Carbon Emissions Inventories and Energy Policy Research

Research Team

Project Description

The emission of carbon dioxide (CO2) from fossil-fuel combustion is the principal factor responsible for human-driven changes in the global climate system. Consequently, there is considerable interest in human-related sources and sinks of CO2 and a new global agreement to try to limit these net emissions. For both geochemical and geopolitical reasons, there is widespread interest in the spatial distribution of CO2 sources and sinks. Knowledge gaps occur in both discrimination between humand and "natural" emissions and in attribution of emissions to specific political entities or specific activities.

In efforts to limit emissions, it is necessary to be able to evaluate both responsibility for emissions and successes in mitigating emissions. Although there are multiple sources for estimates of emissions at the national level, and several attempts to estimate emissions at finer scales (e.g. within citits or with geographically specified spatial grids), there is considerable uncertaintly in all of these numeric estimates. There has been little previous success in quantitative evaluation of the uncertainties or in addressing how to deal with uncertainty for spatially-distributed emissions data. In previous work, we have described the differences displayed in multiple analyses of the distribution of emissions (Hutchins et al., 2016) and have developed a new statistics function (PSUM - Point Spatial Uncertainty Measure) that permits us to quantitatively describe the uncertainty in gridded emissions estimates (Woodard et al., 2015).

Current gridded inventories of emissions from fossil-fuel use and industrial processes rely heavilty on related, proxy and re-purposed data. Rayner et al. (2010) noted that "none of the pointwise fossil emission products available today include estimates of uncertainty." They also point out, importantly, that uncertainties for nearby grid spaces are not independent because, for example, the uncertainty for any given grid space includes consideration that a large point source might be slightly displaced and thus incorrectly located in an adjacent grid cell. Rayner et al. emphasized that "using the uncertainty of this pointwise a serious error since it assumes independecne of erros."

Large point sources make up a large percentage of anthropogenic CO2 emissions for the US and for other industrialized countries (Singer et al. 2014). In 2010, one third of US emissions were reported from only 311 sites of large point sources (US EPA, 2014). Total uncertainty in emissions from any geographic grid space thus has to reflect uncertainty in small or areal sources and in both the magnitude and location of large point sources. Woodard et al. (2015) developed one key, missing component needed to quantify the spatially explicit uncertainty in gridded inventories of CO2 emissions - an approach for dealing with the uncertainty in the locations of large point sources.

In our initial quantitative analyses of uncertainty (Hogue et al. 2016), we separated emissions from large point sources (e.g. from electricity generating plants as characterized by EPA's eGRID dataset (US EPA 2014)) and emissions from small and areal sources and we have used population density as a proxy for distributing all fossil-fuel related emissions beyond power plants. There are a then six components to uncertainty that need to be combined for an estimate of total uncertainty for the individual cells:

  • Uncertainty in total national emissions
  • Magnitude uncertainty for large point sources
  • Spatial uncertainty for large point sources
  • Magnitude uncertainty of the population proxy
  • Spatial uncertainty for the population proxy
  • Uncertainty in using population density as a proxy for emissions

Our analyses to date have focused on the US to demonstrate the analytical methods and possibilities and because of the quantity and quality of the data available for the US.

Our research goal has been to produce a data set that reveals the uncertainty in spatially explicit analyses of anthropogenic CO2 emissions, plus an analysis that begins to establish how to reduce this uncertainty or to deal with it. We have succeeded in developing an approach and a set of tools for doing this analysis, and we have applied it to the US using data on power plants in the US. The next steps are underway and are looking at the influence of scale and the impact of being able to deal separately with large point sources.