.

The author, historian Philostorgius, describes the eclipse: "When Theodosius [Emperor Theodosius II] had reached adolescence, on the nineteenth of July at about the eighth hour, the Sun was so completely eclipsed that stars appeared."

Philostorgius lived in Constantinople from around 394 CE until his death, in around 439 CE. It is therefore most likely that he viewed the solar eclipse from there. The previous model for ΔT for this time would have placed Constantinople outside the path of eclipse totality – so the record has allowed the team to adjust ΔT for this time.

The other records show slight adjustments too.

"Our new ΔTdata fill a considerable gap and indicate that the ΔTmargin for the 5th century should be revised upward, whereas those for the 6th and 7th centuries should be revised downward," Murata says.

Although the tweaks may seem slight, they have considerable implications. They place tighter constraints on the variability of Earth's rotation on century timescales, and may inform future studies of other geophysical phenomena, such as modeling the planetary interior, and long-term sea level changes.

The research has been published in Publications of the Astronomical Society of the Pacific.