Gridded estimates of CO2 emissions: uncertainty as a function of grid size

• DOI: 10.1007/s11027-017-9770-z
• Publication Date: 2017-12-07T23:09:27Z
• Authors: S. Hogue; D. Roten; E. Marland; G. Marland; T. A. Boden
• Keywords: Proxy (statistics); Grid cell; Single point; Point estimation

A crucial aspect of constructing a gridded model of anthropogenic fossil fuel CO2 (FFCO2) emissions involves careful consideration of uncertainty. Both the spatial resolution of the emissions estimates (grid scale) and the selection of proxy data to represent the spatial distribution of emissions, plus the quality of data on point sources of emissions, have important impacts on uncertainty. In earlier papers, we explored the uncertainties associated with grid selection and the available data on large point sources. In this work CO2 emissions data are spatially distributed using population density as the selected proxy, using three different treatments of large point sources, and with five levels of grid resolution (1o, 2o, 3o, 4o, and 5o). The methods of calculating uncertainty associated with grid size, proxy selection, and reported point-source emissions data are presented, with particular attention being drawn to grid size selection. We find that as the resolution becomes coarser, relative uncertainty (total uncertainty as a percentage of total emissions) at the grid cell level decreases. Relative uncertainty in most grid cells decreases as the portion of emissions attributed to specific point sources increases. Good data on large point sources is very important for spatially explicit emissions inventories.