If you want to figure out how many different species of birds are in a forest, all you have to do is listen. But there are a few ways you can go about such a study: You and your colleagues can hang out in the forest for 12 hours a day, five days a week. Or you can set up microphones and record the forest sounds for analysis back in the lab, in which case you’ve got a few more options — let a computer listen and identify species, make your students listen to all of it, or listen to only some of the recordings.
But which way is best? The computer isn’t quite talented enough at this point, and though students are cheap, they’re not cheap enough to listen to all of the sound you’d collect. So the options are to count in person or listen to only some of the recordings. But which ones? And are they better than going into the forest yourself?
To figure that out, researchers from Queensland University of Technology in Australia set out to determine how many species of birds were in a eucalyptus forest in southeast Queensland. The study (in press) appears online in Ecological Applications.
They recorded five days of sounds from several sites in the forest and also had two experienced bird surveyors conduct on-site surveys of those areas at dawn, noon, and dusk for five days. Once the recordings were back in the lab, they were sampled in one-minute bites in five different ways (listening to recordings longer than 240 minutes is cost prohibitive and impractical for most researchers): randomly over the 24-hour period, randomly only in the three hours after dawn, randomly only in the three hours before dusk, randomly after dawn and before dusk, and every half hour over the full 24 hours.
Across all the sites in the forest, the researchers found 96 species using acoustic methods and 66 species with the traditional, in-person survey. And when they looked at their samplings of the acoustic data, they were able to find the most species in those three hours right after dawn, almost twice as many as can be found in the same time by the traditional method.
Sampling the recordings isn’t perfect — it doesn’t find all species, especially rare and cryptic ones — but the researchers say that using a computer to help identify the parts of a recording more likely to have a bird call would help avoid getting random samples of only wind. Thus, combining computer analysis with human listeners of these recorded forest sounds could help scientists get a better picture of biodiversity.
Relying solely on the computer to identify so many species isn’t an option for now, the researchers say, because the programs are plagued with false positives and negatives. However, one day they may get good enough that they can detect birds with high accuracy and cut out the need for humans to listen to hours of sound. (But there’s something about that that makes me a little sad.)
Image of a forest kingfisher courtesy of JJ Harrison, via wikimedia commons