For the first time, new maps are using radar to forecast nighttime clouds of migratory birds and track their flights in near real-time.
Scientists with the Cornell Lab of Ornithology and the University of Oxford, England, made the breakthrough: processing weather radar data to produce maps that visualize migration on the Cornell Lab’s BirdCast website.
One map shows an animated visualization that tracks migration in near real-time. Cornell Lab postdoctoral associate Adriaan Dokter designed an algorithm that rapidly estimates the intensity and flight directions of migrating birds detected by the weather radar network. The system processes incoming radar data continuously and updates the animated map every 10 minutes.
“We’re able to isolate bird data from atmospheric information because of the way weather radar works – a process called dual-polarization,” Dokter said. “This means the radar stations transmit and receive radio waves in both vertical and horizontal directions. It provides a much clearer picture of the size, shape and direction of the targets it picks up. And with the power of cloud computing, we can analyze all radar data incredibly fast.”
Another map forecasts migration three days ahead. The color-coded displays combine projected weather conditions and bird movements to show where and when the most intense migrations are expected. Most songbirds migrate in darkness, usually when weather conditions are favorable. Tailwinds can produce massive migratory movements. Rain can shut down flights entirely.
“Knowing when and where a large pulse of migrants will pass through is useful for conservation purposes,” said Benjamin Van Doren ’16, a doctoral candidate at the University of Oxford. “Our forecasts could prompt temporary shutdowns of wind turbines or large sources of light pollution along the migration route. Both actions could significantly reduce bird mortality.”
Noted Kyle Horton, Cornell Lab postdoctoral associate: “This is the most significant update since we first began using radar to study bird movements. From the birdwatcher’s perspective, if you know where and when migrants will be flying at night, you stand a better chance of seeing them, especially if the birds make a stopover in your area.”
Van Doren and Horton designed the system that generates the migration forecast maps. They used machine-learning models based on 23 years of radar and weather data to predict suitable conditions for migration occurring three hours after local sunset.
“These forecast and live migration maps, and the research that produced them, represent a breakthrough nearly 20 years in the making,” Cornell Lab migration researcher Andrew Farnsworth said. “We hope these maps will provide perspective to the expert and novice alike on the amazing spectacle – and the sheer magnitude – of migration. Beyond that, we believe these maps will become powerful tools for conservation action to help reduce the impacts of human-made hazards birds face during their incredible journeys.”
This research was supported by funding from the National Science Foundation, Leon Levy Foundation and NASA. Additional funding was provided by the Edward W. Rose Postdoctoral Fellowship at the Cornell Lab of Ornithology, Marshall Aid Commemoration Commission in the United Kingdom, and Amazon Web Services Cloud Credits for Research.
BirdCast is a collaboration among the Cornell Lab of Ornithology, University of Massachusetts Amherst and Oregon State University, and was funded by grants from the National Science Foundation and Leon Levy Foundation.
Pat Leonard is a staff writer at the Cornell Lab of Ornithology.