Last week I ran two workshops for BCT on bat identification. The first whole day workshop was looking at manual identification of bat echolocation. We looked at call characteristics and how we need to understand issues such as directionality, attenuation and analysis protocols to understand what is going on - basically to 'read' a sonogram for the story it tells. Most of this was done through Batsound, which I still think is the best single package for sound analysis, though it does lack some of the other features such as GPS extraction (though I believe the new version of Batsound Touch can do this in some form - it can certainly tag sound files with GPS location anyway).
The second whole day workshop was looking at automatic identification and some of the packages that can be used to do it. We reviewed the principles behind it, some of the assumptions it makes, then looked at the range of software available. We then focussed on Wildlife Acoustics 'Kaleidoscope' and Biotope's 'Sonochiro'. We ran the same sets of sample calls through each and looked at how we interpret the outputs. The participants then had an opportunity to try each of the packages themselves on sets of pre-identified calls and also on some standard field data from a range of different hardware types. Thanks to both Wildlife Acoustics and Biotope for the training licences for each.
In general we found that both packages were pretty good at picking out pipistrelles, though identification of other species was variable. The take home message was really that these packages help to scan huge volumes of files and identify species of interest, but those files then really have to be manually checked, or at least a sub-set manually checked for validation.
This is pretty much what BCT recommends in some guidelines it has published on its website. With fifteen people on each course, it was a great opportunity to explore bat identification. I certainly learnt a lot, and I was running it. We're running it again in September.
No comments:
Post a Comment