Bats and the Internet of Things.

In the summer of 2017, fifteen customised acoustic sensors were attached to lampposts across London’s Queen Elizabeth Olympic Park. Equipped with ultrasonic microphones, these sensors are using machine learning algorithms to pick out bat calls from the cacophony of the city’s soundscape. Described as the ‘shazaam for bats’, this project is the world’s first end-to-end open source system for monitoring bats.

Developed and installed by UCL and Intel scientists in collaboration with Arup, the Bat Conservation Trust and the London Wildlife Trust, each bat sensor contains a chip that processes three-second long audio samples. Traditional bat detectors convert ultrasonic echolocation signals into an audible frequency that can be heard by humans, but these sensors turn the audio recording into a spectrogram image that’s scanned by deep learning algorithms to identify potential bat calls. Once detected, the calls are sent to an Internet of Things (IoT) platform called OpenSensors.com where the data is open and accessible to anyone with an internet connection. This open access to the data has lead to some creative applications and provides a glimpse of how we can use remote sensors and the Internet of Things to democratise data.

One of the acoustic sensors that is now installed in the Queen Elizabeth Olympic Park. Photo credit: Nature-Smart Cities.

Hearing bats in the dark.

Emerging at dusk, bats leave their roosts in search of insects. Using echolocation to navigate and hunt for prey, a single Common Pipistrelle can eat 3,000 tiny insects in one night. Scientists study bats because changes in their populations can be the first sign that something is happening in their environment, such as a decline in insect population or loss of habitat.

Up to seven species of bats live in Queen Elizabeth Olympic Park and on a warm summer’s night, the sensors can record more than 15,000 calls. Although the algorithm can’t yet identify individual species, this is something the team of researchers is working on. If machine learning can distinguish between the calls of different species, we’d know exactly how many bat species are living in our cities, and which species are most abundant. With this information, we’d quickly detect any changes in biodiversity and have more time to address any emerging issues.

Arcade games and bat gnomes.

A consequence of making data open and more accessible is that anyone can come along and play with the data. Anyone is free to find new insights and develop more creative applications that can engage a new audience.

For example, a Human-Computer Interaction researcher from UCL took the data collected by the bat sensors and made an 80s style video arcade game with it. Players can test their knowledge of bats, watch video clips, listen to samples, and explore an interactive map to see where most bats are being detected. The game’s aim is to address the common misconception that bats are scary, and show how amazing (and cute!) they actually are.

The bat video arcade game made by the Human-Computer Interaction researcher at UCL. Photo credit: UCL

Another team of researchers, this time from UCL’s Centre for Advanced Spatial Analysis, was inspired by the data to 3D-print bat ‘gnomes’ and put them in an exhibition at the Queen Elizabeth Olympic Park. The aim of the exhibition is to make people think about the security issues around our adoption of tools like Google Home that are always listening for certain activation words.

As visitors walk around the park they can ‘talk’ to a variety of gnomes via a bluetooth connection on their phone and learn about the park as well as the technology they are using. In a world where everything is seemingly connected to the internet — even rubber ducks and plant pots — this exhibition is a timely opportunity to educate ourselves about internet security and learn how acoustic sensors, like the bat monitors, work.

Always listening or selective hearing?

In the same that Dads ignore their kids when they ask for extra pocket money, acoustic sensors are programmed to ignore everything but the cues they are listening out for. The algorithms in the bat sensors installed in Queen Elizabeth Olympic Park ignore all sounds below 20kHz, the range in which human speech falls. Therefore they ‘hear’ humans talking but don’t record it. In fact, anything that isn’t a bat call is deleted from the system.

The sensor’s ability to listen 24/7 is what makes it such a good tool for long-term monitoring. There’s no way you could mobilise enough people to conduct surveys in the same continuous, consistent way. By allowing machines to collect data and publish it openly, we give ourselves an opportunity to focus on the analysis. At a recent talk, Professor Kate Jones, one of the researchers leading this bat monitoring project described this approach as a way of democratising data. Anyone can access the data and they don’t have to go through any gatekeepers to get to it. This means we have fresh eyes spotting new insights and testing new ideas to help us understand what’s happening in our world.

In this particular project, two API’s make it easier for people to access and query the data: one is a REST API that can query historic data, and the other is a Realtime API that allows applications to register for instant notifications as the sensors generate new data. For example, the Realtime API provides a live bat call count at www.batslondon.com.

Point number 7 is the location of a previously unknown bat roost. Screenshot of the map taken at 14.27 on Wednesday 6 December. http://www.batslondon.com/ Bats spend most of their winter hibernating so there are few calls at this time of year.

Scaling up.

UCL’s approach to monitoring bats could potentially be scaled up and used anywhere in the world, for any species. At the talk where this project was presented, a member of the audience voiced concern that this kind of technology would be hard to use in places where internet connections are poor, potentially furthering the divide between countries with remote sensing capabilities and those without. This risk of division is exactly why improving connectivity and digital literacy must be an important priority across the globe. However, remote monitoring and open data could have the greatest impact in places where conservation resources are limited, so we need to find ways to make this technology more accessible.

When we think about using the Internet of Things for conservation aims, we need to be aware of what data is being recorded and who it’s being shared with but that doesn’t mean we should be fearful. With careful design and the correct application, sensors that are connected to the Internet of Things can teach us more about our cities soundscapes and the wildlife that live unseen and unheard alongside us. An open system is the gateway to more a democratic approach to data access, offering the potential to ignite excitement and renew people’s interest in nature. And that’s good news for our planet.


Want to learn more? Visit the Nature-Smart Cities website.