Carbon emission estimation at the urban functional zone scale: Integrating multi-source data and machine learning approach

Abstract

With the acceleration of urbanization, urban carbon emissions have become a major contributor to global warming. Although the spatial distribution of carbon emissions has been studied, more accurate fine-scale results are needed. This study utilized Open Street Map (OSM) and point-of-interest (POI) data to identify urban functional zones. Carbon emissions of sample zones were calculated from electricity and natural gas data as target variables. The identified functional zones were combined with multi-source remote sensing data as features to construct a machine learning-based carbon emission model, predicting the spatial distribution of emissions at the urban functional zone scale. The study area was divided into 3,861 functional zones with total emissions of 51,124,900 tons, where residential, commercial, and industrial zones were the main sources. Multifunctional urban green space plazas also contributed significantly to emissions, a factor often overlooked in previous studies. Compared to conventional methods relying on nighttime light data, multi-source data significantly improved accuracy and spatial resolution, especially in industrial and green space zones. This study confirms that analyzing carbon emissions based on urban functional zones is effective and supports low-carbon city construction and management.