A common value of carbon emission factor, t-CO2/GWh, in nuclear power generation reported in the literature varies by more than a factor of 100. Such a variation suggests a margin of uncertainty and reliability. In this study, we employ a bottom-up approach to better define the system, its input and output, and boundaries. This approach offers improved granularity at the process level and consistency in the results. Based on this approach, we have developed a methodology to enable comparison of carbon emissions from nuclear power generation. The proposed methodology employs the principle of energy balance on a defined power generation system. The resulting system boundary facilitates the use of the "Kaya Identity" and the decomposition technique to identify the carbon emission streams. Using nuclear power as a case study, we obtained a carbon emission factor of 22.80 t-CO2/GWh, which falls to within 2.5% of the median of globally reported LCA results. We demonstrate that the resulting methodology could be used as a generic tool for life cycle analysis of carbon emissions from other power generation technologies and systems. (C) 2013 Elsevier Ltd. All rights reserved.