Automated cameras record Serengeti life

By Jonathan Amos
BBC Science Correspondent

Image source, SnapshotSerengeti
Image caption,
On the menu: Predators are seen to walk past the cameras casually carrying a meal

Hundreds of thousands of images of animals going about their daily lives have been caught on automatic cameras in Serengeti National Park, Tanzania.

The pictures have allowed scientists to observe the natural behaviour and interactions of the creatures, free from any disturbance by humans.

Eating, sleeping, grooming, playing, fighting - any activity in the field of view was recorded.

An army of public volunteers helped to classify all the images.

And their work - as part of the snapshotserengeti citizen science project - is now reported in a scholarly paper in the journal Scientific Data.

Image source, SnapshotSerengeti

As well as informing ongoing ecological research in the Serengeti, the giant picture pool has uses in computer research.

Programmers need large annotated datasets with which to "train" pattern-recognition systems, and the Serengeti bounty is perfect for this application.

Dr Ali Swanson and colleagues placed 225 cameras across more than 1,000 sq km of the Serengeti - some attached to trees, others to metal posts.

The cameras' trigger was an infrared and motion sensor, with each trespass in front of the lens initiating a burst of up to three shots.

The network was set up in 2010, and by 2013 had collected 1.2 million burst events.

Out of these image sets, the snapshotserengeti community identified more than 322,000 animals, distributed across 48 separate species.

Image source, SnapshotSerengeti
Image caption,
Cheetahs tend to get beaten up by the local lions

The most frequently seen animals were those participating in the plains' great migrations.

There were more than 100,000 wildebeest sightings, for example. Likewise, passing zebra triggered the cameras on more than 70,000 occasions.

But included also in the image sets are some rare and elusive animals, such as the aardwolf and the zorilla.

Ecologists are already mining the pictures to learn how different species interact.

"For part of my dissertation, I looked at how lions and cheetahs divide up habitat hotspots on very fine timescales," said Dr Swanson, who was affiliated with the University of Minnesota, US, for the research.

"The lions beat up cheetahs - they kill them, and steal their food. So, I was interested to see how cheetahs managed to co-exist given that the lions were so mean to them.

"And with the cameras, you can see how they avoid the lions, kind of in the moment, sneaking in just when the lions leave the scene."

Image source, SnapshotSerengeti
Image caption,
The triggers use infrared sensing, which also detects fire outbreaks

Dr Swanson has spent much of the past five years in a vehicle driving from camera to camera to change batteries and retrieve data cards.

She reports a surprising amount of damage to the camera positions.

"The herbivores will rub up against them, but the hyenas will just chew through them," she told BBC News. "And then the elephants rip the cameras off their mounts and throw them across the ground. Sometimes I'd get to a camera position and find bits of plastic strewn all around."

Dr Swanson said snapshotserengeti had now classified up to 1.9 million burst events, with perhaps another 200,000-300,000 still on the SD cards in the cameras waiting to be recovered.

How much longer the project can continue is uncertain, however. It is about to run out of funding.

Image source, SnapshotSerengeti
Image source, SnapshotSerengeti
Image caption,
Elephants rip the cameras from their mounts and hurl them across the ground
Image source, SnapshotSerengeti
Image caption,
The dataset can be used to "train" pattern-recognition systems and follow me on Twitter: @BBCAmos