Analyzing the impact of three-dimensional building structure on CO2 emissions based on random forest regression

Abstract

Carbon dioxide (CO2) is the primary greenhouse gas that increasingly threatens environmental conditions and public health. In addition to conventional socio-economic mitigation measures, a healthy urban design can substantially contribute to the reduction of CO2 emissions. Nevertheless, previous attempts only concentrated on the impacts of horizontal landscape pattern and spatial structure on CO2 emissions. The relationship between three-dimensional building structure and CO2 emissions remains to be explored. To fill this knowledge gap, our study analyzed which building indicators matter most to CO2 emissions in high-density areas. First, we discovered the linear relationships between CO2 emissions and various potential spatial drivers based on Pearson correlation test. Second, we examined whether the additional consideration of different building-related indicators can better explain the variation in CO2 emissions using random forest regression. These experiments indicated that building coverage ratio, mean building number, spatial congestion degree, and floor area ratio can exert substantial impacts on CO2 emissions in the study area. Building structure is a key factor affecting CO2 emission volumes. For example, our improved model yields a lower root relative squared error (32.53%) than the benchmark model (34.68%). This methodological framework, which can be easily applied to any other regions, is expected to provide valuable information for the reduction of CO2 emissions from the perspective of vertical urban planning. Policy-makers should carefully consider the impact of building structure on CO2 emissions at an earlier stage of the healthy urban design.