ZSL’s Alasdair Davies tells us how it’s done.
Alasdair Davies is a Technical Advisor for the Zoological Society of London’s Conservation Technology program and a web developer for the EDGE of Existence program. His current focus is the delivery of ZSL’s Instant Wild project, the advancement of camera trapping technology, and the future development of the EDGE website. Alasdair is also a founder and director of the primate conservation organization The Great Primate Handshake. We conducted this interview via email.
When and where was the idea for Instant Wild hatched?
ZSL’s Instant Wild programme started life on a staircase within Conservation Programmes at Regent’s Park, London—better known as the location of ZSL London Zoo. It was one of those “Have you seen the new GSM-enabled camera traps? Aren’t they great…” conversations whilst holding the morning’s first mug of coffee and checking in on the day’s schedule.
Is there a story behind it?
Although the conversation on the staircase was brief, our Director of Conservation Programmes, Jonathan Baillie, was luckily the other person on the stairs that morning. Later that afternoon, he called me into his office. It was evident that he had been pondering the morning’s conversation and I could sense that there was an exciting idea on the table. Within in an hour, the name Instant Wild was decided upon and the concept of sending the very photos the cameras take to the general public for instant analysis was founded. We’d have hundreds of thousands of eyes scanning thousands of photos from across the globe every single day. It could answer so many questions, and there wasn’t a second to lose.
Where were the first cameras installed, and what kinds of lessons did you learn right away?
We’d start with just one or two cameras. That was the plan. We’d need a website, and of cause, an app, so we hired the services of James Sanford, an iPhone app developer, and drafted how it could work technically. We just needed to figure out how the general public were going to help us identify the animals captured in each photo and where our first location would be.
It took about 6 months to get the basics together, and in November 2011, Instant Wild was launched with a handful of cameras in Kenya, Sri Lanka, and the United Kingdom. We asked friends, volunteers, and colleagues in-country to buy a few local SIM cards, set the cameras up to email the photos through to our server, and then we’d push them out to the app and website in real time. It would take less than 60 seconds to display a photo from the wild.
What technologies are currently used for data collection, transmission, storage, and analysis? (hardware, software, etc.)
We currently use GSM-enabled camera traps, but have recently developed satellite-enabled camera traps, as we are of course always limited by the GSM coverage, especially in remote or difficult-to-reach locations. The cameras detect the movement of wildlife by the use of PIR sensors (passive infrared). Essentially, they are able to detect tiny changes in the ambient temperature (i.e., when a warm-blooded animal walks past), and it is this that triggers the photo to be taken.
What’s the process by which an image is captured, transmitted, identified, stored, and put to use in research?
Diagram A details how we deliver images today. Diagram B details our future system in development at the moment, to be released in 2014.
Who maintains the cameras in the remote locations?
At present the cameras are maintained by volunteers or field conservationists. We are incredibly grateful to everyone who has helped thus far, as well as those who continue to help maintain our cameras by replacing batteries and configuring them for transmission at each new location.
Are there interesting and/or difficult challenges involved in keeping the cameras working?
The biggest issue we face is poor or unreliable GSM coverage. The cameras need to transmit data and so require a GPRS connection as a minimum, but fare much better if EDGE or higher is available. We sometimes only acquire 25% of the actual images taken by a camera in a given day, as the signal can sometimes fluctuate. The second major issue is resilience to environmental conditions. Moisture can erode battery containers and ants often decide that the camera’s internal electronics are a good place to make a nest and call it home.
Of the million+ animal identifications thus far, how many extremely rare animals have been spotted?
Our two most important achievements have been the sighting of a critically endangered Javan Leopard on our Indonesian camera, proving its existence in the area and the discovery of a Mountain mouse-deer on the Sri Lankan cameras. Found in the highlands, it was believed to have only been photographed once in the wild three years ago until it was spotted by members of the public on Instant Wild.
What happens with the identification information? Who uses that data, and to what end? Is the data made publicly available?
Instant Wild supports ZSL’s conservation work in several ways. First of all it engages members of the public in conservation research and improves awareness and knowledge of the species identified. Secondly, by passing the identification of camera trap images to the general public, conservationists can analyse the data quicker, which helps us make informed conservation decisions. In our last analysis of user identifications, we recorded a success rate of over 90%. Thirdly, Instant Wild provides ZSL with the ability to quickly know if a rare and threatened species has been spotted, and helps us to build up a picture of species type and density in a particular area.
In 2013, ZSL was awarded a $750,000 Google Global Impact Award to fund next-generation camera traps equipped with automated sensors to better protect threatened wildlife. What new technologies will you be bringing online as a result of the Google grant? Will ZSL be building the new cameras/mics? Can the (tech-savvy) public help in any way?
We’re building a system that utilises a number of evolving technologies. It’s a system that will combine a number of new low-powered sensors in the form of cameras, but also seismic and audio-based sensors to sense the presence of wildlife or human encroachment (vehicle entry, gunshots, etc…).
By using low-powered Atmel processors and RF, we will transmit the information the sensors capture to our satellite-enabled “nodes” up to 500m away. We will then be able to analyse the sensor data received prior to transmission over the Iridium satellite link by filtering photos through visual recognition software, or matching sound clips against audio signatures using a Raspberry Pi at the heart of each node. We’re always open to suggestions from the tech community as to how we can improve our system, whether it’s by exploring new power supply methods or just improvements to our app.
What makes the new technologies better than the equipment and software you’re currently using?
From a software perspective, visual recognition algorithms are constantly evolving and we intend to maximise their potential to help us optimize our system. If we can correctly identify the presence of a human in a photo 100% of the time, for example, then we can guarantee that the system is able to alert rangers and law enforcers even before human eyes scan each photo.
From a hardware perspective, one of the exciting developments is the backwards compatibility of the Iridium modems in our satellite nodes to send 10 x the amount of physical data when Iridium launches the Iridium NEXT satellite constellation in 2016. We’ll then be able to send higher resolution photos, audio, and even video.
Will you be launching an Android IW app?
Get a closer look at Instant Wild in the video featured in the upper right corner of the page. For more images and news about the project, follow Instant Wild (@InstantWild) and the ZSL (@OfficialZSL) on Twitter.