Unravelling food carbon footprint heterogeneity in metropolitan areas using Tokyo as a case study

As cities gear up toward carbon neutrality, the food sector can play a crucial role in decarbonization. Food related carbon emissions likely vary across urban areas due to the interplay of urban form, food environments, and dietary habits, affecting the intensity of emissions. Existing consumption-based carbon accounting methods fail to capture spatial heterogeneity effectively and have not fully explored opportunities to enhance spatial resolution in urban contexts. This study proposes a novel Service Point-Based Carbon Accounting (SPBCA) method to systematically understand how these factors influence CO₂ emissions in urban food systems. Unlike traditional approaches, SPBCA focuses on meal provision points rather than consumption locations, allowing for more accurate spatial representation of emissions. We applied SPBCA to census tracts in the Tokyo metropolitan region and validated its effectiveness using LightGBM, an advanced machine learning approach. The model achieved a high validation accuracy (R² = 0.874) through cross-validation, demonstrating SPBCA's capability to capture the heterogeneous nature of urban food-related emissions. This method enables identification of key actors in urban food systems, important for developing effective decarbonization roadmaps for climate policy and urban planning at the urban neighborhood scale.