BatEchoMon, India’s first automated bat monitoring, detection system

For her PhD analysis, bat biologist Kadambari Deshpande made in a single day recordings of bat echolocation calls within the Western Ghats. A “good night time” would generate about 30 GB of information from 11 hours of recording with a bat detector. To course of the information, Deshpande would undergo a number of one-minute recordings, scanning each millisecond for bat calls, and make notes on the species and different data on their behaviour and ecology.

“It took me 11 months to course of 20 nights of information,” Deshpande mentioned. “BatEchoMon can most likely give me that in a number of hours.”

BatEchoMon, brief for “Bat Echolocation Monitoring”, is an autonomous system able to detecting and analysing bat calls in real-time. It’s India’s first automated bat monitoring system, developed by Deshpande and Vedant Barje underneath the steering of Jagdish Krishnaswamy, as a part of the Lengthy-Time period City Ecological Observatory within the Faculty of Atmosphere and Sustainability on the Indian Institute for Human Settlements (IIHS), Bengaluru.

Deshpande is a postdoctoral fellow on the Observatory and the Faculty; Barje, who leads the WildTech Mission on the Wildlife Conservation Belief, is a marketing consultant there.

BatEchoMon marks a brand new chapter for bat analysis within the nation, in keeping with Deshpande. The monitoring system permits chiropterologists — scientists who research bats — to “transcend information processing and in the direction of asking attention-grabbing questions on bat ecology”.

“Not solely will it result in a smoother workflow, it’s going to assist individuals transition to recording bats in several elements of the nation, permitting us to achieve extra insights on the pure historical past and ecology of various bat species,” Rohit Chakravarty, a bat researcher and conservationist on the Nature Conservation Basis, mentioned.

“I don’t know of any system internationally with an in-built recording plus name classifying unit. If my data serves me properly, BatEchoMon is a milestone in bat analysis globally.”

The bat within the machine

BatEchoMon is greater than a bat detector. Apart from a recording system, it contains parts that may report, retailer, course of, and analyse species-wise bat exercise on the fly. Bat detectors are solely the specialised recording gadgets that may convert the ultrasonic echolocation calls of insectivorous bats into audible sounds for people. “In [BatEchoMon], Audiomoth, a preferred low-cost ultrasonic detector, has been configured to work as an ultrasonic microphone,” Barje mentioned.

BatEchoMon is programmed to activate routinely at sundown, when bats start flying, and constantly listens and analyses audio. The system’s mind is a Raspberry Pi microprocessor, which processes the information captured by Audiomoth. “It first isolates bat calls from different ultrasounds, similar to these of bugs or anthropogenic and environmental noises. Then the height frequency and construction of a bat name are analysed to match a recognized pre-trained mannequin, which helps establish the bat species,” Deshpande defined.

“The system makes use of a [convolutional neural network] based mostly algorithm to do that,” Barje added. The output from the system is a spectrogram — a visible illustration of the frequencies of an audio sign because it varies with time — of all detected bat calls, together with audio recordings of the parts with simply the calls. The system additionally generates statistical information on which species has been most energetic by the night time, which species was energetic when, and so forth.

“Earlier, all of this wanted to be interpreted manually, after painstakingly combing by hours of information,” Deshpande mentioned.

The Raspberry Pi and its related processing parts are enclosed in a field measuring 200 mm × 80 mm × 80 mm. Different auxiliary parts within the system embrace a photo voltaic panel plus battery and a WiFi communication unit for energy provide and information transmission, respectively. Within the absence of the solar, the system can final for as much as eight days, in keeping with Barje.

BatEchoMon additionally has a modular design, and its battery, charging equipment, and the extent of automation and information relay might be customised to the area it’s put in in. However the staff was reluctant to disclose extra concerning the setup course of at the moment.

‘Instantly, it turned potential’

Bat ecology and acoustics is a nascent subject in India, with only a handful of bat researchers recording bat calls and analysing them for ecological research. International bat-call databases similar to ChiroVox and Xeno-Canto have few recordings submitted by Indians.

Deshpande has been utilizing bat detectors since 2008 and has noticed their evolution worldwide. In Europe, she mentioned, detectors outfitted with the related software program and reference libraries have saved scientists lots of time. Since then she has wished to develop one thing comparable, however customised for the insectivorous bats which are extra widespread in India.

An opportunity assembly with a fellow researcher and engineer from her hometown in Nashik, Maharashtra, kickstarted the BatEchoMon mission. “Assembly [Barje] simply revolutionised the entire thought. Instantly, it turned potential,” Deshpande mentioned.

The duo went by quite a few iterations, attempting completely different microprocessors, algorithms, and energy options — a “large problem”, in keeping with Barje — earlier than arriving on the present model of BatEchoMon. Their main purpose was to incorporate all the specified performance in a user-friendly, low-cost bundle.

The core system of BatEchoMon prices one-third of superior detectors and comparable techniques, in keeping with Barje. He didn’t want to disclose actual numbers, nevertheless.

The primary problem

In the previous couple of months, BatEchoMon has efficiently accomplished pilot checks in an IIHS website in Nashik. The staff plans to check it for longer durations and in numerous situations in addition to to beta check the system with choose customers outdoors the organisation.

“All people comes with their very own experiences of acoustics and completely different species. So ultimately, we wish our colleagues to check it for themselves and share their experiences in order that we will enhance on the present system,” Deshpande mentioned.

The first impediment for BatEchoMon is the restricted availability of reference libraries for the calls of many bat species. “Presently, the system can establish six to seven widespread Indian bat species. Going ahead, we want to embrace as many bat species as potential,” Deshpande mentioned.

In addition they mentioned they hope the system in its current kind will be capable to establish the species generally seen in city, peri-urban, and human-modified forested areas. The primary problem is to create sturdy coaching datasets to make good detection fashions for various species — a problem they hope to beat by collaborating with different bat researchers.

Fortuitously, collaborations amongst Indian researchers are enhancing due to initiatives such because the ‘State of India’s Bats’ workshop performed by the Nature Conservation Basis and Bat Conservation Worldwide, in keeping with Chakravarty.

“One of many key data gaps recognized by individuals on the workshop was the dearth of an intensive reference name library,” he mentioned. “We additionally want extra funding to conduct surveys in several elements of the nation. It’ll enable researchers to establish extra species and collect extra recordings for coaching datasets.”

Nikhil Sreekandan is an unbiased journalist.