From 1879 to 1887, keepers at about two hundred lighthouses and lightships around Britain and Ireland filled in schedules of the birds they saw for a committee of the British Association, which printed their entries by name. Here is every night of those reports. Pick a night on the calendar to see it on the chart: every light at its own rhythm, the keepers' words at the hours they printed, the birds counted dead where they fell.
Each row is one lunation, from 1879 at the top to 1887 at the bottom; new moon at both edges, full moon in the middle. A cell's brightness is that night's moonlight at midnight. Each white stroke is one station whose keeper recorded birds striking the lantern or killed.
Every light on that year's lists, flashing its printed character in real seconds while the night runs at 7½ minutes a second. Tallies in fives mark the birds counted dead at a light (a small ring where a keeper wrote "many killed" without a number). Short arrows show the wind as a keeper printed it (lamp colour) and as the 20th Century Reanalysis reconstructs it (blue), each pointing the way the wind blew. Choose a light to see its station.
Treated as instruments, the keepers can be tested: against a modern reconstruction of the 1880s weather, against the moon's known positions, and against each other. Every test below had its check and its null written down before it was run; the details, including what failed, are in the notes.
The keepers' printed winds fall within 45° of the 20th Century Reanalysis's wind at the same place and time 63% of the time, against 25% when each is paired with the same station's wind on another day. They measured the weather.
The light-vessel crews' winds are turned about 17° from true north, against a magnetic declination of about 20° then: they read the ship's compass. The lighthouses are a mixture, some true, some magnetic, a few far off.
Counted over every night of each station-month with a strike, birds struck the lantern on 3.1% of nights under a bright moon (more than half lit and up at midnight) and 7.9% of the rest. Within station and month, strike nights are half as often moonlit (odds ratio 0.49).
When a station's keeper changed, the amount it reported changed more than when the keeper stayed, beyond season and coast (log ratio +0.29, 95% 0.11–0.47). A station's record is partly a portrait of its keeper. Whether the birds killed follow the keeper is not settled.
The largest night in the record, 21 October 1887: 507 birds recorded killed at 16 stations, 272 of them at the South Bishop off Pembrokeshire between 11 P.M. and 2.30 A.M.
On the one coast with enough timed records, a rush of birds does not pass from light to light at any speed a bird flies. The great nights come broadside.