Artificial intelligence against the bark beetle infestation

Trees respond to stress by emitting volatile organic compounds known as VOCs (Volatile Organic Compounds). Bark beetles also communicate via volatile messengers called pheromones. Trained dogs can smell them and thus indicate a bark beetle infestation on spruce trees. This is very strenuous for the animals, they tire quickly. “This is where our idea came from: we are developing a digital nose, i.e. artificial intelligence (AI), that can recognize trees and sniff out stress chemicals and pheromones,” explains Claudia Probst.

The FH professor and head of the Agricultural Technology and Management course on the Wels campus of the FH Upper Austria is a specialist in phytomedicine. It mainly treats plant diseases. Together with the computer scientist Georg Roman Schneider, who teaches the course, the project “DigiWald” was launched. A forest expert and two chemists from bio- and environmental technology complete the university team.

AI must detect diseased trees early
There is already artificial intelligence that can recognize different types of trees using image processing. What is new, however, is that this is combined with an olfactory recognition. The aim is to detect tree diseases in the forest as early as possible. The FH team is breaking new ground with its basic research. This is supported by funds from the forest fund, an initiative of the Federal Ministry of Agriculture, Forestry, Regions and Water Management. The term is three years.

Tree stressors can include drought, extreme heat, fungal infestation, deer hunting, or too many stocks, for example. “But for starters, we’ll tackle the problem of bark beetles in spruce trees,” says Probst. For this purpose, the behavior of the bark beetles is studied. Their pheromones, like the messenger substances of the trees, are chemically and biologically analyzed and scientifically processed. The collected data is then used to develop the neural network (algorithms based on how the human brain works) and artificial intelligence. The research team coordinates this practically with forestry. The federal forests, the Gmunden district administration and nature conservation organizations are present as cooperation partners.

Looking for sensors
The development of the technological side is in the hands of Schneider. “We are investigating which available sensors for gaseous substances can be used to detect these subtle changes in the air,” says Schneider. A technological laboratory is being set up in the Institute for Sustainable Robotics on the Wels campus. Here, both the computer-aided artificial intelligence and the digital nose must be trained. Ultimately, for example, the system should be able to recognize a tree as spruce, recognize a hole in it as an entrance for the bark beetle, and detect pheromones – all indications that this spruce was actively infested by bark beetles. With the information, it would be easier for forestry companies to remove an affected tree at an early stage to protect the remaining forest stand.

Until then, the team has a lot of research work to do. For example, in experiments it is important to find out whether it is enough to hold a measuring device in your hands so that the AI ​​can detect the pheromones, or whether it makes sense to fly a drone over the treetops to detect relevant measurement data. “We will gain important insights over the next three years. However, it will take another three to five years before the digital nose is used in practice,” Probst is convinced.The project is financed by the Forest Fund of the Federal Ministry of Agriculture, Forestry, Regions and Water Management.