On the initiative of biologists Diederik van Liere (wolf behavior scientist) and Nynke Osinga, a group of Data Scientists from Cmotions set to work developing a machine learning (AI) model that can not only predict where wolves will settle, but can also explain what factors play a role in this. Applying these Data Science techniques in this way and for this purpose is a novelty in biology/ecology that can be incredibly valuable.

The results of this prediction model, using XGBoost and SHAP, are presented in a scientific article which was published in Environmental Management magazine.

The main conclusion of the scientific prediction model is that wolves are more likely to settle in rural areas. Previously, wolves were thought to settle only in vast natural areas far from populated areas. Now it appears that they prefer to mark their territory close to other packs, even if it is near an inhabited area. "An area with up to 40 percent forest is already suitable," van Liere said. "Cropland is also attractive because that is where deer, roe deer and smaller game species live that the wolf hunts."

The wolf prediction model is an important tool for taking timely measures when wolves enter built-up areas. "Wolves are no longer leaving the Netherlands, which is why we need to learn how to live with them without incidents," van Liere said. "Wolves, like dogs, have an excellent learning ability. Only we can teach them to stay away from people and livestock. If a wolf walks through the built-up area, it must be chased away immediately in a professional manner so that they associate people with danger. Because when wolves become familiar with people, their cubs will copy that behavior. In West Germany, but also in the Netherlands, we unfortunately already see this happening."

We are proud of what we have collectively accomplished in this project. Nice to see how something that starts small can grow into a scientific article. An incredibly inspiring project!