Drones deployed in the hunt for bark beetles

What does BFH hope to achieve with the research project on drone and satellite images of trees?

For decades, bark beetles have brought devastation to trees and forestry. They swarm out several times a year, burrow into the bark of the trees, make their nests there and reproduce. An infested tree may still look outwardly healthy at first, but it is actually already dying because the beetles are preventing it from transporting water and nutrients.

The aim of the research project is to recognise trees infested by bark beetles at an early stage and to show foresters where they are located. This can be done more efficiently with drone and satellite images than if the experts have to inspect the forests on foot. If diseased trees can be felled before the beetles fly out and infest other trees, the spread of the pest can be contained and the forest better protected.

What approach did the researchers take?

In an experimental forest in Entlebuch, the researchers used drones to take images in several different bandwidths. They then analysed the images based on their reflections. They began with a static algorithm, and later tried algorithms that self-improve through machine learning. They found that the difference between the data from healthy and infested trees was often too small to be able to draw a clear distinction.

The researchers are currently trying to work out why the data from both should be so similar. If they manage to eliminate as much unwanted information as possible from the data, the trained algorithms will deliver significantly better results.

How is it possible for an algorithm to learn which trees are infested with bark beetles based on imagery?

The algorithm searches for specific signals in the multispectral bands of the drone and satellite imagery. The signals from diseased and healthy trees differ from each other. The researchers are training the algorithm to discern this difference and make an accurate statement as to whether a tree is infested with bark beetles or not.

What challenges does the project have to overcome?

To make the algorithm more reliable, it needs a sufficient quantity of high-quality data. The challenge lies in deriving this data from the drone and satellite images. The researchers apply various algorithms here, which are also based on machine learning. By learning to recognise patterns and correlations from data, a system is able to enhance its own performance.

How does the project benefit society?

Bark beetles cause immense damage to forests. According to an estimate by the Swiss Federal Institute for Forest, Snow and Landscape Research WSL, these pests destroyed over 700,000 cubic metres of wood in 2023.

If it is possible to reduce the extent of the destruction, this will bring much relief to the forests as ecosystems, but also to their owners, who suffer in many cases substantial economic losses as a result of the destruction. Healthy forests are not only important because of their protective function against rockfall and avalanches, they are also very important for sustainable development. Wood is a climate-neutral raw material, and trees absorb vast amounts of CO₂.

When will the method be available for use in forests?

Further research is needed to refine the method using drone and satellite imagery and to enhance the precision of the calculations. The researchers anticipate that they will be able to offer a method with a sufficiently robust algorithm within a few years. “Sufficient” means that the system makes an accurate statement in just under 70 per cent of cases as to whether bark beetles have nested in a tree. A 100 percent hit rate will not be possible with this method.