Introduction

Atmospheric CO concentrations are increasing by about 1.5 ppm every year. This 1.5 ppm increase integrated over the volume of the atmosphere accounts for about 50% of the CO released to the atmosphere from fossil fuel burning and deforestation (henceforth ``anthropogenic'' CO). The rest of the anthropogenic release (the part that does not remain in the atmosphere) goes mainly into the oceans and the terrestrial biosphere and soils. Knowing how CO is partitioned between oceanic and terrestrial CO sinks is critical for making CO forecasts into the future. This paper provides a brief review concerning the oceanic uptake of anthropogenic CO. (A more detailed review is provided by Siengenthaler and Sarmiento, 1993.)

Thermodynamic considerations suggest that a 1.5 ppm per year rise in atmospheric CO levels should lead to a CO increase in ocean surface water of about 1 mole/l. Until the late 1980s, dissolved inorganic carbon (DIC) measurements in seawater could not resolve CO differences less than 15-20 moles/l. Thus direct measurements of the anthropogenic CO increase were out of the question. Measurement capabilities have now improved to the point where DIC measurements can be made to an accuracy of 2-3 moles/l. Under the auspices of the World Ocean Circulation Experiment (WOCE) and the Joint Global Ocean Flux Study (JGOFS), ocean chemists are accumulating an ocean-wide data set from which the ocean's current CO distribution can be mapped. Unfortunately, it may still be 10 or 20 years before the ocean's CO levels will be mapped well enough a second time to document the CO increase by difference.

In the meantime, our primary methods for estimating oceanic CO uptake will have to be based on models which help us evaluate indirect or proxy constraints on CO uptake. Before 1990, most modelling efforts employed simple box models which attempted to predict the oceanic penetration using some kind of upper ocean mixing parameterization. In these models, ocean mixing rates were determined by adjusting or `tuning' mixing parameters until the model could reproduce the transient penetration of bomb-produced tracers like tritium (H) and bomb C. If one assumes that the mixing coefficients appropriate for bomb tracers are also appropriate for anthropogenic CO, the tuned model can be used to simulate the increase in oceanic CO using the observed atmospheric CO time history as a boundary condition.

Recently, two attempts have been made to calculate anthropogenic CO uptake using ocean general circulation models (GCMs) in which CO uptake is driven by a fully resolved ocean circulation field (Maier-Reimer and Hasselmann, 1987; Sarmiento et al., 1992). The GCMs and the tracer-calibrated box models are in basic agreement that the oceanic uptake of anthropogenic CO during the 1980s was about 2 GtC/yr (2 x 10 g C per year). This is about 30% of total anthropogenic emissions, or about 60% of the portion attributed to the ocean/terrestrial sink.

It is important to point out that ocean models generally treat the entry of anthropogenic CO into the ocean as if it is a simple perturbation of the ocean's natural carbon cycle. ``Anthropogenic CO'' is treated as a separate tracer from the background or natural DIC in the ocean. Of course, anthropogenic CO and natural CO are not separable. CO only goes into the ocean when the atmospheric partial pressure of CO (pCO) exceeds the CO partial pressure in the ocean. Maps of pCO, the pCO difference between ocean and atmosphere, document large CO influxes and outfluxes which were presumably in balance in pre-industrial times. The anthropogenic perturbation skews the influx-outflux balance to favor influx overall.

The oceanographic perspective on CO uptake was dealt a big blow at the outset of the 1991-1994 review period. Keeling et al. (1989) and Tans et al. (1990) made estimates of the ocean's CO uptake by trying to fit the observed latitudinal variation of atmospheric CO concentrations using atmospheric transport models. The atmospheric models in these exercises are given known source distributions (anthropogenic CO enters the atmosphere mainly in the northern hemisphere) and different ocean CO sink scenarios. The atmospheric models then predict latitudinal CO distributions for each source-sink scenario. These can then be evaluated against the observed latitudinal distribution as derived from the network of CO sampling stations (Conway et al., 1988).

During the 1980s, atmospheric CO levels in the high latitudes of the northern hemisphere were only 3 ppm higher than CO levels in the high latitudes of the southern hemisphere. The atmospheric transport calculations of Keeling et al. and Tans et al. suggest that this gradient is too small to allow much transport between the hemispheres. Thus the main sink for atmospheric CO would appear to be in the northern hemisphere.

According to Tans et al., the terrestrial biosphere and soils must be the main CO sink because the ocean sink in the northern hemisphere is fairly small. Tans et al. conclude that there must be a substantial net increase in carbon storage in northern hemisphere terrestrial vegetation and soils every year. They partition roughly 80% of the global ocean/terrestrial sink into the terrestrial biosphere and soils, and give the oceans only about 20%. They estimate that the oceans are currently taking up only 0.3-0.8 GtC/yr.

Tans et al.'s increased terrestrial carbon storage in temperate northern latitudes may be evidence for massive amounts of CO fertilization, the direct stimulation of vegetative growth by enhanced CO concentrations in the atmosphere. In their favored scenario, the annual carbon storage in northern terrestrial biomass and soils exceeds carbon losses due to tropical deforestation by about 1.0-1.5 GtC/yr. If the oceanic uptake is larger, the level of CO fertilization is still substantial, but northern terrestrial carbon storage is reduced to a level which more-or-less balances the release of CO by tropical deforestation.

The Tans et al. conclusion presents oceanographers with a major problem. Given the huge area of ocean in the southern hemisphere, the ocean's capacity to absorb CO should be skewed toward the south (Sarmiento et al., 1992). Bomb tracers are observed to penetrate vertically throughout much of the ocean's upper kilometer in both hemispheres. How could the southern hemisphere ocean be such a small sink for anthropogenic CO?

The problem here, in one important aspect, is mainly a problem of definition. The atmosphere's north-south CO gradient is influenced by ocean-atmosphere CO fluxes which include both natural and anthropogenic components. The ocean models used to estimate oceanic uptake of anthropogenic CO explicitly ignore the natural cycle. Natural fluxes of CO between the ocean and atmosphere can be quite large with respect to the anthropogenic perturbation. The region between 5N and 5S in the equatorial Pacific by itself outgasses 1 GtC/yr which must be balanced by CO uptake elsewhere.

Of particular interest here is the possibility that the ocean is naturally transporting carbon across latitude circles or between the hemispheres. Meridional transports within the ocean imply that there are natural CO fluxes between the ocean and atmosphere which influence atmospheric CO gradient. The ocean and land also exchange carbon (via rivers). These aspects of the natural cycle are not taken into account in the Tans et al. analysis. The question then becomes: are there characteristics of the natural carbon cycle which can reconcile 2 GtC/yr of ocean uptake with a small interhemispheric transport in the atmosphere? The answer, as will be reviewed below, would seem to be ``yes,'' although our knowledge of the natural carbon cycle is still not adequate to close the carbon budget completely.