Snotbot

Mapping drones aid in wildlife monitoring and conservation

Image credit: Christian Miller

Smarter, faster, and cheaper – drones are helping conservationists to map, track and sample hard-to-access species like migrating whales and denning polar bears without disturbing the animals.

In order to develop a conservation plan for a species, scientists need to know how many animals there are, where they are, and how they use the environment. The same data is needed when human changes to an environment take place so that the impact can first be anticipated, and then can be mitigated from the outset.

The introduction of offshore wind farms or the conversion of rainforest habitat to palm oil production can have major impacts on the animals that live there. The usual way to get this data was to put biologists’ boots on the ground – not so easy when the animal is nocturnal and cryptic, or lives most of its life underwater, or in extremely remote areas in challenging conditions.

Serge Wich, a conservation biologist and professor of primate biology at Liverpool John Moores University, was one of the first biologists to deploy drones in field research in 2011, and like most of the pioneers in the field, started by building the drones himself.

“Our first flight over the rainforest in Sumatra clearly showed the structure of the forest and animals that lived in the forest canopy and the potential of the tool. We built our first drones ourselves and they were relatively cheap,” says Wich. “Back then, off-the-shelf technology was extremely expensive, but of course all that has changed. Compact, lightweight systems with a choice of optical, multispectral, and thermal sensors and excellent obstacle avoidance systems are freely available now.”

After working on many different primates, including orangutans and langurs, Wich and his colleagues headed to Madagascar to Lake Aloatra to collaborate with the Durrell Wildlife Conservation Trust, which has been involved in the conservation of the critically endangered Bandro lemurs that live exclusively in the reed beds and papyrus at a single lake. The only surveys previously had been made by canoe along existing channels used by local fishermen and did not cover the entire area.

Drone surveys could cover a far greater area, previously inaccessible to biologists. Using a thermal camera just after dawn, the warmer animals really pop out against the background vegetation, permitting an accurate population count, which just is not possible using any other means. In a single 20-minute flight, the drone covered more marshland than a canoe team paddling around over two days searching for the elusive mammals.

Wich is also a co-founder of Conservation AI, set up to address one problem with the use of drones – the need for a human to review the huge amounts of data generated. “We are building deep-learning models to automate detection of animals in drone footage,” he explains. “Human observers look at the accuracy of these models and then improve them. We are trying to find a sweet spot – a model that makes the same amount or fewer errors than a human researcher, but can review footage much, much faster.”

Even a trained human can take a minute to look at a photo, Wich says. “One drone can generate an image every two seconds, 1,800 images in an hour. A researcher would take 1,800 minutes or 30 hours to analyse the data from one hour of flying – a trained AI can do that in a few minutes.”

He added: “The aim is to make data collection and analysis equally fast. We already have an AI trained on the animals of the African savannah that can correctly identify giraffes, lions, cheetahs and zebras and many more species from aerial images.

“The number of species that can be accurately identified is growing all the time, and for example, will help inform managers of national parks or game reserves about their animal population trends.”

Drones can be an efficient alternative to expensive helicopter or fixed-wing plane surveys when surveying polar bear populations. Scientists on Wrangel Island in Russia have been using drones to count the polar bear population. Four hundred and twenty bears were identified from an aerial survey on the Umka-2022 expedition and the count will be combined with ground observations, using machine learning to eliminate any double counting of animals.

Drones are not equipped for some of the hazards encountered during their work, and polar bears are scared of little and naturally very curious. One drone had to make an emergency landing due to strong katabatic winds down-draughting from the mountains and glaciers on the island. “The bears chewed him extremely badly,” says an expedition member, “but we managed to retrieve him and save him eventually.”

In Svalbard and East Greenland, bear researchers have been using drones equipped with thermal cameras to locate the dens of female polar bears, which are 10-15°C warmer that the surrounding snow slope. It is easier to spot the thermal differential using a drone than to hang out of a helicopter to spot a one-metre den entrance from height.

Working with polar bears is inherently dangerous and the use of drone technology can keep scientists off the avalanche-prone mountainsides, and away from unwanted close encounters. The zoom cameras on the drones mean that the mother and cubs are not disturbed by noise. But the technique, known as FLIR (forward-looking infrared), is far from perfect, and can fail to detect deeper dens or operate effectively in poor conditions like high winds or driving snow.

Snotbot

Image credit: Christian Miller

Conservation scientist Dr Ryan Witt of the University of Newcastle in New South Wales, Australia, has been using drones to help his team detect cryptic arboreal species – well hidden and often camouflaged animals like koalas that live in the country’s forests. Counting the furry marsupials is difficult using conventional means such as scat surveys or spotlighting for eyeshine at night. Researchers searching forests at night on foot with spotlights found, on average, about one koala every seven hours.

Flying the thermal drone at night in the same forests was extremely efficient – the team found an average of one koala every two hours. And this was in an area with a notoriously dispersed population. The team started by borrowing a pilot and  thermal-camera-equipped drone in 2019 but now have their own drone. “The latest thermal imaging drones from DJI are giving us great results, at least in some forest types,” says Witt. “We are still learning where drones are most effective at spotting koalas. They are good in coastal eucalypt forests and open woodlands. Using drones we can find the actual animal very efficiently and have the field team move to the location to collect scat.”

Witt added that the team can also use the RGB camera sensor to photograph the animals to assess sex, or potentially determine if a koala is back with young. “Before drones, gathering these types of data sets were nearly impossible simply because the animals were so hard to find,” he adds.

The team have never killed a drone and considering that obstacle avoidance technology does not work at night that is quite an achievement. “Right now we manually review all our drone imagery but are working with another team to develop machine learning to take on that onerous and time-consuming task,” Witt explains.

If working with rare and often elusive land animals was not hard enough, the scientists who study mammals like whales, which spend most of their lives underwater, have a challenge of an altogether different order. For decades, researchers would chase whales using small boats and attempt to be close to them as they surfaced to breathe. That meant anticipating where the animal would surface, which is hard to do. Balanced precariously on a moving boat, researchers would use a long pole to stick a tag on the animal to monitor its movements or take a sample of its skin.

“Collecting biological data from a whale was incredibly frustrating. You’d race over to where you thought the animal would surface and much of the time it would come up somewhere else,” says whale biologist Iain Kerr of Ocean Alliance, a 50-year-old non-profit in the US. “I got to thinking – there had to be a better way. My hobby was flying, and crashing, model aircraft and radio-controlled helicopters at the time, and I wondered if these could be modified to do a better job. That was back in 2012, and after much trial and error... the rest is history.”

One of Kerr’s first field drones was SnotBot, in operation since 2015. A plethora of biological data can be gleaned from analysing the snot deposited on petri dishes mounted on the drone as it flies through the whale’s exhalation. “We can collect DNA, check for stress and pregnancy hormones, and assess the animal’s microbiomes – the living organisms that live inside the whale’s guts and lungs,” Kerr explains. “Later, we added dual cameras to accurately measure the whale using photogrammetry, the science of obtaining accurate measurements from images.” Most importantly, very few of the whales sampled reacted to the presence of the drone.

Whales live in a world of sound; some whale calls can travel extremely long distances through the water. Sounds identify the species and, in some cases, the specific individual. They can be used to find out how many whales are in a given area and have important social cues. With this knowledge, Kerr has also developed EarBot, a waterproof drone which can fly to an animal, land on the water surface nearby and switch off and listen in, recording the whale vocalisations and song.

Kerr hopes to expand on this work in the future by deploying a swarm of drones that could be deployed near a large group of whales and be able to assign the calls to the individual whale making the sound. “We want to listen in on a group of whales and be able to work out who is talking to whom – smart and affordable acoustic research,” he says. “One day we might even be able to work out what they are talking about!”

Using new technology with endangered species entails an ethical requirement to ensure that the animals are not stressed by the experience. Initially, behavioural biologists are usually on hand to review the pilot’s work. They can spot a stress response; for example, if an animal is regularly hunted by birds of prey, they might hide from a drone in the same way as they would an eagle or start to make alarm calls to alert others to perceived danger.

Adding any stress to the life of an animal on the verge of extinction is clearly not good. Any additional stress may impact on breeding success or its ability to feed. All uses of drones in conservation are subject to rigorous tests for stress responses first, but the success stories are so widespread that it is clear that they are here to stay.

World first

Researchers use drone to tag whales

Simultaneous biological sampling and tagging of a whale without disturbing the animal has long been a goal of biologists. Tags have been used for over 30 years, but most of the technology development went into the tags, and not the delivery systems. Working together, Iain Kerr at Ocean Alliance and research ecologist Dr David Whiley at NOAA decided to try and change that. To put what happened next into perspective – to tag 3 Sei whales over a 14-day period was considered ambitious, involved a very large budget, 11 people and three boats.

The two scientists and their team headed to the Sea of Cortez in the Gulf of Mexico in 2022 to test out a drone delivery tag system on a modified SnotBot, seeking to sample and tag the giants of the ocean – blue whales and fin whales, which were there to breed and feed. The results were extraordinary in many ways. But first the delivery system had to be perfected.

The tags attach to the whale’s skin using suction cups, so the team had to calculate how high the drone had to be above the swimming whale to create an impact force sufficient to compress the cups and how fast the drone had to be flying to securely attach a tag. Using ballistic gel to simulate the whale’s skin and a GoPro camera to film their deployments in slow motion helped the team to succeed in the lab. But they also had to develop a delivery system to release the tag from the drone. Ultimately, a radio-controlled release operated by the pilot was used.

Drones are launched after the whale takes its first breath when surfacing from a deep dive, which gives the drone enough time to be on the spot for the animal’s next breath, flying at a fixed altitude over the whale. The drone’s speed means that the scientists’ boat can be over 400m away from the animals, causing minimal disturbance. The whales barely register the presence of the drone.

This first ever test of drone-dropped whale tagging was published in April 2023 and on this first tagging expedition, the team tagged 21 whales in eight days, from a single 8m boat, with most tags being deployed in under five minutes. These are unheard-of success rates, according to Kerr.

“This opens up the possibility of tagging every member in a family group of these highly social animals and will allow us to learn much more about their lives together under the surface,” he says. The future of drone-based whale tagging looks assured.

Sign up to the E&T News e-mail to get great stories like this delivered to your inbox every day.

Recent articles