This paper proposed a global optimization method integrating fuzzy clustering and Kriging to improve the accuracy of air pollution forecasting, addressing the limitations of conventional approaches that often overlook spatial heterogeneity and local variations in pollution patterns. The proposed method first partitions input data into clusters with similar characteristics using fuzzy clustering, which captures the inherent uncertainty and gradational boundaries in environmental data. Kriging is then applied within each cluster to model spatial dependencies and predict pollution levels, ensuring localized precision while maintaining statistical robustness. A global optimization framework is employed to jointly optimize the parameters of both clustering and Kriging models, minimizing prediction errors through an iterative refinement process. The novelty of our approach lies in its hybrid structure, which combines the flexibility of fuzzy clustering to handle ambiguous data groupings with the geostatistical rigor of Kriging for spatial interpolation. Experimental results demonstrate significant improvements in forecasting accuracy compared to standalone methods, particularly in regions with complex pollution dynamics. This work contributes to environmental science by providing a scalable and adaptive tool for policymakers and researchers to monitor and mitigate air pollution more effectively. The method’s generalizability also suggests potential applications in other spatially correlated phenomena beyond air quality assessment.