Appendix B presents a mass-balance analysis for the 20-largest CO₂ emitting nations. Table B.1 shows that anthropogenic emissions increase the abundance of CO₂ by, on average, a fraction of a part per million (ppm) in the whole column (ranging from 0.06 ppm for Australia to 0.76 ppm for the United States). These small signals are measurable, but their detection is confounded by the much larger and incompletely understood signals from terrestrial ecosystems. Figure 4.2 shows that, even at global scales, and when fluxes are averaged over an entire year, the apparent fraction of fossil-fuel CO₂ that remains in the atmosphere varies from year to year by as much as a factor of 2. Most of this variation is caused by the response of the terrestrial biosphere and oceans to climate anomalies (Francey et al., 1995; Keeling et al., 1995; Bacastow, 1976) and to increased fire activity during El Niño events (van der Werf et al., 2004, 2008). The magnitude of annual perturbations (sources or sinks) can be as large as one-quarter of the magnitude of global fossil-fuel emissions (e.g., Battle et al., 2000). To monitor anthropogenic emissions with tracer-transport inversion, one must be

FIGURE 4.2 Interannual variation in the airborne fraction of fossil-fuel accumulation in the atmosphere. The airborne fraction is defined as the change in the mass of atmospheric CO₂ divided by the mass of fossil-fuel CO₂ emitted. The dark line shows successive 5-year averages. SOURCE: Figure 7.4b from IPCC (2007a), Cambridge University Press.

able to separate these terrestrial and ocean fluxes from the anthropogenic emissions.

The best tests of self-reported emissions using atmospheric chemistry models and observations have involved the global or hemispheric budgets of long-lived, synthetic, fluorinated gases produced solely by human activities. Rowland et al. (1982) measured CF₂Cl₂ at several remote sites to determine a global mean abundance. With knowledge of the long atmospheric lifetime (i.e., slow chemical loss), they were able to infer annual global emissions to high accuracy (e.g., 10 percent) from the annual increase in the atmosphere. The magnitude of CF₂Cl₂ emissions estimated from inverse modeling contradicted that claimed by the chemical industry, which subsequently retracted its reported emissions in favor of Rowland et al.’s derived emissions. This scenario was replayed in 2008 for the long-lived greenhouse gas NF₃, which is used in rapidly increasing quantities in the manufacture of large flat-panel displays and photovoltaic cells. Prather and Hsu (2008) reviewed the production and lifetime of NF₃, disputing an industry estimate of emissions (Robson et al., 2006) as unrealistically low and argued that this gas should be detectable and increasing in the atmosphere. Within months Weiss et al. (2008) made the measurements and confirmed that the reported NF₃ emissions were indeed too low.

Inverse modeling based on atmospheric measurements also indicates much larger emissions of HFC-134a and SF₆ than the sum of emissions in national inventories reported to the United Nations Framework Convention on Climate Change (UNFCCC). Höhne and Harnisch (2002) showed that post-1998 emissions of HFC-134a are 50 percent higher than reported by Annex I countries (see Figure 4.3), which are expected to be the source of nearly all HFC-134a emissions. Similar results have been shown for SF₆ (Geller et al., 1997; Höhne and Harnisch, 2002). Although inverse modeling shows that emissions of many HFCs and CFCs are underestimated in UNFCCC inventories (and occasionally overestimated; see discussion of halon-1301 in Clerbaux and Cunnold, 2006), it offers no insight as to the source of error. Global or hemi-