The addition of carbon dioxide to the atmosphere through fossil fuel burning and deforestation is an inadvertent global experiment, the outcome of which has profound implications for the future of the earthÕs climate and hence for humankind itself. One millennium from now, long after all fossil fuel supplies will have been exhausted, the ocean will have absorbed and retained about 7 out of every 8 molecules of anthropogenic CO₂ ever emitted to the atmosphere. Confidence in this prediction comes from the agreement between related simulations in global ocean carbon-cycle models, from the simplest (which divides the ocean into a few boxes) to the most complex (which contains thousands of grid cells). Less confidence can be given to any one model's prediction about more immediate changes, e.g., over the next couple of centuries and less agreement exists concerning today's regional uptake patterns, as well as those during other times. Models differ substantially even without consideration of potential changes in ocean circulation and possible shifts in the composition of oceanic plankton, much less the large uncertainty in future emissions of anthropogenic CO₂. Marine carbon cycle models have provided important constraints on the large-scale patterns of the marine uptake of anthropogenic CO₂, and remineralization; important fluxes that have largely been elusive to direct observation over large spatial scales.

The challenge to oceanographers is to synthesize the available data sets and incorporate into models that can be used for simulation of the partitioning of CO₂ between the atmosphere and the ocean. GAIM is responding to that challenge in the Ocean Carbon-Cycle Model Intercomparison Project (OCMIP), a coordinated effort of evaluation and intercomparison of 3-D global marine carbon cycle models.

Evaluation of marine carbon cycle models using observations of carbon-system and related parameters is necessary in order to establish the reliability of using such models for future prediction. A highly simplified ÒperturbationÓ approach, in which the natural carbon cycle and its attendant biological complexity are ignored, is feasible if ocean circulation and biogeochemistry can be assumed as invariant. However, there are many indications that the earthÕs climate and ocean circulation are indeed changing and may change dramatically in the future. In that case, the potential for complex marine biogeochemical feedbacks is large, and we are therefore obliged to develop the capability to model those aspects of the natural marine carbon cycle that are relevant to the air-sea partitioning of carbon dioxide. There are important data sets being compiled such as those by the Joint Global Ocean Flux Study (JGOFS), the next generation of satellite ocean color products, and a number of existing seasonal, global scale syntheses of nutrients, dissolved oxygen, surface carbon dioxide and chlorophyll. These data present an unprecedented opportunity for the evaluation of models of the natural marine carbon cycle.

Model intercomparisons are necessary for model improvement as part of an iterative process (Fig. 1.1). The customary cycle for the development and verification of model predictions include: (1) model output is evaluated by comparison with observations and strengths and weaknesses are identified; (2) models are reformulated or parameters are adjusted with the aim of model improvement; (3) models are run again and the output is re-evaluated. This iterative process is time-consuming and expensive for global 3-D carbon cycle models. Model intercomparisons are an intervention in the normal iterative process. A coordinated intercomparison can accelerate model improvement because different models perform with varying degrees of success, and the identification of the causes of model strengths and weaknesses is facilitated, leading to a more rational model adjustment. The intercomparison methodology also provides a better estimate of the uncertainty of model predictions, so long as the models are significantly different, but equally viable.

During the first phase of OCMIP, the focus was to identify differences between simulations of both natural and anthropogenic CO₂ in four 3-D models: (1)Max Planck Institut fur Meteorologie (MPIM, Germany), (2) Princeton University/GFDL (USA), (3) Hadley Centre (U.K. Met. Office), and (4) IPSL/CFR-LMCE-LODyC (France). Since that time, several additional models have been added. Additionally, OCMIP compared measured vs. simulated 14C (for both natural and bomb components, separately) as a means to validate the model circulation fields which drive each of the four carbon-cycle models.

Figure 1.1: A schematic diagram showing the intervention of model intercomparison (shaded circle) in the normal model development process (unshaded boxes) undertaken by individual investigators.

Objectives

The ocean is by far the largest active reservoir of carbon on earth and is the ultimate residence for the majority of anthropogenic CO₂, despite relatively slow oceanic uptake which cannot keep pace with excess CO₂ emissions to the atmosphere. Ocean models offer the most appropriate means to estimate past and present oceanic uptake of anthropogenic CO₂, and they provide the only option to predict future changes. Simplistic calculations have shown that about 7 out of every 8 molecules of anthropogenic CO₂ emitted to the atmosphere will eventually reside in the ocean; and the same result is found with idealized simulations in 3-D models. But, however, these simulations do not consider additional uncertainties due to changes in ocean circulation, chemistry, and biology. Even without these complications transient uptake by the ocean is poorly constrained, particularly at the regional level. Previous model results have demonstrated that predictions of today's air-sea flux of anthropogenic CO₂ differ by 20% globally, but they differ by nearly a factor of two in the Southern Ocean, the largest sink. As the Southern Ocean sink absorbs proportionally even more carbon (relative to other regions), global differences between models will likely increase. OCMIP aims to better constrain our understanding of CO₂ uptake by the ocean, the major long-term sink, through standardized validation and comparison of independently developed ocean models. The long-term objectives of OCMIP are:

1. To improve 3-D marine carbon cycle models for their use in predicting the future partitioning of carbon dioxide between the atmosphere and the ocean
2. To provide estimates of important but poorly-known quantities that are relevant to the marine carbon cycle: new production, aphotic zone remineralization and anthropogenic CO₂ fluxes and inventories

Since the above objectives require an understanding of the natural marine carbon cycle and the consequences of anthropogenic perturbations, both of which are profoundly influenced by ocean circulation, OCMIP defined the following shorter-term objectives:

1. To evaluate and intercompare the ability of 3-D global models to simulate the natural marine carbon cycle
2. To evaluate and intercompare the ability of 3-D global models to simulate the marine uptake of anthropogenic CO₂
3. To evaluate and intercompare those aspects of the ocean circulation that are relevant to the natural marine carbon cycle and its anthropogenic perturbation

To achieve these objectives OCMIP has been running a series of simulations (Table 1.1), most of which are designed so that they may be readily evaluated with existing data sets.

Model evaluation (often referred to as validation) is critical to any exercise where model predictions are to be taken seriously. OCMIP evaluates the behavior of ocean models by comparing their simulated distributions of appropriate tracers to measurements taken in the real ocean. The tracers of the greatest interest to the first objective (storage of anthropogenic CO₂) are bomb 14C and CFC's, which like anthropogenic CO₂ are of a transient nature and have permeated only near-surface waters. One benefit to using both these tracers is that their surface-ocean/atmosphere equilibration times bracket that for anthropogenic CO₂. Participating models incorporated two very different tracers of deep-ocean circulation: natural 14C and 3He. Natural 14C is produced cosmogenically in the upper atmosphere, is rapidly attached to molecules of CO₂, and is transferred to the ocean through gas exchange at air-sea interface; 3He is unique in that its only supply is from the ocean floor along mid-ocean ridges.

Experimental Methods

Natural CO₂

It was assumed that the natural carbon cycle was the state of the ocean-atmosphere carbon system prior to any significant anthropogenic forcings on the global carbon budget. The time frame of the natural carbon cycle is usually considered to be the millennia leading up to the 19th century, when atmospheric CO₂ concentrations were "steady" at about 280 matm. The natural marine carbon cycle plays an important role in the partitioning of carbon dioxide between the atmosphere and the ocean through two important processes: the solubility pump and the biological pump, both of which act to create a global mean increase of dissolved inorganic carbon (DIC) with depth, maintaining atmospheric CO₂ at a level considerably lower by about a factor of three than it would otherwise be.

The solubility pump maintains a vertical DIC gradient due to the fact that cold waters, which originate in high latitudes and fill up the deep ocean, can hold more DIC than warm waters at equilibrium with a fixed atmospheric pCO₂, a result of the higher solubility and dissociation of CO₂ (into carbonate and bicarbonate ions) in cold water. The vertical DIC gradient depends not only on the vertical temperature gradient but also the degree to which surface waters equilibrate with the atmosphere before sinking. Unlike most gases, the equilibration of surface water CO₂ with the atmosphere takes about one year, and therefore the strength of the solubility pump may depend critically on the kinetics of air-sea gas exchange.

The biological pump consists of two separate pumps: that of organic matter and calcium carbonate. The organic matter pump affects the DIC distribution through the photosynthetic formation of organic carbon in surface waters and the sinking and subsequent remineralization of this organic matter deeper in the water column. The carbonate pump affects the DIC distribution though the biogenic precipitation of calcium carbonate in surface waters and the subsequent sinking and dissolution of this material deeper in the water column. These two pumps also affect the alkalinity of sea water through the nitrate and dissolved calcium distributions.

Solubility pump

It is difficult to determine directly from observations the relative strengths of the carbon pumps because the poorly-quantified ocean circulation patterns play a large hand in determining the DIC distribution. Ocean carbon cycle models, however, provide an attractive means for estimating the relative strengths of the carbon pumps. The solubility pump is determined in a fairly straightforward fashion by specifying atmospheric CO₂ and determining the oceanic DIC distribution that results from air-sea gas transfer of CO₂ and its dissociation, as given by the inorganic chemistry of CO₂ in sea water.

In view of the fact that ocean circulation is the most important but least constrained aspect of the solubility pump, this series of experiments determines the strength of the solubility pump in each ocean model, with the only difference being the ocean circulation. The remaining parameterizations, which include specification of the atmospheric pCO₂, CO₂ gas transfer velocity, CO₂ solubility, equilibrium constants for the carbonate system, surface temperature, salinity, alkalinity, wind-speed, sea-ice cover, were identical for all OCMIP models (Table 1.2).

It was possible to evaluate models of the solubility pump by removing the biological pump and anthropogenic contamination from the DIC distribution. The organic matter pump was removed using the apparent oxygen utilization and an estimate of the respiration quotient. Alkalinity, corrected for nitrate was used to eliminate the calcium carbonate pump. The anthropogenic component could be eliminated using related techniques, and by incorporating transient tracers that reveal age information.

The biologic pump

A much more demanding task was simulation of the biological pump. Models must simulate the net biogenic uptake of dissolved inorganic carbon in surface waters (new production) and the net remineralization of the organic carbon thus formed in the aphotic zone. Simulation of new production was complicated by the many factors that govern the rates of photosynthesis such as light and the availability of a variety of nutrients (nitrate, ammonia, silicate, iron, and phosphate), each of which depend on the physical environment (e.g. mixed layer depth and the flow field, particularly vertical velocity).

If C, N and P in organic matter are cycled in relatively fixed proportions on a global scale, then simulation of the organic matter pump was tantamount to simulating the distributions of nitrate or phosphate. Clearly there are deviations from an invariant ÒRedfield stoichiometryÓ, but we have, at least initially, regarded these deviations as relatively minor. Even with this simplification, however, simulating the oceanÕs organic matter pump was complicated by the variety of processes that govern the formation of organic matter in the euphotic zone, its export to depth and its subsequent remineralization. Unlike ocean circulation, no fundamental set of equations that govern nutrient cycling has been found. A wide range of biological models of varying sophistication have been put forth for specific applications. Model development is in its early stages but proceeding rapidly.

OCMIP chose a two-tiered approach to simulating the organic matter pump. In the first approach, the complexities associated with process-oriented modeling of new production were circumvented by imposing the observed, monthly mean distribution of surface phosphate on the models ("nutrient restoring" approach). Thus the models determined the necessary new production that was consistent with the observed surface nutrient field. In practice, this was done by calibrating the models to observations on a time scale that were shorter than the characteristic time scale of the circulation field so as to prevent surface nutrients in the model from departing markedly from observations. A time step of one month was used to restore the top 100 m of the water column, which is roughly equivalent to the euphotic zone in the open ocean as defined by the 1% light level.

Nutrient restoring simulations

In this approach, the observed, monthly mean distribution of surface phosphate was imposed on the model. Thus, when a modelÕs ocean circulation supplied phosphate-rich water to the euphotic zone (taken to be the upper 100 m), enough nutrients were removed to bring surface waters in agreement with observations. In practice, this was done by restoring nutrients to observations on a 30-day time scale. The removal of nutrients in surface waters represents new production of organic matter which was split evenly between particulate organic matter (POM) and as dissolved organic matter (DOM). POM was assumed to be remineralized immediately below 100 m according to the length scale given by organic carbon sediment trap observations. DOM was advected and diffused into the thermocline and remineralized with a time constant of 10 years.

The advantage of the nutrient-restoring approach was that it provided a data-constrained, but model-facilitated estimate of an important but poorly constrained quantity on large scales. Given that the estimated new production depends critically on the model ocean circulation field, the variety of estimates from the different circulation models produced confidence intervals for predicted regional and global new production. Further, these simulations provided estimates of how the models differed in their nutrient supply rates, information that is crucial for interpreting the process-oriented models. Ecosystem modeling studies show that this physical supply is often the cause for model error. A second advantage of this approach was that it facilitated a focus on aphotic zone remineralization. Modeling studies suggest that the aphotic zone nutrient and oxygen distributions are highly sensitive to the type of organic matter (dissolved or sinking particulate) exported from the euphotic zone, as well as the lifetime of that organic matter. Thus, improved estimates of aphotic zone remineralization was determined by evaluating models with aphotic zone nutrient and oxygen distributions. The nutrient-restoring models were run until most of the water column comes into equilibrium (several thousand years).

There were two disadvantages of the nutrient-restoring approach. First, small mismatches between the observed and simulated ocean circulation field, particularly with regard to vertical motion, can lead to erroneous and patchy estimates of new production. A second disadvantage of the nutrient restoring approach was that it was of little use for predicting future variations in the natural marine carbon cycle, such as might be expected if ocean circulation changes. Thus we recognize that a more process-oriented approach is ultimately necessary.

The calcium carbonate pump

Simulating the calcium carbonate pump with a process-oriented model presented another level of complexity beyond simulating the organic matter pump: the distribution of particular phytoplankton species (mainly coccolithophorids) must be simulated. The dynamics of CaCO3 dissolution are very complex, depending largely on sedimentary processes. Further, it is difficult to evaluate models of the CaCO3 pump because the alkalinity distribution is the essentially the only data set that can be brought to bear on the problem, and this distribution is dominated by the fluxes of fresh water across the air-sea interface. Fortunately, the calcium carbonate pump contributes relatively little to the vertical DIC gradient compared to the organic matter and solubility pumps. In order to treat the CaCO3 cycle, the observed nitrate- and salinity-corrected alkalinity (A) was used to constrain the carbonate pump. Upper ocean changes in A were distributed according to new production, and at depth dissolution was performed in such a way to ensure that the global mean profile of A remained in agreement with observations. This simplified approach reproduced inter-basin differences in deep ocean alkalinity and the carbonate ion saturation horizon.

The combined pumps

The dissolved inorganic carbon distributions were simulated by combining the three carbon pumps. Atmospheric CO₂ was specified at its pre-industrial value of 278 matm, with gas transfer and carbonate chemistry modeled as for the solubility pump. The effect of the organic matter pump on DIC was simulated by linking the source/sink terms for phosphate to DIC through a fixed Redfield ratio. The model-predicted DIC distributions were evaluated using DIC measurements from GEOSECS, TTO, and SAVE in addition to the JOGFS CO₂ survey data. These data were processed to remove the anthropogenic component.

Modeled Results

Although the models all had the same overall pattern of enhanced uptake in upwelling and convective regions and weak uptake in downwelling regions, there were dramatic regional differences among the models, particularly in the Southern Ocean. For example, dramatic differences between the Princeton model predictions observations were seen in the Southern Ocean, with the model overestimating the uptake of anthropogenic CO₂ (Fig. 1.2) Because this region of the ocean plays a critical role in determining atmospheric CO₂, these results have initiated the careful evaluation of the models with relevant observations and the identification of the processes that caused such differences among the models.

Figure 1.2: (a) Marine anthropogenic Dissolved Inorganic Carbon (DIC) in the Eastern Indian Ocean (along 95^oE) in 1995 computed using measurements from the JGOFS CO₂ survey and the method of Gruber, 1996. This method employs a new quasi-conservative tracer, C*, which is defined as the difference between the measured DIC concentration of a water parcel, corrected for biology, and the concentration the water parcel would have at the surface in equilibrium with a pre-industrial atmosphere. C* reflects both the anthropogenic signal and the air-sea pCO₂ difference. The air-sea disequilibrium component can then be discriminated from the anthropogenic signal using either information about the water age (e.g., from transient tracers such as CFCs or ³H-³He) or the distribution of C* in regions not affected by the anthropogenic transient.

(b) A section of anthropogenic DIC estimated from the Princeton ocean carbon cycle model for the same year in the grid boxes covering the longitude range of the cruises. Each point in (a) represents a computed value. Dashed lines in (b) indicate the midpoints of the vertical grid boxes of the model.

Anthropogenic CO₂

Since pre-industrial times, the rise in atmospheric CO₂ has caused the change in the air-to-sea CO₂ flux to be positive everywhere. This perturbation to the natural system, termed anthropogenic CO₂, is difficult to measure in the ocean. Four OCMIP models have been used to estimate oceanic uptake of anthropogenic CO₂ (Fig. 1.3). Although the models agreed to within 20% for global uptake of anthropogenic CO₂ during the 1980's (Table 1.3), regional uptake differed by much more. The range of estimates of global uptake of anthropogenic CO₂ during the 1980's from all OCMIP models is from 1,500 to 2,200 Gt C yr-1. This OCMIP range falls well within the spread of flux estimates from previous compilations of 1-, 2-, and 3-D ocean model results : 2005, 2006, and 2008 Gt C yr-1.

Figure 1.3: Comparison of 4 OCMIP model simulations of air-sea CO₂ flux.

Ocean uptake was highest in the high latitudes and at the equator (Fig. 1.4), i.e., in zones where deep waters uncontaminated with anthropogenic CO₂ communicate readily with the surface (via upwelling and convection). The Southern Ocean dominates as the major sink because of its large surface area. Models differed by more than 100% in their predictions of how much anthropogenic CO₂ was absorbed in the Southern Ocean. Additionally, the predicted position of maximum uptake in the same region differed between models by nearly 20%. The position of maximum uptake varied by roughly 15^o and the quantity absorbed differed by nearly a factor of two between models. All models absorbed from one third to one half of their uptake in the Southern Ocean, as a function of the simulated intensity of mixing between the surface and deep ocean in that region (the process which limits oceanic absorption of anthropogenic CO₂). Most of the difference in global uptake between models were explained by discrepancies in the Southern Ocean. In the 30^oS-30⁰N equatorial band (53% of the ocean's surface) all models absorbed 0.8 Gt C yr-1. North of 30^oN, differences between models were large, but the small ocean area caused little difference in the total absorbed by each model. Large differences in regional uptake, particularly in the Southern Ocean which will become a proportionally larger sink, suggest that future global uptake estimates from different models may differ by much more than they do at present (~20%).

Figure 1.4: Global ocean, zonal mean air-to-sea flux of anthropogenic CO? given in Pg C yr^-1 degree^-1 as calculated by four OCMIP models for 1990. The oceanic uptake per band of latitude was most important in the Southern Ocean, but even there predictions varied considerably between models, both in the total absorbed and in the position of maximum absorption.

C-14

For proper analysis, one must separately compare natural and bomb C-14 (Figure 1.5). The sum of both components is the total quantity which is measured. Deep ocean waters remain essentially uncontaminated by the bomb signal. For bomb-contaminated waters nearer the surface, pre-nuclear C-14 measurements come from coral records and a limited number of water samples taken more than 30 yr ago; otherwise, the natural C-14 is derived from relationships based on correlations with SiO2 and vertical penetration of bomb tritium.

Figure 1.5: Global mean profiles of (a) total, (b) natural, and (c) bomb ¹⁴C. The total ¹⁴C, for which measurements exist, reveals discrepancies between observed values and the simulations from GFDL (in the thermocline) and the MPIM (in the deep ocean). The former problem is a well known deficiency of the GFDL model. These discrepancies are both linked to the simulation of natural ¹⁴C, because virtually no bomb C-14 penetrates below 1000 m if one considers the global mean distribution.

Natural radiocarbon

Due to the weak flows characteristic of most of the deep ocean, it is not possible to directly measure much of the ocean's deep circulation. One can use natural 14C, however, as a type of clock for the deep ocean circulation. Atoms of 14C are produced naturally when nitrogen is bombarded with cosmic radiation in the upper atmosphere. This 14C rapidly attaches to molecules of CO₂, enters the ocean through air-sea gas exchange, and subsequently decays (with a half-life of 5730 yr), mostly in the deep ocean. Gradients of 14C result, both vertically and horizontally, that depend upon how long waters have been isolated from the ocean's surface.

Natural radiocarbon is particularly useful for evaluating the deep circulation fields of ocean models. Along the Western Atlantic GEOSECS section, the observations reveal that the basin north of the equator is filled with waters penetrating from the north, which are at most 500 yr old, a signature of North Atlantic Deep Water (NADW). The youngest water penetrating far to the south is a tongue at mid-depth, centered at 2500 m. Antarctic waters have an older signature, despite their recent contact with the atmosphere, due to mixing with the ocean's oldest waters in the Pacific. Antarctic waters penetrate northward both at intermediate depths (AAIW) and along the bottom (AABW).

The GFDL model failed to produce NADW which fills the deep North Atlantic. Thus it produced its oldest deep water in the north, in the opposite sense to the observations. In contrast, the other models did fill the deep North Atlantic with young waters of roughly the right age; however, each of these three models had its own problems. Analogous natural 14C sections for the Pacific revealed that all four models predicted rates for the northward penetration of bottom waters which were consistently too slow. In the West Pacific all simulations reproduced the gross features of the observations.

Table 1.4: Difference in natural delta¹⁴C (in ä) between Antarctic surface waters and the deep North Pacific minimum (along the W. Pacific GEOSECS track).

* Stratification at the surface is particularly notable in the IPSL model, an artifact which appears mostly due to the semi-diagnostic forcing (where T and S are restored toward the observations throughout the water column over much of the ocean, not only at the surface (prognostically) as in the other 3 models). More recent simulations in the IPSL model have eliminated this problem. Note that the IPSL difference (and thus age) fall to within the observed range when instead the "surface" Antarctic C-14 level is taken at 200 m.

A closely related problem is that along-bottom transport of Antarctic Bottom Water (AABW) appeared to be too slow in all models (Table 1.5). By simplistically assuming infiltration only by advection from the south, i.e., neglecting transport by other processes (e.g. vertical turbulent diffusion), the calculated northward AABW advection rate suggested that the models over-predicted the observed difference by 50 to 120%, indicating modeled AABW renewal rates were too slow.

Table 1.5: Along bottom natural delta¹⁴C (in ä) difference (between 45^oS and 45^oN) in the Pacific

Bomb radiocarbon

Beginning with the atmospheric nuclear weapons tests during the 1950s, the 14C content of the atmosphere increased sharply (Fig. 1.6). At its peak in 1963, atmospheric 14C was nearly double that of pre-industrial times. Subsequently, atmospheric 14C has declined, following implementation of international treaties banning atmospheric weapons tests. Today atmospheric 14C is only about 10% greater than during pre-industrial times.

Figure 1.6: Atmospheric record of CO₂, combining data from the Siple ice core and the smoothed Mauna Loa record, and atmospheric bomb C-14 above the 20^oS -20^oN equatorial band.

By carefully separating out the nuclear component through correlations with other ocean tracers, oceanographers have been able to use the bomb 14C signal in the ocean as a primary data set with which to evaluate ocean circulation models (Fig 1.7).

Figure 1.7: Observed and modelled radiocarbon distribution in the surface waters of the Eastern Pacific in 1973 and 1993. 1973 observations are an average of GEOSECS observations east of the date-line and the 1993 observations are from the WOCE P17 line along 135^oW. Observed values are taken from the shallowest sample above 100 m. The model used is from Toggweiler and Samuels (1995), and values are taken from the surface grid box, which is 51 m deep, along 135^oW.

Bomb 14C is of particular interest because of its similarities to anthropogenic CO₂ (which is mostly 12C) (Fig. 1.8). Both are isotopes of carbon, both are transferred from the atmosphere to ocean by air-sea gas exchange and concentrations of both in the atmosphere have changed substantially since pre-industrial times. Figure 1.7 shows the utility of using bomb C-14 in the upper ocean to evaluate ocean circulation models. The observed surface D14C values (shallowest sample less than 100m depth) from the eastern Pacific GEOSECS stations and WOCE P17 were compared with the global ocean circulation model of Toggweiler and Samuels (1995).

Figure 1.8: Bomb ¹⁴C inventory (vertical integral of concentration) in units of 10⁹ atoms ¹⁴C cm^-2 for the GFDL, MPIM, Hadley, and IPSL models. GEOSECS stations are marked with circles shaded according to their concentration using the same scheme displayed for the 3-D model results.

³He

As opposed to natural 14C which enters the ocean through the air-sea interface, 3He is produced along mid-ocean ridges at the ocean floor. Thus, boundary conditions of these two tracers are completely different; each can be used for unique insight into the patterns and timing of deep-ocean circulation.

Primordial He, which also invades the deep sea from the earth's interior via mid-ocean ridges, has a distinctly different isotopic composition (3He/4He ratio) than does the atmosphere. Vent waters, along with their 3He signature, enter the ocean at several hundred degrees centigrade and typically rise a few hundred meters in the water column, before coming to equilibrium and being transported along surfaces of constant density (isopycnals). When it was found that oceanographic circulation patterns deduced from 3He data contradicted the simple, yet classic model of deep-ocean circulation, it became obvious to many that the unique features of 3He could shed much light on the mysteries of deep-ocean circulation. Since that time, many more samples have been collected in the ocean. At present, this tracer has even been used to help validate deep-ocean circulation in one 3-D model. Modeling studies suggest that the boundary condition for 3He is proportional to the spreading rate of the mid-ocean ridges at least in the Pacific, Indian, and Southern Oceans.

In OCMIP, all 3-D models ran to equilibrium one standard simulation for 3He. Primordial 3He was added to the ocean at mid-ocean ridges along the sea floor as a function of ridge spreading rates. At the air-sea interface, all ocean models exchanged 3He with the atmosphere using boundary conditions which were analogous to those used for CO₂. Validations with 3He lead to insights unobtainable by other tracers into the performance of each model's deep ocean circulation. Correctly simulating the movement of deep waters is essential when one wishes to predict the behavior of the real ocean's natural carbon cycle and realistic distributions of artificially injected CO₂.

Chlorofluorcarbons

Chlorofluorcarbons (CFCs) are excellent tracers of ocean circulation for several reasons. First, they readily equilibrate in surface waters, so that the Òpre-formedÓ CFC concentration is easily estimated from the well know atmospheric CFC concentration and the CFC solubility in sea water. Second, CFCs are entirely anthropogenic, having no natural Òcontamination,Ó and are essentially inert in sea water. Third, the ratio of CFC-11 to CFC-12 in sea water can be used, in many cases to estimate the age of water parcels. Finally, accurate measurements of CFCs can now be made relatively easily. As a result of their great utility, CFCs are included in almost every major oceanographic campaign studying ocean circulation, including WOCE. Nearly 100,000 measurements for CFC's now exist.

The distributions of CFCs have yielded insight into the many circulation processes relevant to the uptake of anthropogenic tracers, such as ventilation of the thermocline, intermediate and deep water formation, and ventilation of the interior deep ocean through the exchange with the deep western boundary currents. CFCs are also now frequently used to evaluate oceanic general circulation models.

Summary

Bomb 14C is of particular interest because of its similarities to anthropogenic CO₂ (which is mostly 12C) (Fig. 1.8). Both are isotopes of carbon, both are transferred from the atmosphere to ocean by air-sea gas exchange and concentrations of both in the atmosphere have changed substantially since pre-industrial times. Figure 1.7 shows the utility of using bomb C-14 in the upper ocean to evaluate ocean circulation models. The observed surface D14C values (shallowest sample less than 100m depth) from the eastern Pacific GEOSECS stations and WOCE P17 were compared with the global ocean circulation model of Toggweiler and Samuels (1995).

1765-present

In the first scenario, the models were run from 1765 to the present and forced with the corresponding history of atmospheric pCO₂, as determined from ice core records and direct measurements. Surface pCO₂ was specified from the ÒbestÓ simulation of the natural carbon cycle in surface waters as determined from an evaluation of these simulations with surface nutrients, dissolved oxygen anomaly and the seasonal variations of surface pCO₂. The results of these simulations were evaluated mainly with data-based estimates of marine anthropogenic CO₂ for the early 1990s (mainly from the JGOFS CO₂ survey data) as well as from the 1970s and 1980s (from GEOSECS, TTO, SAVE etc.).

Future stabilization scenarios

The IPCC has proposed several atmospheric CO₂ scenarios as potential targets for the stabilization of atmospheric CO₂ over the next few hundred years. OCMIP has forced the ocean circulation models with two of these stabilization scenarios for the next 300 years. In the first scenario (S450), atmospheric pCO₂ increases monotonically from its present value to a constant value of 450 matm by about the year 2075. The second scenario (S650) is similar, except that a constant value of 650 matm is reached by about the year 2200. The models predicted the marine uptake of anthropogenic CO₂ necessary to be consistent with these scenarios and provided estimates of what the fossil fuel plus terrestrial CO₂ inputs must be to achieve these targets. The models were initialized at the present using the surface pCO₂ climatology and the surface alkalinity fields for the solubility pump simulations.