Diagnosing Environmental Drivers of Carbon Monoxide Concentration in Kaduna Metropolis Using Interpretable Machine Learning

Reliable spatial information on urban air pollution is essential for effective environmental management and sustainable urban development. This study modelled and interpreted the spatial distribution of carbon monoxide (CO) concentration in Kaduna Metropolis, Nigeria, using remote sensing, geospatial analysis, Random Forest regression and SHapley Additive exPlanations (SHAP). Sentinel-5P CO data were analysed alongside land surface temperature, NDVI, NDWI, NDBI, NDBaI and proximity to roads, infrastructure and industrial sites. A total of 10,000 valid samples were used for model development, with spatial cross-validation and independent testing applied to assess predictive performance. The model demonstrated good spatial generalisation. SHAP analysis showed that anthropogenic variables were more influential than land-surface factors, with road proximity emerging as the dominant predictor, followed by industrial and infrastructure proximity. Together, these anthropogenic variables accounted for 68.01% of total model importance. The novelty of the study lies in extending interpretable machine-learning assessment of urban air pollution in Kaduna to CO and quantifying the relative contributions of environmental and anthropogenic spatial drivers. The findings provide a basis for prioritising major transport corridors, industrial areas and infrastructure- intensive locations for targeted CO monitoring, traffic-emission control and industrial emission management. Unpacking the Constraints to Sustainable Rural Community Development: Perspective from Indigenous Settlement in South Africa Olusegun Oguntona1*, Omokolade Akinsomi2, Magnus Andersson3, Chijioke Emere1, Andreas Lundin4 1 cidb-Walter Sisulu Centre of Excellence, Department of Built Environment, Faculty of Engineering,