Most migratory birds fly at night. Weather surveillance radar stations — designed to detect rain — inadvertently detect birds. The biological signal was noise to meteorologists but data to ecologists: algorithms that separate bird returns from precipitation returns now quantify the nightly passage of billions of birds across continental flyways.
BirdCast forecasts use multi-scale terrestrial and atmospheric predictors to achieve real-time, high-resolution migration forecasts explaining up to 65% of national variation in nocturnal migration intensity. The models integrate wind speed, temperature, barometric pressure, and cloud cover with habitat variables and seasonal timing to predict where and when birds will fly.
The practical applications are surprising in their diversity. Wind energy companies need migration forecasts to curtail turbines during peak passage, reducing bird mortality without unnecessarily losing generation. Aviation needs them for bird strike risk assessment. Urban lighting campaigns use them to promote “lights out” during peak migration nights, reducing the fatal attraction of illuminated buildings.
The counterintuitive finding: integrating terrestrial predictors (vegetation greenness, land cover, latitude) with atmospheric predictors improves forecasts more than refining the atmospheric model alone. Birds aren't just responding to weather — they're responding to the landscape below. A favorable wind over agricultural monoculture produces less migration than the same wind over forest-wetland mosaic, because stopover habitat quality modulates whether birds depart.
The radar wasn't built for birds. The forecast wasn't built for wind farms. The conservation impact wasn't planned. The entire field of radar aeroecology emerged from repurposing a military/meteorological instrument for biological observation — perhaps the most productive unintended application of a sensor network in history.