Quantifying heterogeneous impacts of 2D/3D built environment on carbon emissions across urban functional zones: A case study in Beijing, China

•Multi-scale semantic segmentation deep model for urban-function-zone mapping.•Estimating function-zone carbon emissions by combining nighttime light and statistical data.•Exploring the differences in carbon emissions between different types of function-zones.•Quantifying heterogeneous impacts of the 2D/3D built environment on carbon emissions across function-zones. Rapid urbanization has led to a significant increase in carbon emissions (CES), adversely affecting the urban ecological environment. As fundamental units in urban planning and management, urban functional zones (UFZs) exhibit similar energy consumption patterns. Previous studies on urban CES have often focused on large-scale regions, lacking research on CES differentials at the scale of UFZs. This study, using Beijing as a case study, employed multi-source geospatial data to estimate CES and analyze influencing factors at the UFZ scale. First, a multi-scale semantic segmentation network was constructed to produce UFZ map using high-resolution remote sensing images and OpenStreetMap road data. Second, a top-down approach was employed, incorporating nighttime light data to calculate the UFZ-scale CES to analyze differences in CES among various UFZs. Finally, factors influencing urban CES, including socioeconomic factors and two-dimensional (2D) and three-dimensional (3D) urban built environment, were analyzed using Pearson analysis and GeoDetector analysis. Results indicate that: 1) Among the 2813 UFZs in Beijing, the majority belong to residential zones (34.23 %), followed by institutional zones (20.33 %); 2) The total CES in study area amount to 4.07 × 107 tons. Commercial zones contribute significantly higher CES than other UFZs, followed by residential zones. 3) For different types of UFZs, the impact of various factors on CES differs significantly. Population density, the highest building height, base area, and average building volume are the main correlation factors. Spatial analysis of CES at the UFZ scale contributes to the management of low-carbon cities, energy allocation, and optimization of urban planning implementation.