Comparisons were made among three different emissions data sets: CDIAC global total emissions, CDIAC country-level emissions, and EIA country-level emissions. These revealed small discrepancies with two origins. First, different data sets include different components of total emissions, Eq. 1. The CDIAC global total includes all terms, CDIAC country-level data omit FBunkers and FNonFuelHC, and EIA country-level data omit FCement but include FBunkers by accounting at country of purchase. The net effect is that the EIA and CDIAC countrylevel data yield total emissions (by summation) that are within 1% of each other, although they include slightly different components of Eq. 1, and the CDIAC global total is 4–5% larger than both sums over countries. The second kind of discrepancy arises from differences at the country level, the main issue being with data for China. Emissions for China from the EIA and CDIAC data sets both show a significant slowdown in the late 1990s, which is a recognized event (16) associated mainly with closure of small factories and power plants and with policies to improve energy efficiency (17). However, the CDIAC data suggest a much larger emissions decline from 1996 to 2002 than the EIA data (SI Fig. 10). The CDIAC emissions estimates are based on the UN energy data set, which is currently undergoing revisions for China. Therefore, we use EIA as the primary source for emissions data subsequent to 1980.

- Concluding paragraph (on page seven of the PDF), my emphasis.