TJ5.1 The Uncertainty of Gridded CO2 Emissions Data

Monday, 11 January 2016: 1:30 PM
Room 356 ( New Orleans Ernest N. Morial Convention Center)
Eric Marland, Appalachian State University, Boone, NC; and S. Hogue, R. J. Andres, and G. Marland

With numerous demands for spatially explicit estimates of CO2 emissions we need now to have estimates of the uncertainty that accompanies geographically gridded emissions data. We recognize 6 discrete elements of uncertainty and develop an approach for addressing each of them individually. We focus on the U.S. as a test case because of the quality and ready availability of appropriate data. 1.) Because the best data on fuel use are at the national level, most geographically explicit emissions datasets start by estimating national totals of emissions and then distributing these national totals within each country. There is uncertainty in the national total. 2.) Large point sources of emissions account for 30-50% of emissions in many high-emitting countries and available data allow us to separately analyze these emissions values. 3.) Data sets on large point sources contain, for a variety of reasons, spatial uncertainty on where the emissions are actually discharged. 4 and 5.) For emissions beyond large point sources we generally rely on proxy data, such as population density and/or satellite-observed night lights, to estimate the spatial distribution of emissions. There is uncertainty in both the magnitude and location of the proxy values. 6.) There is uncertainty on the extent to which proxy values truly represent CO2 emission. A combination of these six elements comprises the total uncertainty in the emissions figures.

For the U.S. at 1 degree latitude/longitude resolution, using a population proxy for areal sources of emissions, uncertainty associated with large point sources dominates in cells with the highest emissions, proxy uncertainty dominates in cells with lower total emissions, and uncertainty in the population data becomes important in only a small number of cells. Having completed a preliminary assessment of the uncertainty for the U.S. we start to illustrate the requirements for both understanding and reducing uncertainty: 1.) for a dataset that relies heavily on proxy data, we need some way to, at least periodically, calibrate the proxy data against real emissions data, 2.) data on large point sources are very important both in their magnitude of emissions and the accuracy of their locational reporting, 3.) proxy uncertainty is large but can be reduced by greater availability of discrete data for large point sources to remove them from the areal totals, and 4.) the magnitude of uncertainty is scale dependent and the uncertainty of discrete cells is not independent of the emissions and uncertainty in nearby cells.

- Indicates paper has been withdrawn from meeting
- Indicates an Award Winner