Technical Comments

Comment on "The Ocean Sink for Anthropogenic CO2"

Science  17 Jun 2005:
Vol. 308, Issue 5729, pp. 1743
DOI: 10.1126/science.1109620

Using data from recently completed hydrographic surveys of dissolved inorganic carbon (DIC) and related tracers in the world's oceans, Sabine et al. (1) arrived at an estimate of 118 ± 19 Pg C for the uptake of anthropogenic CO2 by the oceans through 1994. This estimate uses the “ΔC* method” pioneered by Gruber (2) and now widely applied to estimating ocean carbon changes. Here, I highlight several complications associated with the ΔC* method that have not been previously discussed. Consideration of these factors suggests that estimates of ocean uptake of anthropogenic CO2 may need revising.

One complication is that the ocean inventory of anthropogenic CO2 is an incomplete measure of the change in the ocean carbon content. The term “anthropogenic CO2,” as used by Sabine et al., refers to the excess carbon dioxide that has accumulated in the ocean as a direct response to rising CO2 levels since preindustrial times. Changes in carbon accumulation driven by processes within the ocean, such as warming (whether anthropogenic or otherwise), or changes in ocean stratification are not counted as “anthropogenic” CO2. These contributions are certainly much smaller than the component driven by rising atmospheric CO2 levels, but they are not necessarily negligible. A complete global carbon budget must therefore also include a term for these ocean-driven exchanges of CO2.

It is relatively straightforward to estimate the direct effect of long-term warming on air-sea carbon exchanges (3, 4). For example, using the box-diffusion model (5) tuned for consistency to match the Sabine et al. estimate of 118 petagrams of carbon in the absence of warming, the net uptake in the presence of warming is found to be 105 Pg C. This is consistent with a warming correction of –13 Pg C (68). Warming releases CO2 from the ocean primarily because CO2 is less soluble in warmer water.

It is likely that warming over the past century has also influenced CO2 exchange indirectly through increases in the density stratification of the upper ocean, thereby decreasing vertical mixing and increasing the trapping of nutrients and metabolic CO2 in subsurface waters. There is no simple way to estimate this stratification effect, but based on the results of general circulation models (3, 4), the effect through 1994 is most likely on the order of +6 Pg C, offsetting the warming effect. Combining the warming and stratification effects thus leads to an estimate of –7 Pg C for the ocean-driven term, with an uncertainty that is not well constrained but probably is about ±10 Pg C.

Another complication is that the ocean inventory of anthropogenic CO2 may not be determined reliably using the ΔC* method in the presence of warming or other ocean variability. The ΔC* method is a hybrid of two methods, the first applied in the upper ocean where the isopycnal layers are contaminated everywhere with anthropogenic CO2 and the second applied in the deeper waters where the isopycnal layers are partially uncontaminated. In the upper ocean, the ΔC* method principally relies on chlorofluorocarbon ventilation ages in combination with the known history of atmospheric CO2. If a different atmospheric history were used, the estimate of anthropogenic CO2 uptake in these water would change proportionally, which illustrates that the method is not a direct observation of carbon accumulation but effectively a model-dependent estimate, albeit a highly constrained one. This model assumes that the ocean circulation has remained steady with time and is therefore subject to error if circulation rates have changed. For example, the box-diffusion model indicates that if the vertical diffusion rates were 20% higher before 1980, the Sabine et al. estimate (1), which effectively projects modern diffusion rates over the entire period, would underestimate the actual uptake by 7 Pg C.

In the deeper waters, the ΔC* method depends on the spatial gradients in the tracer ΔC*, which is a mathematical function of total carbon, oxygen, nitrate, phosphate, silica, alkalinity, temperature, and salinity. The assumption is made that the gradient in ΔC*, along a given isopycnal surface from the older uncontaminated waters to the younger contaminated waters, is a measure of anthropogenic CO2. However, the gradient in ΔC* can be produced not only by uptake of CO2 but also by other processes, particularly air-sea exchanges of heat and O2 (9). Based on the ΔC* sensitivity to heat and O2 (10) and plausible estimates of changes in heat and O2 content of the deeper waters (1116), corrections on the order of 3 to 5 Pg C are implied, with the heat and O2 effects partly canceling. More work is needed to assess these corrections as well as the impact on variable circulation, which are complicated by our limited knowledge of global hydrographic changes before the late 20th century. Until these corrections are properly assessed, it seems appropriate to revise upward the uncertainty in the estimate of anthropogenic CO2 uptake. A reasonable revision might be from ±19 to ±23 Pg C/year.

In summary, the Sabine et al. (1) estimate of anthropogenic CO2 uptake should be combined with an additional term of about –7 ± 10 Pg C to account for air-sea CO2 exchanges driven by warming and stratification. The error estimate on the anthropogenic contribution should furthermore be increased to about ±23 Pg C to reflect uncertainties associated with changes in ocean circulation, heat content, and O2 content on the ΔC* method. Combining these corrections yields an estimate of 111 ± 25 Pg C for the net oceanic uptake of CO2 from 1800 to 1994. I hope that this comment will initiate a process to examine these corrections more closely.

References and Notes

Related Content

Navigate This Article