Bioacoustics, an emerging technology using sound recordings to monitor wildlife, can be useful to support human ecologists but should not be considered as a replacement for on-site observation, an industry insider said on Tuesday.
While bioacoustics offers significant advantages for wildlife monitoring, it faces the persistent challenge of misidentification due to the similarity of animal calls, said Geoff Carss, CEO of UK-based bioacoustics company Wilder Sensing during a webinar.
“This is really a tool to extend the work of ecologists, not to replace it,” said Carss, discussing the use of the technology to identify bird species during the online event organised by the Biological Recording Company.
Analysing recordings of animal sounds to understand more about a landscape has been around for decades, but with the recent advances in AI, machine learning, and data storage it has developed into a more scalable way to track biodiversity.
False positives can occur when a species is incorrectly identified as present, mistaking one bird call for another, while false negatives can happen when a species is missed.
“False positives are far easier to recognise because it’s a misidentification. False negatives are much harder because you miss stuff consistently – starlings are a good example of that,” said Carss.
For example, a nightjar bird sounds similar to a mole cricket insect, he said.
“It’s very hard to improve some of the machine learning algorithms, because the calls are so similar.”
Environmental factors, such as background noise and wind direction, can also interfere with recordings.
In one case, Wilder Sensing’s technology suggested the presence of a corncrake, another relatively rare bird in the UK. Human analysis revealed the noise was likely to be the sound of rain hitting electricity pylons.
“Background noise can mask species, which can create a bias,” Carss said.
ADVANTAGES
One advantage of bioacoustic recorders is their ability to operate continuously in areas that are difficult for humans to access, such as deep marshes.
“We can run this for a few hundred pounds a year per recorder. It’s just not physically possible to do that with an on-site ecologist,” Carss said.
Moreover, bioacoustics can often detect species that on-site ecologists might miss. Human presence on site can sometimes alter animal behaviour, which bioacoustics avoids.
“If there’s not much surveying done before, we’ll typically find 20% more species than the on-site ecologists will,” Carss explained.
On-site ecologists, however, can detect quieter species, making the two approaches complementary. “The two together can be really compelling,” he said.
“We can generate pretty accurate, large-scale species data sets at very moderate costs.”
Earlier this year, an executive at Irish data company Treemetrics, said claims made by data companies about the ability to gather accurate AI-powered image-based information remotely via satellites was “complete and utter hype”.
By Thomas Cox – t.cox@carbon-pulse.com
*** Click here to sign up to our twice-weekly biodiversity newsletter ***
