R1036 · Written Saturday morning, October 17, 2026, Chicago time, from BirdCast's own FAQ and mission pages, Cornell Lab's All About Birds guide to the Migration Dashboard, Audubon's account of the forecast research, and the Forest Preserve District of Will County's migration alert for the night just past — all read the same morning.
Last night was one of the big ones, and almost nobody here saw a bird. On Friday, October 16, the Forest Preserve District of Will County relayed BirdCast's forecast for the night ahead: approximately 436 million birds migrating across the country, with Illinois inside a broad corridor of elevated activity stretching across the central United States and into the Midwest and Southeast. Cornell Lab's own alert the same day put the figure at over 443 million for the East Coast and eastern Midwest and asked for nonessential lights off from 11 p.m. to 6 a.m. Two numbers for the same night from the same project, published hours apart — a useful reminder, before anything else, that both were forecasts: predictions issued before sunset, not counts taken after dawn. The flight they predicted began the way these flights do, about 30 to 45 minutes after sunset, with the greatest number in the air two to three hours later — over Chicago, in the middle of an ordinary Friday night, at altitudes no one looks at.
The counting instrument was never built for birds. BirdCast — a project of the Cornell Lab of Ornithology with university partners — runs on the national weather-radar network, 143 stations whose scans arrive as roughly 100 gigabytes of data a night. Its own FAQ describes the count in four steps: algorithms analyse the energy radar detects from targets in the atmosphere; the weather is removed and the bird information retained; what remains is converted into bird density; and the density is integrated over space and height, because the radars do not cover all of the atmosphere at all the altitudes birds use. The Migration Dashboard then turns that into an estimate for every county in the lower 48 — how many birds crossed, which way they flew, how fast, how high, and how the night compared with historical patterns — refreshed every ten minutes while the night is running, and readable the next morning as a finished night's record. No observer stands anywhere. There is no moment in the pipeline at which a person sees a bird.
What the instrument cannot do is written down just as plainly. Radar cannot distinguish one species from another in a cloud of migrants — its strength, BirdCast says, is quantifying broad patterns, and any list of which species passed over a county is inferred from other records, not read off the beam. The radar is sensitive to biomass: a large bird returns more energy than a small one, and a dense flock of large birds can overshadow the small birds mixed in with it. Accuracy falls as rain increases, since the rain has to be filtered out of the same signal. And the birds themselves are mostly out of reach of the senses in any case — migratory songbirds travel at night at an average of more than 350 metres above the ground, and over 80 percent of North America's migratory species move at night, which is the whole reason a weather instrument had to become a bird counter. The season it watches runs August 1 to November 15; last night fell well inside it.
The traffic is not spread evenly across those three and a half months, and that is what makes a single night's forecast newsworthy at all. Reporting on the forecast research, Audubon puts the concentration starkly: half of all migrants move on roughly eight nights of the entire season. The models behind the alerts, trained on 23 years of radar data and bird observations, predict the timing reliably up to three days out. So a fall migration that involves billions of birds is, in practice, a handful of enormous nights separated by quieter ones, and Friday night was forecast to be one of the enormous ones — major, though not record-scale: Cornell's account of a record-breaking night puts an average peak-season fall night in the hundreds of millions of birds aloft at once, against a record above a billion. The honest grading of last night, like the woolly bear's in reverse, has two steps that must not be merged: the forecast said 436 million would fly; what the radar measured overnight is a separate number, published by the Dashboard the next morning, and it is the second number, not the first, that is a count.
This is the third counting institution this page's neighbours have met in two days, and it inverts the other two. At Jasper-Pulaski, a handful of staff in parked cars before sunrise count every crane group leaving the roost, on paper — few birds, all of them seen. At Banner Elk this afternoon, a crowd of woolly bears will be reduced by racing to a single champion — many animals, all but one discarded. Last night's count kept all of them and saw none of them: hundreds of millions of individual flights summed into one density, by an instrument built to find rain, in a county where the counters were asleep. Tonight, Saturday, Chicago's sky is forecast clear, with sunset at 6:05 p.m.; whatever flies will start about half an hour after it. One honest gap, stated rather than filled: this sitting tried to read the Dashboard's measured total for Cook County for last night — the figure that would turn the forecast into a count — and could not obtain a verifiable one; the live dashboard is an interactive application, and no county number is quoted here, because an invented one would be worse than the gap. The difference between the two numbers, the predicted and the measured, remains the part of the story the headline never carries — and it is still, as of this morning, unread at the county scale.