Smart Birding

A new birdsong app identifies feathered friends by their tweets

Squinting into windblown trees and bushes is for the birds—especially if it’s birds you’re looking for.

“You have to listen. There’s no way around it,” says Mark Berres (photo right), an ornithologist and CALS animal science professor. “The most difficult aspect of bird-watching is call identification, but calls are the most important tool for identifying birds.”

Even experienced birders have trouble matching more than a handful of songs with species, but Berres may have answered the prayers of bird-watchers, researchers and even the most casual naturalist.

Not surprisingly, salvation comes in the form of a smartphone app: WeBIRD, the Wisconsin Electronic Bird Identification Resource Database.

WeBIRD users can record a nearby bird’s call, submit that song wirelessly to a server and retrieve a positive ID of the species.

“I am amazed at how good it is,” says Berres, who also has used WeBIRD to identify grasshopper species by their clicks and frogs by their croaks. “Not only can WeBIRD tell you which species you’re hearing—in some cases it’s good enough to identify individual birds from their song.”

That’s no mean feat. Birdcalls can differ throughout the day, among groups just miles apart, and by individual birds.

“When a bird sings, the song itself may have varying amplitudes and frequencies,” Berres says. “It can also speed up a little bit and slow down a little bit. They may throw in a note here or take out a
note there.”

WeBIRD dices songs into time-ordered chunks, using data-organization techniques often applied by geneticists to jumbled bits of DNA to “align temporally misaligned data, working around a lot of the variation,” says Berres.

Berres expects WeBIRD—which could be available to the public in time for the 2012 spring migration—will enable field research through remote recording and analysis. More important, he hopes WeBIRD will help birds.

“If people can appreciate intrinsic beauty—and birds have got that part down—a closer awareness of the natural world will follow,” says Berres. “Fostering a connection with wildlife is one of the ways we’re going to save it, and WeBIRD puts that connection to birds in the palm of your hand.”

Click here to watch a WeBIRD demonstration with Mark Berres.

A JUNE 2013 UPDATE FROM MARK BERRES in response to many inquiries about WeBIRD:

Hi Folks,

Sorry to be quiet for so long. Spring semester is always a very busy time for me (I teach three 500-level courses) and this year doubly so as I have been working in Vietnam as well. Now that the semester is finished a little more time can be devoted to WeBIRD activities.Many of you are wondering about the status of WeBIRD. I can assure you that we are still working on it, albeit slowly. Last spring and early summer we tested WeBIRD in the field. Identification of resident and local species (i.e. those in Madison WI) presented no difficulties as expected. But when the first migratory birds started returning, WeBIRD did not perform well; the match significance was particularly poor for these non-resident species.A little inferential work led us to an issue that we already knew about, but thought we had covered. One aspect of avian vocalization is that most species exhibit substantial variation in their songs (and calls). Moreover, this variation is structured geographically. There are several biological explanations for this phenomenon (an important one is natal philopatry), many of which may also explain how human speech dialects are also structured geographically. Thus, a Tufted Titmouse in WI does not sound like a Tufted Titmouse in ME. Similar, yes, but in many species (like the White-crowned Sparrow along the west coast) the differences can be considerable.

How does this affect WeBIRD? If you query a song of a species for which there is no corresponding entry in the WeBIRD database, you will obviously not obtain a match. Alternatively, if you have a song in the database that is “reasonably similar” to your query, you will obtain at least an indication of a match. Once a candidate match is made, WeBIRD evaluates the statistical significance of the match (this is the most important step). The level of significance will vary depending on the degree of similarity between the query and specific database entry. Therein lies the problem: if no “reasonably similar” songs exist in the database, accurate identification of the species is impossible (or at least statistically unlikely). Given that substantial geographic variation in bird vocalizations exists makes this a formidable problem.

Although we had numerous exemplars of songs from non-resident species included in the WeBIRD database, it wasn’t sufficient for WeBIRD to work with. These songs were derived from commercial sources (e.g. bird vocalization CDs, internet sources, etc.) but almost all were recorded at locations very distant from WI.

More digging led us to a curious discovery: many songs, although from different sources, featured the same – or nearly the same – song of a given species. And many of these could be traced back to entries accessioned into the Macaulay Library of Natural Sounds! Sadly, what this really means is that there are surprisingly few [good] recordings of bird species throughout their range. This also explains, in part, why listening to audio CDs of bird calls can be very frustrating for beginners when learning avian vocalizations: their usefulness depends on where you live).

Practically speaking, limited numbers of song exemplars – specifically songs from the same species but recorded in different geographical regions [cf. dialects] create both false-positives and false-negatives for WeBIRD. Thus, for anyone outside of the general southern WI area, WeBIRD is probably not going to work very well with its current database of Madison residents. This is precisely the reason why we have not yet released WeBIRD publically.

Is there a solution to this problem? Yes, there is. We are planning to engage the assistance of citizen scientists – people like you – who wish to make this project a reality. Crowd-sourcing vocalization exemplars will be necessary as no single team of researchers, no matter how diligent or organized, could ever hope to amass different songs of birds distributed throughout the country. To this end, we are now working on a beta version of an app that will allow participants to record and upload songs for inclusion into the WeBIRD database. We hope to integrate simultaneously a website to monitor the progress of participation and accumulation of bird songs (perhaps even turning it into a game). If enough contributions are made, then WeBIRD will really become a reality that everyone can enjoy.

Sincerely,
Mark