Spatial Patterns of Energy-Related Carbon Emissions from Residential Land: A Hybrid Physics-Machine Learning Study of Shenzhen

Accurate estimation of residential building energy consumption and energy-related carbon emissions is essential for supporting urban low-carbon development. This study proposes a hybrid modelling framework that integrates physics-based simulation and machine learning to estimate residential building energy use in Shenzhen for 2020. Representative building archetypes are simulated, and the results are used to train machine learning models for large-scale application. A bottom-up inventory combined with spatial-proxy-based downscaling is further employed to generate high spatiotemporal resolution maps of carbon emissions. Results show that the model achieves strong estimation performance across multiple scales. Daily mean temperature is the dominant driver of energy-use variability, while building type significantly influences consumption levels. Residential energy use is generally higher on weekends than weekdays. Spatially, emissions are concentrated in central and western districts, with Longgang having the highest emissions (11.19 Mt), followed by Bao’an, Longhua, and Nanshan. High-emission buildings are mainly located along road-adjacent areas. The proposed framework provides a robust and scalable approach for fine-resolution CO₂ emission estimation and supports accurate emission source attribution for urban carbon management.