Health in human-animal interactions
Making sense of mosquitoes with machine learning
How do bumblebees land on a flower? Does LED lighting in their greenhouse affect their pollination activities? How can you best combat the deadly malaria mosquito? Smart software can be a useful tool in answering such questions. Professor Florian Muijres and his team are teaching computers to recognise the flight movements of insects and birds. They hope this will help them to better understand animal behaviour.
Florian Muijres, professor of Experimental Zoology, is quite open about his admiration for insects. “They have small brains with relatively little processing power, yet they are capable of exhibiting complex behaviour. Think of mosquitoes that avoid fly swats, and bumblebees that land precisely on a moving flower. Insects do this perfectly.”
Muijres’ admiration for the behaviour of insects is closely connected to his expertise in technology. He studied Aviation and Aerospace Technology at Delft University of Technology and now he works at WUR, where he’s figuring out how animals that fly function in a changing natural environment with human influences. “I have always watched birds as a hobby and was particularly interested in watching them fly. It’s really cool to finally understand how that works.”
Muijres and his team investigate that by teaching the computer to recognise the movements of animals. For example, how much does LED lighting hinder pollination by bumblebees in a greenhouse? And how can you combat the deadliest animal on the planet, the malaria mosquito? Technological advances are making his research ever more efficient and widely applicable, from one bumblebee in the lab to a thousand mosquitoes in the air.
Muijres gives an example to illustrate the changes in his field of research. “When I graduated in 2006 with a study of the flight behaviour of bats, it took us a week to create a one-second video, manually digitising all the camera images from three points on the bat’s body and processing them to create a 3D reconstruction. All the algorithms for analysing the wing positions during a wingbeat were in our own heads, as it were.”
Since then, cameras have become much faster: these days, you can record how a mosquito in flight flaps its wings 600 times a second in 13,500 frames per second. The result is terabytes of data to analyse. But the most important development has been the rise of deep learning algorithms in which an artificial neural network, analogous to our brains, is taught to perform complex tasks. That kind of network can recognise images automatically from large volumes of data.
Muijres uses machine learning to perform precise analyses of complex flight movements. The knowledge this produces can be used to design targeted measures that benefit society, for example by preventing diseases and making food production more sustainable.
‘We install cameras on the house that let us track up to a million mosquito flight movements’
Muijres studies creatures that are important to society, such as the bumblebee. Bumblebees are key pollinators both in nature and in sustainable forms of agriculture and horticulture. That is why it is useful to know how bumblebees forage and how they deal with changing circumstances, such as too little light, or cold or windy conditions.
Over the past few years, growers have been using red and blue LED lighting to boost plant growth, but bumblebees’ eyes are more sensitive to green light. Muijres’ team investigated whether the insects find the LED lighting a problem when landing on flowers and if this makes their pollination less efficient.
First, the researchers observed the bees in the lab. “We looked at how bumblebees land on a flower in a static situation. It turns out they often slow down, or even hover, and then speed up again.” That lets them land quickly and in the intended spot. Once they had a picture of the standard pattern, the researchers added side wind and a flower in motion.
The bees also appeared to be able to adapt well to the LED lighting: the landing manoeuvre as a whole took just as long with both light colours. “Bumblebees aren’t just cute creatures, they’re also tireless and robust. They can fly well in difficult conditions and they can forage for food relatively quickly thanks to their efficient landing pattern.”
Muijres’ group used the same real-time tracking as they had for the bumblebees in a study of the flight behaviour of malaria mosquitoes. In Africa in particular, this insect poses a major threat to public health because it transmits the deadly malaria parasite. But if you know how a malaria mosquito enters the home, you can take measures to prevent that. In the past, researchers would set mosquito traps inside and outside the home. Then they would seal a ventilation grille, for example, and compare the number of mosquitoes in the two traps. But that cost a lot of time.
‘It is a complex dynamic between human and insect, a kind of intricate cat-and-mouse game’
Muijres: “Now we install cameras on the house that let us track up to a million flight movements. Algorithms use all that data to give an accurate picture of how mosquitoes behave around a trap and how they enter the house. Do they fly straight through the ventilation grille or the window, or do they fly just above the ground first? This will tell you rather quickly where best to locate the mosquito trap. So by quantifying things more precisely, you can take much more effective measures.” His research has given him quite some respect for these insects. “They’re incredibly good at tracking down humans. The smell of the carbon dioxide we exhale is an important enticement.”
His team also discovered that malaria mosquitoes use a special avoidance manoeuvre. “When you try to slap them, they detect the air movement produced by the hand and use that as a kind of bow wave to surf away on,” explains Muijres. “It is a complex dynamic between human and insect, a kind of intricate cat-and-mouse game.” “Perhaps one day we will be able to track that behaviour and see close up how they interact.” Experiments have been carried out with transgene mosquitoes where the males are sterile.
The trials with sterile mosquitoes are aimed at reducing the number of females as they are the mosquitoes that bite and thereby transmit malaria. “But the transgene males still need to be sufficiently fit to mate in the swarm. We want to see how they move, so we are collaborating internationally on this.”
Muijres is also keen to study bumblebees in their natural habitat. What is more, the bumblebee is being used as a model for a drone that Muijres is developing in partnership with Delft University of Technology. A drone that can fly and land autonomously would be useful in civil aviation, for example in drone ambulances. “We want to programme the landing behaviour of bumblebees in drones to see whether that lets them make precise landing manoeuvres independently. This is quite the challenges, indoors and without GPS.” It only increases his admiration for these tiny insects. “It’s so impressive. When you see bumblebees out there in nature, it’s as if it never goes wrong.”
His background as an aerospace engineer is evident when he describes these creatures as a black box. “You don’t know what they get up to in woods, in a house or in a greenhouse. These new techniques let us get closer to answering the key questions and opening up that black box.”
Share this article
This project is part of the Next Level Animal Sciences (NLAS) innovation programme.
Participating researchers of Wageningen University & Research collaborate with various partners to develop new research methods and technologies within the field of animal sciences. NLAS consists of three research directions, namely sensor technology, complex cell systems and data and models.