Uncertainties Associated With Activity Data

Activity data are often more closely linked to economic activity than are emission factors are. However, unlike emission factor data, there is typically no statistical sample of alternative activity data estimates readily available to fit distributions and estimate uncertainty. There are often well established price incentives and fiscal requirements for accurate accounting of economic activity. Activity data therefore tend to have lower uncertainties and a lower correlation between years than emission factor data. Activity data are often collected and published regularly by national statistical agencies, which may have already assessed the uncertainties associated with their data as part of their data collection procedures. These previously developed uncertainty estimates can be used to construct PDFs. This information will not necessarily have been published, so it is recommended to contact the statistical agencies directly. Since economic activity data are not usually collected for the purpose of estimating greenhouse gas emissions and removals, it is good practice to assess the applicability of the uncertainty estimates before using them.

There are several approaches that may be helpful in assessing the uncertainty of activity data in particular circumstances:

Activity data based on complete samples (censuses): Census data are activity data that are based, in principle, on counting every instance of a particular activity. Census typically includes both systematic and random errors. Systematic errors arise through systematic undercounting or double counting. Random errors are typically the sum of a range of commonplace errors. Random errors usually can be expected to be normally distributed and serially uncorrelated. Because activity data are usually collected by the same people, using the same processes, for each observation, systematic errors are likely to take approximately the same value each year. There are several approaches to identifying the potential uncertainty of activity data for complete samples. These approaches are often an integrated part of a QA/QC plan:

• To check for the size of random errors, look for fluctuations over time, and differential fluctuations in series that ought to be highly correlated with the data of interest.

• To check for bias errors, cross-check the data of interest with other, related information. One might, for instance, look up and down the supply chain for fuels, comparing coal production, coal import/export, and reported consumption. Or, one might study activities for which data are collected independently but which ought to be highly correlated with the data of interest, for instance reported fuel input vs. electricity output. One might also look at activity data of different frequencies (e.g., monthly, annual), if they are collected using different approaches.

• Interpretation of statistical differences, within, for instance, national energy data are an example of crosschecking. The comparison between energy-related carbon dioxide emissions derived from the IPCC reference approach is a formal cross-check with emissions estimates derived from other sources.

Census-based activity data are often 'precise but inaccurate' in the taxonomy shown in Figure 3.2, the random errors are small, but there may be larger bias errors. Cross-checking can suggest upper and lower bounds for possible bias errors, and sometimes will permit an actual estimate of the bias error. A possible bias error lurking within these bounds may often be characterised as a truncated uniform distribution: cross-checking shows that the unobservable true value must lie within a particular range, but there may be no reason to think any point within that range is more or less likely. However, because the bias errors in activity data are likely to be highly correlated, the difference between the reported value and the unknown true value is likely to be about the same every year, and this characteristic should be taken into account when estimating trend uncertainty.

Activity data based on random samples: Some kinds of activity data are derived from sample surveys, for instance consumer surveys, land use surveys, or forest cover surveys. In these cases, the data will be subject to sampling errors, which will be normally distributed and uncorrelated over time. The agency conducting the sample will normally be able to advise on sampling error. If this information is unavailable, it may be possible to identify or infer the sample and population sizes and calculate sampling error directly.

Was this article helpful?

0 0
Guide to Alternative Fuels

Guide to Alternative Fuels

Your Alternative Fuel Solution for Saving Money, Reducing Oil Dependency, and Helping the Planet. Ethanol is an alternative to gasoline. The use of ethanol has been demonstrated to reduce greenhouse emissions slightly as compared to gasoline. Through this ebook, you are going to learn what you will need to know why choosing an alternative fuel may benefit you and your future.

Get My Free Ebook

Post a comment