Dynamic scenario analysis and prediction of embodied carbon emissions in China's building sector: A hybrid interpretable machine learning model

Abstract

With the rapid urbanization in China, excessive carbon emission has emerged as a primary constraint in sustainable development. Notably, the building sector accounts for a substantial portion of carbon emissions, with embodied carbon emissions particularly prominent. Under the strategies of carbon peaking by 2030 and neutrality by 2060, this study focuses on embodied carbon emissions in China's building sector. Utilizing a hybrid interpretable machine learning model, this study endeavors to predict national and provincial carbon peaking times and emission values through dynamic scenario analysis. The findings offer unique insights: First, based on the materials and resources consumed by the building sector, this study reveals a significant probability of China's embodied carbon emission peaking in 2030 at 6.58 billion tons. Second, the interpretability analysis of the model underscores the profound impacts of economic and demographic factors on national and provincial embodied carbon emissions, collectively explaining over 50 % of its variability. Third, regional heterogeneity analysis reveals the reasons for the rapid carbon peaking in coastal and developed provinces due to the varied developing conditions. These results present a novel perspective on carbon reduction and sustainable development, offering crucial guidance for global carbon mitigation efforts and specific policy recommendations.