Understanding the adoption of agri-environment practices in Europe through explainable artificial intelligence

Agri-environment practices (AEPs) are vital for sustainable agriculture in Europe, yet adoption patterns vary considerably across regions and farm types. This study employs Categorical Boosting machine learning models and eXplainable Artificial Intelligence to analyze AEP adoption across 6,947 farms in Germany, Czechia, and the United Kingdom using data from the Integrated Administration and Control System. We examine associations between farm characteristics, environmental conditions, and neighborhood effects with both the decision to adopt AEPs and the area dedicated to them. Results reveal that farm size and economic capacity are the strongest predictors, with larger farms implementing more AEP types across greater absolute areas, while smaller farms allocate higher proportions of their land. Environmental conditions, particularly temperature and soil quality, and practice-specific neighborhood effects also shape adoption patterns. These findings call for agri-environment policies that differentiate support by farm size, align measures with local environmental conditions, and account for practice-specific neighborhood effects.