Recently at work I noticed several new Wildlife Acoustics boxes on my colleague’s desk, and I had to ask about them. They were bioacoustics monitors, used to record wildlife sounds for identification later on. These are typically set up in remote areas (or in our case, soon-to-be-developed areas) where they’re strapped to trees, left for a few days or weeks, and then the acoustic data is downloaded for analysis. My colleague is an avid birder and can identify a ton of different bird calls, but also listens and compares them to known recordings of certain bird species in order to confirm what was recorded. This is secondary; however, to being out in the field for several days during the early mornings (when bird calls are most active) to listen and identify birds in person.
Then I read this post on AskMetafilter by cardinality on the issues with using software such as Shazam to identify bird calls. The difficulty seems to be in the minor variations in bird calls due to regionalization that reduce the ability of software to compare and identify the bird song, since there essentially needs to be a huge database built up in order to have a basis for correct identification. And indeed, my colleague confirmed this – there is working software available to automatically catalog the recorded sounds; however, all of the recordings they make still have to be sorted through by hand (er, ear) to confirm the software categorization. The software does end up making the whole task a lot quicker though.
By the way, all of this work to identify birds in an area is usually done in order to determine what possible effects the development of the area (be it a mine, a housing development, new highway, etc.) may cause. This includes the assessment of noise (my area), as well as air pollution, fish, groundwater, and a whole host of other considerations.