Research Information System
Ecological Processes, vol.15, no.1, 2026 (SCI-Expanded, Scopus)
Background: Climate change is reshaping key ecological processes such as drought cycles, heat stress, disturbance regimes, and air quality regulation. These changes generate cascading impacts on ecosystems and the human societies that depend on them. There is therefore a pressing need for frameworks that not only map where adaptation is required but also explain why these needs arise. Existing approaches often remain limited: top-down climate scenarios overlook local stressors, multi-criteria decision-making (MCDM) applications suffer from subjectivity, and machine learning models frequently lack interpretability. To address these limitations, this study introduces an integrated framework that combines ecological, climatic, and socio-economic indicators with MCDM, Geographic Artificial Intelligence (GeoAI), and explainable machine learning to provide spatial mapping and a sensitivity-based interpretation of adaptation needs. Results: A Climate Change Adaptation Index (CCAI) was constructed for 1,187 NUTS-3 regions using 44 indicators of vulnerability and resilience, including land surface temperature, burned areas, hydrological stress, demographic sensitivity, renewable energy use, and institutional effectiveness. Fuzzy Analytical Hierarchy Process (F-AHP) was employed to derive factor weights, and SHapley Additive exPlanations (SHAP) was applied to analyse indicator interactions. The findings highlight three major discoveries: (i) heat stress, precipitation, and renewable energy share emerge as the dominant ecological and socio-economic contributors to variation in the constructed index; (ii) Southern and Mediterranean Europe exhibit significantly higher adaptation requirements than Northern Europe, reflecting marked spatial disparities; and (iii) SHAP-based explainability uncovers regional heterogeneity by showing how indicator effects vary across contexts, thereby extending expert-based weighting results with data-driven sensitivity evidence within the constructed index. Conclusions: By linking ecological processes with socio-economic conditions and supporting the analysis with interpretable machine learning, this study advances an interpretable understanding of index-based regional climate adaptation needs. The framework combines MCDM, GeoAI, and SHAP to provide transparent evidence of both global and local contributors to variation in the constructed adaptation-needs. Moving beyond descriptive spatial distributions, the framework clarifies how indicator contributions and their combinations shape the constructed CCAI scores and associated adaptation-needs patterns, enhances transparency, and provides actionable insights for policymakers. The approach is scalable and policy-relevant, supporting targeted strategies to strengthen Europe’s climate resilience.