Understanding factors influencing carbon emissions is important for achieving carbon abatement goals. Traditionally, decomposition approaches, e.g. Logarithmic Mean Divisia Index (LMDI) method, are used to uncover factors influencing the change of carbon emissions. This study aims to develop a novel spatio-temporally stratified heterogeneity (STSH) model to better identify, analyze and understand influencing factors of carbon emissions. Compared with LMDI, this model can analyze more independent intensity and quantity factors and reveal the complex interactions between factors. Using China’s construction industry as a case study, the model can successfully identify top influencing factors in the same order of importance similar to LMDI, although a significantly larger number of factors are considered. In the whole construction stage, cement usage, steel usage, completed floor area and construction output value have the highest contributions of 0.599, 0.528, 0.448, 0.446 respectively. In construction and construction-related transportation activities, the top influencing factors are fixed capital assets, construction output value, completed floor area, and electricity usage, with contributions of 0.632, 0.613, 0.599, and 0.597, respectively. The model can also reveal the complex interactions between factors, including bi-variate enhanced interaction and nonlinearly enhanced interaction. The results demonstrate that the model is more useful to evaluate the individual and aggregate impact of a large number of independent factors on carbon emissions when compared with previous models.