CLIMATE CHANGE IMPACT ON AGRICULTURAL LAND SUITABILITY: AN INTERPRETABLE MACHINE LEARNINGBASED EURASIA CASE STUDY

Abstract

Climate change is significantly altering temperature patterns, precipitation regimes, and the frequency of extreme weather events, thereby affecting the suitability of agricultural land across diverse regions. This study presents an interpretable machine learning-based framework to assess and predict the impact of climate change on agricultural land suitability across Eurasia. By integrating multi-source datasets including climatic variables, soil characteristics, topography, and land-use patterns, the proposed model leverages advanced algorithms such as Random Forest and feature selection techniques to identify key factors influencing land suitability. Unlike conventional black-box models, this approach emphasizes interpretability using techniques such as SHAP (Shapley Additive Explanations) to provide insights into feature importance and decision-making processes. The model is trained and validated on historical and projected climate data to evaluate shifts in suitability patterns over time, enabling the identification of vulnerable regions and potential areas for agricultural expansion or adaptation. The results demonstrate that climate change has heterogeneous effects across Eurasia, with certain regions experiencing declining suitability due to increased temperature stress and water scarcity, while others may benefit from extended growing seasons. The proposed framework not only enhances predictive accuracy but also supports policymakers and stakeholders in making informed decisions for sustainable agricultural planning, climate adaptation strategies, and food security management in the face of ongoing environmental changes.