Continuous time-series of highly precise measurements of the atmospheric composition are of central importance to current understanding of the contemporary carbon cycle. CO₂ has been measured at the Mauna Loa and South Pole stations since 1957 (Keeling et al., 1995; Figure 3.2a), and through a global surface sampling network developed in the 1970s that is becoming progressively more extensive and better inter-calibrated (Conway et al., 1994; Keeling et al., 1995). Associated measurements of ¹³C in atmospheric CO₂ began in 1977 (Francey et al., 1995, Keeling et al., 1995, Trolier et al., 1996). More recently, complementary information has been available from O₂ concentrations (measured as ratios of O₂:N₂, see Box 3.5), which have been regularly measured since the early 1990s (Keeling and Shertz, 1992; Keeling et al., 1993; Bender et al., 1996; Keeling et al., 1996b; Battle et al., 2000; Manning, 2001; Figure 3.2a). O₂ concentration data for the 1980s have been gleaned by two methods: sampling of archived air flasks that were collected during the 1980s (Langenfelds et al., 1999), and measuring the air trapped in Antarctic firn (Battle et al., 1996).

In addition to fossil fuel CO₂ emissions, Figure 3.3 shows the observed seasonally corrected growth rate of the atmospheric CO₂ concentrations, based on the two longest running atmospheric CO₂ recording stations (Keeling and Whorf, 2000). It is evident from this comparison that a part of the anthropogenic CO₂ has not remained in the atmosphere; in other words, CO₂ has been taken up by the land or the ocean or both. This comparison also shows that there is considerable interannual variability in the total rate of uptake.

O₂ and CO₂ measurements are used here to provide observationally-based budgets of atmospheric CO₂ (

3.1). CO₂ budgets are presented here (Table 3.1) for the 1980s (for comparison with previous work; Table 3.3), and for the 1990s. The reported error ranges are based on uncertainties of global fossil fuel emissions, determination of the decadal average changes in the atmospheric CO₂ concentration, and O₂:N₂ ratio; and uncertainties in the assumed O₂:CO₂ stoichiometric ratios in the combustion of fossil fuels and in photosynthesis and respiration. The error ranges reflect uncertainties of the decadal mean averaged values; they do not reflect interannual variability in annual values, which far exceeds uncertainty in the decadal mean rate of increase, as is further discussed in Section 3.5.2. The salient facts are as follows:

• During the 1980s, fossil fuel emissions were on average 5.4 ± 0.3 PgC/yr and atmospheric CO₂ content increased on average by 3.3 ± 0.1 PgC/yr. Partitioning of CO₂ uptake was estimated based on archived flask O₂ measurements (Langenfelds et al., 1999) for the 1979 to 1997 period, taking the O₂ trend during 1991 to 1997 (Battle et al., 2000) into account. The resulting estimate of the ocean-atmosphere flux was -1.9 ± 0.6 PgC/yr and of the land-atmosphere flux -0.2 ± 0.7 PgC/yr. This partitioning is adopted here in Table 3.1. It is corroborated by independent O₂ measurements in Antarctic firn (Battle et al., 1996). Restricting the analysis to the Battle et al. (1996) data for the 1980 to 1989 period, an ocean-atmosphere flux of -1.8 ± 1.0 PgC/yr and a land-atmosphere flux of -0.4 ± 1.0 PgC/yr were obtained, i.e., results indistinguishable from the values in Table 3.1.
• Despite a greater emission rate of 6.3 ± 0.4 PgC/yr (see Section 3.4.1), the average atmospheric increase during the 1990s was 3.2 ± 0.1 PgC/yr, i.e., about the same as during the 1980s. An exceptionally low rate of increase during the early 1990s was balanced by a high rate during the late 1990s. Based on the longest existing O₂ records from La Jolla (California, USA) and Alert (northern Canada) (Keeling et al., 1996b; Manning, 2001; see Figure 3.4), the ocean-atmosphere flux during the 1990s was -1.7 ± 0.5 PgC/yr and the land-atmosphere flux was -1.4 ± 0.7 PgC/yr.

Ocean uptake in the 1980s as estimated from O₂ and CO₂ measurements thus agrees with the estimates in the SRRF (Schimel et al., 1995) and the SAR (Schimel et al., 1996) (although these were model-based estimates; this section presents only observationally-based estimates (Table 3.3)). Considering the uncertainties, the ocean sink in the 1990s was not significantly different from that in the 1980s. The land-atmosphere flux was close to zero in the 1980s, as also implied by the SAR budget. The land appears to have taken up more carbon during the 1990s than during the 1980s. The causes cannot yet be reliably quantified, but possible mechanisms include a slow down in deforestation (Section 3.4.2), and climate variability that resulted in temporarily increased land and/or ocean uptake in the early 1990s (Section 3.5.2). These budgets are consistent with information from atmospheric ¹³C measurements (see Box 3.6 and Table 3.4) and with budgets presented in the SRLULUCF (Bolin et al. 2000) except that estimated ocean uptake is smaller, and land uptake accordingly larger, than given in the SRLULUCF (see Table 3.3, footnote i).

Table 3.3: Comparison of the global CO2 budgets from Table 3.1 with previous IPCC estimates a,b,c (units are PgC/yr).
	1980s				1990s	1989 to 1998
	This chapter	SRLULUCFd	SARe	SRRFf	This chapter	SRLULUCFd
Atmospheric increase	3.3 ± 0.1	3.3 ± 0.1	3.3 ± 0.1	3.2 ± 0.1	3.2 ± 0.1	3.3 ± 0.1
Emissions (fossil fuel, cement)	5.4 ± 0.3	5.5 ± 0.3	5.5 ± 0.3	5.5 ± 0.3	6.4 ± 0.4	6.3 ± 0.4
Ocean-atmosphere flux	-1.9 ± 0.6	-2.0 ± 0.5i	-2.0 ± 0.5	-2.0 ± 0.5	-1.7 ± 0.5	-2.3 ± 0.5i
Land-atmosphere flux*	-0.2 ± 0.7g	-0.2 ± 0.6	-0.2 ± 0.6	-0.3 ± 0.6	-1.4 ± 0.7	-0.7 ± 0.6
*partitioned as follows
Land-use change	1.7 (0.6 to 2.5)g	1.7 ± 0.8	1.6 ± 1.0	1.6 ± 1.0	insufficient data	1.6 ± 0.8j
Residual terrestrial sink	-1.9 (-3.8 to 0.3)	-1.9 ± 1.3	-1.8 ± 1.6h	-1.9 ± 1.6		-2.3 ± 1.3
a Positive values are fluxes to the atmosphere; negative values represent uptake from the atmosphere. b Previous IPCC carbon budgets calculated ocean uptake and land-use change from models. The residual terrestrial sink was inferred. Here the implied land-atmosphere flux (with its error) is derived from these previous budgets as required for comparison with Table 3.1. c Error ranges are expressed in this book as 67% confidence intervals (±1). Previous IPCC estimates have used 90% confidence intervals (±1.6). These error ranges have been scaled down as required for comparison with Table 3.1. Uncertainty ranges for land-use change emissions have not been altered in this way. d IPCC Special Report on Land Use, Land-use Change and Forestry (SRLULUCF) (IPCC, 2000a; Bolin et al., 2000). e IPCC Second Assessment Report (SAR) (IPCC, 1996a; Schimel et al., 1996). f IPCC Special Report on Radiative Forcing (SRRF) (Schimel et al., 1995). g Ranges based on Houghton (1999, 2000), Houghton and Hackler (1999), and CCMLP model results (McGuire et al., 2001). h The sink of 0.5 ±0.5 PgC/yr in “northern forest regrowth” cited in the SAR budget is assigned here to be part of the residual terrestrial sink, following Bolin et al. (2000). i Based on an ocean carbon cycle model (Jain et al., 1995) used in the IPCC SAR (IPCC, 1996; Harvey et al., 1997), tuned to yield an ocean-atmosphere flux of 2.0 PgC/yr in the 1980s for consistency with the SAR. After re-calibration to match the mean behaviour of OCMIP models and taking account of the effect of observed changes in temperature aon CO2 and solubility, the same model yields an ocean-atmosphere flux of -1.7 PgC/yr for the 1980s and -1.9 PgC/yr for 1989 to 1998. j Based on annual average estimated emissions for 1989 to 1995 (Houghton, 2000).

Several alternative approaches to estimating the ocean-atmosphere and land-atmosphere fluxes of CO₂ are summarised in Table 3.4. Alternative methods for estimating the global ocean-atmosphere flux, based on surface-water pCO₂ measurements and ocean ¹³C changes (Quay et al., 1992; Tans et al., 1993, Heimann and Maier-Reimer, 1996; Sonnerup et al., 1999), respectively, have yielded a range of -1.5 to -2.8 PgC/yr (for various recent periods). The total anthropogenic CO₂ added to the ocean since pre-industrial times can also be estimated indirectly using oceanic observations (Gruber et al., 1996). A global value of 107 ± 27 PgC by 1990 can be estimated from the basin-scale values of 40 ± 9 PgC for the Atlantic in the 1980s (Gruber, 1998), 20 ± 3 PgC for the Indian Ocean in 1995 (Sabine et al., 1999), and the preliminary value of 45 PgC for the Pacific Ocean in 1990 to 1996 (Feely et al., 1999a) with a large uncertainty of the order of ± 15 PgC. Assuming that accumulation of CO₂ in the ocean follows a curve similar to the (better known) accumulation in the atmosphere, the value for the ocean-atmosphere flux for 1980 to 1989 would be between -1.6 and -2.7 PgC/yr. Although each individual method has large uncertainty, all of these ocean-based measurements give results comparable with the fluxes presented in Table 3.1. Consideration of model-based estimates of ocean uptake in Table 3.4 is deferred to Section 3.6.2.2.

The land-atmosphere flux based on atmospheric measurements represents the balance of a net land-use flux (currently a positive flux, or carbon source, dominated by tropical deforestation) and a residual component which is, by inference, a negative flux or carbon sink. Using the land-atmosphere flux estimates from Table 3.1, assuming that land-use change contributed +1.7 PgC/yr to the atmosphere during the 1980s (Section 3.4.2), then a residual terrestrial flux of -1.9 PgC/yr (i.e., a residual sink of similar magnitude to the total ocean uptake) is required for mass balance. This is the term popularly (and misleadingly) known as the “missing sink”. The central estimate of its magnitude agrees with previous analyses, e.g., in the SAR (if “northern forest regrowth” is combined with “residual terrestrial sink” terms in the SAR budget; Schimel et al., 1996) and the SRLULUCF (Bolin et al., 2000) (Table 3.3). The uncertainty around this number is rather large, however, because it compounds the uncertainty in the atmospheric budget with a major uncertainty about changes in land use. Using an error range corresponding to 90% confidence intervals around the atmospheric estimate of -0.2 PgC/yr (i.e., 1.6s, giving confidence intervals of ±1.1 PgC/yr), and taking the range of estimates for CO₂ released due to land-use change during the 1980s from Section 3.4.2, the residual terrestrial sink is estimated to range from -3.8 to +0.3 PgC/yr for the 1980s. Model-based analysis of the components of the residual terrestrial sink (Table 3.4) is discussed in Section 3.6.2.2.

IPCC Homepage