A solar eclipse occurs when the moon is positioned between the Earth and the Sun. This can only happen at the New Moon (obviously) and therefore eclipses interested early medieval Christians – not because they were necessarily mysterious or portentous, but because they helped to check the accuracy of the calculations about luni-solar cycles for working out the Easter date. If you need to know when the Easter full moon is each year, you also need to know the day of the new moon, and Easter tables across the Carolingian world carried both bits of information for each year. It is not surprising, then, that noting eclipses was common in the margins of Easter tables.
One particular list of solar eclipses that circulated late in the reign of Charlemagne poses some interesting issues for historians of science. You can compare them to the NASA historical calculations here. Here is the last part of the Carolingian list, translated from Borst’s 2006 edition:
In 807, there was again an eclipse of the Sun on 11 February in the 26th part of Aquarius around the sixth hour of the day.
In 810, there was an eclipse of the Sun on 7 June, around the second hour, in the 24th part of Gemini.
In the same year, there was an eclipse of the Sun on 1 December, in the 24th part of Sagittarius.
In 811, there was an eclipse of the Sun on 27 April, in the first hour of the day.
In 812, there was an eclipse of the Sun on 15 May, in the 28th part of Taurus.
The list survives as part of the famous 7 Book Encyclopaedia produced in the circles of Drogo of Metz (maybe) between 809 and 812. It can be found in early manuscripts from Monte Cassino and Auxerre too.
The biggest problem lies with the eclipse of 811: there was no eclipse on 27 April that year. To what, then, was the author referring?
The simplest explanation is that this is not a simple list of observed eclipses: it is a list of calculated eclipses supplemented by observational data transmitted from further away (ie Byzantium or Ireland). There was, in fact, a solar eclipse in 811, but it occurred one lunar month later on 26 May, and was only visible in or near Australia. (I doubt this involved observational data from elsewhere – that will come later).
It is possible that there is scribal error here, as the difference between v kal mai (27 April) and vi kal jun (26 May) would not be the biggest slip ever in Carolingian manuscript history by a very long way. Alternatively, there was a small calculation error: 27 April could readily be calculated as a new moon (which, again, is what you need to be in conjunction with the Sun), and 26 May was the next one. Such calculations are easy if you have an Easter table and a calendar. Still, the information is not precise enough to guarantee an accurate answer, and this could be chalked up as a valiant fail – still not bad for pen-and-paper astronomy. But does the fact that the real eclipse was only visible in the Southern Hemisphere cause a problem? It is troubling to think that Carolingian thinkers might have been discussing phenomena only observable so far away. However…
One recently discovered source suggests the problem is not as great as you might initially think. This is a text about how to predict eclipses from 754 and it seems to come from an Irish milieu, but it is transmitted in a tenth-century manuscript from Fleury (Paris, BnF, latin 6400B, ff. 277r-v) (see Warntjes 2016). In that text, the author argue that eclipses were recurrent exactly every 30 years, drawing on comments by Isidore of Seville. He therefore expected one in 754, because that was 90 years after the famous total eclipse of 1 May 664 that triggered the Synod of Whitby. But there was no eclipse in 754 visible in Europe. (The calculation was rubbish). The way he explains away his failure, however, is important: it could have been because of the weather, it could have been because it was only a partial eclipses… but it could have been because it occurred in the Southern Hemisphere and so just could not be seen.
And here, note that the position of the eclipse in the sky for 811 is not given in the Carolingian list of eclipses, in contrast to the others in the list, which were visible in the Northern Hemisphere. That might suggest the author of the list has no point of reference for the sky – as would be the case if they had calculated this for the Southern Hemisphere. Doubters might have underestimated the accuracy of early medieval modelling of luni-solar cycles in the sky here.
The other eclipses listed provide support for interest in calculating rather than observing eclipses. The annular eclipse of 807 was reported on the right day, and was visible in the northern parts of Ireland, Pictland, Northumbria, and Norway – and at least partially in the Frankish kingdoms (or else information could have travelled).
The report of one in June 810 is slightly out as it occurred on 5 June, too late in the day to be visible, it was only partial, and was possibly only visible in the Pacific. 7 June, however, was the calculated, not observable, new moon, so you can see the origin of the two-day discrepancy there. The second 810 eclipse – again only partial – was visible, but it occurred a day earlier than predicted, on 30 November, and too far north and east to be visible fully in the Frankish kingdoms – but still observable in part.
Finally, the eclipse in 812, on 14 not 15 May (it was ‘corrected’ again to the calculated new moon they thought so authoritative), was a total eclipse… but its path was visible in North Africa, Sicily, Calabria, and southern Greece, so again only partially in the Frankish kingdoms.
The trigger for this interest was apparently a Byzantine embassy to Charlemagne’s court in 810 that reported some of this activity. This makes sense given where the second visible solar eclipse was actually visible in 810. A second embassy in 812 might explain the information for that year too. Charlemagne asked Dungal of St Denis to explain how two eclipses might occur in the same year and we have Dungal’s reply (see Eastwood 1994). The mechanics are scientifically questionable, sure. But he ultimately defended the idea that eclipses were predictable if one paid sufficient attention to the interrelation between the paths of heavenly bodies.
All we are really talking about is the ability to calculate the course of the moon and Sun in the sky relatively accurately (which they could certainly do), and a capacity to understand the difficulties posed for observation by living on a globe (and they were sure the Earth was a globe). It doesn’t matter that their model of the solar system was geocentric rather than heliocentric, because their modelling and maths worked well enough for their purposes within the model. Here is one effort at charting the course of the moon alone from Fleury (Paris, BnF, NAL 1615):
It is also not a problem if Irish or Carolingian writers mentioned eclipses they could not possibly have witnessed, because they had methods of calculating when the course of the moon and Sun might intersect wherever this might happen. They even had an awareness of why they could not see every eclipse they calculated. We do not yet know exactly how they calculated eclipses in every case. The reference to an eclipse in 811, however, is only a problem if we start by assuming Carolingian astronomy was hopeless (even if it was error-prone), rather than by trying to work out what astronomers were actually up to.
Bruce Eastwood, ‘The astronomy of Macrobius in Carolingian Europe: Dungal’s letter of 811 to Charles the Great’, Early Medieval Europe 3:2 (1994): 117–134.
Immo Warntjes, ‘An Irish eclipse prediction of AD 754: the earliest in the Latin West’, Peritia 24–25 (2013–2014): 108–115.