March 12, 2008 -- New radar software can quickly and accurately differentiate signals from birds and swarms of insects, a problem that stumped previous efforts to automatically tell the two apart. The findings could aid air-traffic controllers and help biologists better track avian migrations.
Collisions between planes and birds – known as bird strikes – can damage aircraft and are a potential safety risk. They are most likely to occur on lower-altitude flight paths near airports. The UK Central Science Laboratory estimates that, worldwide, bird strikes cost airlines around $1.2 billion annually in repairs and aircraft downtime.
Radar signals bounced off flocks of moving birds are easily confused by those from large numbers of tightly packed flying insects. Trained technicians can recognize subtle pulses in the signals caused by bird's beating wings to tell them apart. But that is not easy, especially for large, mixed flocks of birds and insects.
Air traffic controllers do not always have time to do that, and it causes problems for biologists studying bird migrations too, says Serge Zaugg of the Swiss Ornithological Institute in Sempach.
"If you have to spend several seconds on every signal and you have tens of thousands of signals, it can take months to separate the birds from the insects," he says.
Zaugg and colleagues at research institutions in France, the Netherlands and Germany combined data-mining techniques, statistical analysis, and artificial intelligence to create a computer algorithm that learned to spot birds nearly every time.
The team fed the program thousands of bird and insect radar signals that had already been identified by human experts and verified with visual observations. The signals came from flocks of birds or insects, or a mix of birds and insects, flying over the Sahara desert in Mauritania in 2003 and 2004.
After the training, the program can identify signals it has never seen before with 93 to 98% accuracy.
Kenneth Rosenberg, of the Cornell Lab of Ornithology, Ithaca, New York, US, says others have previously distinguished between birds and insects by correlating wind speed with movements detected via radar.
Whereas insects move with the air stream, birds can fly faster or in the opposite direct of the ambient air flow. Yet as far as Rosenberg is aware, all prior efforts relied on non-automated, human identification.
"Having a system that can quickly and correctly distinguish between birds and insects would be extremely beneficial for air-traffic controllers and biologists interested in bird migration," Rosenberg says. "Radar is potentially one of the most powerful tools we have to study bird movements, especially at night when we can't see them."
Other approaches to spotting birds include infrared detectors and even "audio telescopes" to identify different species from their calls.