Stavros Veresoglou

This article opens a new era of AI-based supervised classification of vegetation and habitat types based on species composition. César Leblanc from the University of Montpellier, with the support of colleagues from Pl@ntNet and EVA (European Vegetation Archive), has developed a classifier for EUNIS habitat types across Europe that seems to classify habitat types better than the logical definitions in the expert system. I was already playing with using artificial intelligence tools to classify vegetation types based on vegetation plots 20 years ago. My student Lenka Černá and I published the first study using artificial neural networks for the supervised classification of vegetation in 2005 (doi.org/10.1111/j.1654-1103.…). At that time, however, AI tools were still in their infancy, the calculations were time-consuming, there were no continental vegetation databases and no standardized international vegetation or habitat classification. Therefore, we focused on developing logical definitions of vegetation and habitat types, which was completed for the European habitats of the EUNIS system in 2020 ( doi.org/10.1111/avsc.12519). In our expert system EUNIS-ESy, we created consistent definitions of habitats across Europe, which paved the way for the application of more flexible AI tools such as those developed under César's leadership. Open-access article: doi.org/10.1111/avsc.12802 Code on GitHub: github.com/cesar-leblanc/hdm…