As dusk descends, there’s a cacophony of sounds in the air at the Queen Elizabeth Olympic Park in London.
Park-goers in this large east London area are listening to a variety of them: human chatter, dogs, insects, birds and, unknowingly, bats.
Bats use high-frequency sounds which are bounced off objects to hunt predators and to navigate. This is called echolocation and it’s inaudible to the human ear, which leaves most of us oblivious to their presence, especially in urban settings.
But a multi-collaborative project in this city is revealing the hidden world of these nocturnal mammals and why it’s important to listen closely to their calls.
Dubbed the Nature-Smart Cities project and launched in June last year, a group of environmental researchers and technologists have been monitoring bat calls for six months.
The team consists of several organisations, including University College London and Intel which are developing the world’s first real-time sensor network to detect and identify bat calls to species.
The resulting data aims to shed more light on the behaviour patterns and trends of these often misunderstood species.
BATS AS BIO-INDICATOR SPECIES
Why bats? Essentially, bats are an ideal indicator species for assessing the health of a natural environment.
They provide pollination and pest control services, feeding on mosquitoes, flies and beetles, among others.
These animals are also sensitive to environmental changes, and healthy bat populations can inform us on the quality of overall biodiversity in a local area.
In spite of their importance to the ecosystem, bat populations around the world, including in the UK, are declining. There are 17 different species of bats in the UK, and they all feed on insects.
In London, at least eight species of bats call this megacity home, and the Nature-Smart Cities project aims to provide the most comprehensive picture of bat life yet throughout the QMOP area.
Automated smart detectors are utilised to effectively capture and monitor their bat calls.
Along with partners from the Bat Conservation Trust, London Wildlife Trust and Arup, findings from this study could help answer questions such as which landscape-scale elements are drivers of bat activity, and how they’re responding to human disturbances in a city environment.
‘SHAZAM’ FOR BATS
To investigate bat populations in east London, researchers have developed a network of prototype smart bat monitors and installed them across 15 sites at QMOP.
These locations encompass different habitats: park squares and amenity places, species-rich grassland, open water, and areas with trees and scrub.
The team from UCL is led by Kate Jones, professor of ecology and biodiversity, whose special interest is in bat research and conservation.
She explains that the Echo Box, the smart bat monitor used for this project, is similar to Shazam (an app that can identify music using the mic on a device) but is specifically for detecting bat calls.
Inside each Echo Box is an Intel Edison chip that processes the acoustic data and a digital ultrasonic USB microphone, and it’s connected to WiFi and power.
The Echo Box performs four steps to capture, process and identify bat calls.
Firstly, the microphone records ultrasonic frequencies up to 96kHz from the environment (20kHz being the limit of human hearing).
Secondly, a three-second audio sample is recorded and stored as a sound file every six seconds, providing consistent sample rates across all smart bat monitors in the park.
Next, the recorded audio is converted visually, known as a spectogram image, which displays the frequency of the calls. Meanwhile, the loudness of the calls are shown by how intense/bright the colours are.
Finally, machine-learning algorithms called Convolutional Neural Networks are applied to scan the spectogram image and determine whether it’s a bat and what species it is in real time.
This audio-stream detection/classification is done rapidly and is transferred to the open source sensor network and an interactive website (www.batslondon.com). The public can access these recordings which are made at night.
POTENTIAL FOR DIVERSE MONITORING
Since the Echo Box was deployed at the park in June last year, over 300,000 bat calls have been detected, with an average of 7,000 bat calls per night and more than 20,000 calls recorded one evening.
These calls have been identified to four species, namely Leisleri (Nyctalus leisleri), Noctule (Nyctalus noctula), Soprano pipistrelle (Pipistrellus pygmaeus) and Daubenton’s bat (Myotis daubentonii).
Moreover, the team discovered that Sensor 7, installed at an open water habitat, has been detecting the most calls.
Other factors that were analysed were the level of precipitation, temperature and big events (e.g. sport matches) carried out in the park affected the frequency of bat calls.
The team hopes that this exploratory network of devices and its platform can be improved.
To make the transition from a prototype to a product, the Echo Box would need to enhance its algorithm development, power sources and data storage. In the future, it has the potential to be used in the wild to monitor a diverse range of wildlife.
“We’re very keen to continue developing this technology into something that can be widely used,” concludes Jones enthusiastically.
“We don’t have any concrete plans, but we’re keen to find new locations, partners as well as sources of funding to continue this project.”