Predicting the multilevel biodiversity index through machine learning methods: A case study

Biodiversity is pivotal in sustaining ecological processes and ensuring ecosystem robustness, underpinning vital services like pollination, water purification, and climate regulation. Nevertheless, its swift decline, largely due to human-induced factors such as climate change, land-use alter- ation, pollution, and invasive species, has triggered significant global concern. The biodiversity reduction has repercussions beyond environmental deterioration, posing substantial hurdles to achieving international sustainability goals, particularly the United Nations Sustainable Development Goals (SDGs), including Goal 13 (Climate Action) and Goal 15 (Life on Land). Several machine learning approaches are compared, including random forests (RF), support vector machines (SVM), extreme gradient boosting (XGBoost), decision trees (DT), and linear regression (LR), to predict the multilevel biodiversity index (MBI) in the South of Italy, based on PM10, temperature, pressure, and precipitation. The results show a significant decrease in biodiversity in areas with high PM10 concentrations, especially agricultural and industrial zones, highlighting the negative impact of air pollution on ecological systems. Among the above-mentioned approaches, the best performance is obtained with DT and XGBoost, reliably predicting MBI based on various evaluation metrics. These findings provide valuable insights for policymakers and contribute to mitigating human-driven ecological impacts.