With the intensification of global warming, the issue of carbon emissions causes more and more attention in recent years. In this paper. China's 30 provincial-level administrative units are divided into five emission regions according to the annual average value of provincial CO2 emissions per capita during 1990 and 2010. The regional differences in impact factors on CO2 emissions are discussed using STIRPAT (stochastic impacts by regression on population, affluence, and technology) model. The results indicate that although GDP (Gross domestic product) per capita, industrial structure, population, urbanization and technology level have different impacts on CO2 emissions in different emission regions, they are almost always the main factors in all emission regions. In most emission regions, urbanization and GDP per capita has a bigger impact on CO2 emissions than other factors. Improving technology level produces a small reduction in CO2 emissions in most emission regions, but it is still a primary way for CO2 reduction in China. It's noteworthy that industrial structure isn't the main factor and improving technology level increases CO2 emissions in high emission region. Different measures should be adopted for CO2 reductions according to local conditions in different regions. (c) 2012 Elsevier Ltd. All rights reserved.