Please note: You are viewing the unstyled version of this website. Either your browser does not support CSS (cascading style sheets) or it has been disabled. Skip navigation.

Assessment of Air-Sea CO2 Exchange Rates in the World’s Oceans Using Bomb 14C Inventories Derived from the WOCE Global Survey

   Print  PDF  Change text to small (default) Change text to medium Change text to large

1 March 2004 - 28 February 2007

Dr. Alison Macdonald
Woods Hole Oceanographic Institution, Woods Hole, MA 02543

Dr. Tsung-Hung Peng, NOAA/AOML
Dr. Rik Wanninkhof, NOAA/AOML
Dr. Robert M. Key, Princeton University
Program Manager: Dr. Kathy Tedesco, NOAA/OAR/OGP/GCC

Related NOAA Strategic Plan Goal:
Goal 2. Understand climate variability and change to enhance society’s ability to plan and respond.

Project Overview
Determining the spatial patterns and variability of carbon sources and sinks at global to regional scales, and documenting the fate of anthropogenic CO2 in the atmosphere and ocean are main goals of the NOAA contribution to the U.S. Interagency Carbon Cycle Science Program (CCSP). Quantifying air-sea CO2 exchange rates is critical to these goals because this exchange is the primary mechanism for CO2 transfer between the atmospheric and oceanic carbon reservoirs. The ocean inventory of bomb-produced radiocarbon (14C) is directly related to air-sea CO2 exchange and thereby provides a powerful constraint on the exchange rates. The large amount of high-quality radiocarbon data collected during the WOCE program provides the opportunity to improve our estimate of these air-sea CO2 exchange rates.
Project Goals
(1) To determine air-sea CO2 exchange rates in the world oceans based on the bomb 14C distribution observed during the WOCE global survey in the 1990s,
(2) To re-evaluate the air-sea CO2 exchange rates based upon the bomb 14C inventory estimated from the 1970s- GEOSECS data using an improved method of separating natural 14C from the observed 14C, and
(3) To use inverse methods to derive the air-sea CO2 exchange rates from the WOCE era oceanic bomb 14C distribution and ocean circulation.

An improved method of separating natural 14C from the observed 14C distribution is being used to estimate the bomb 14C distribution and inventory (Rubin and Key, 2002). The completion of this database was the focus of the first year of the project. During the second year the focus has turned to the lateral transport model analysis in which the transport models developed by Broecker et al. (1985) are being used to assess the regional air-sea CO2 exchange rates. Comparisons of regional 14C inventory estimates derived from gas exchange rates related to both cubic and square of wind speed have been conducted. The existing inverse box model of the Pacific Ocean will be used to derive the air-sea exchange rate of CO2 based on bomb 14C distribution during the third year of the project.

Bomb 14C database

The radiocarbon separation method devised by Rubin and Key (2002; referred to hereafter as the PALK method) is based on the strong linear relationship between potential alkalinity and measured radiocarbon in deep waters not yet contaminated by bomb-produced radiocarbon (Equation 1 and 2).

Natural ∆14C = -59 – 0.962(PALK-2320)
PALK = (Alkalinity + Nitrate)*35/Salinity

The PALK method is an adaptation of the silicate method published by Broecker et al. (1995) and was developed using the same GEOSECS data set. The regression statistics for the PALK method are marginally better than for the silicate method, but the important difference is that the PALK method does not show latitudinal bias. One problem with the PALK method is that many radiocarbon measurements do not have accompanying alkalinity measurements. In these cases some form of multiple-parameter linear regression (MLR) can be used to estimate alkalinity. Alternatively, for mid and low latitude samples the silicate method can be used after applying a minor calibration correction applied (Rubin and Key, 2002; Equation 4).

As part of the Global Ocean Data Analysis Project (GLODAP; work partially supported by this grant) most of the WOCE radiocarbon data as well as that from previous major expeditions were organized into a set of three data files – one for each ocean. All of the measured radiocarbon data were separated into bomb and natural components and then these data were used to produce objective global maps of the distributions at 33 depth layers. The objective gridding and the specific depth surfaces were chosen to match existing climatologies for temperature, salinity and nutrients. Both the maps and the data sets are available via the internet ( and are collectively known as GLODAP version 1.1 (Gv1.1). Preparation of the maps and data sets is described in Key et al. (2004). They also included global radiocarbon and bomb radiocarbon inventory estimates. The distribution of bomb 14C inventory is shown in Fig. 1.

One of the major deficiencies of Gv1.1 was that the WOCE Atlantic Ocean radiocarbon data available at the time was insufficient to objectively map in that basin. Therefore, older radiocarbon data from the Transient Tracers in the Ocean (TTO) and South Atlantic Ventilation Experiment (SAVE) were used. Since release of Gv1.1 GLODAP work has continued (with support from this and other grants). We now have WOCE era data from three additional North Atlantic cruises (plus 2 in the Pacific and 2 in the Indian). The same separation method has been applied to these new data. Once these new results are merged with Gv1.1 we will attempt to remap the North Atlantic radiocarbon distributions. We emphasize attempt because the WOCE era North Atlantic data distribution is such that the mapping results may be unacceptable. In the South Atlantic Gv1.1 used SAVE radiocarbon data and we cannot improve on that since there are almost no WOCE era radiocarbon samples in that basin. Fortunately, SAVE was in the late 1980s and the time difference between that data and WOCE data is sufficiently short relative to the time scales of interest for this work.

No one has yet determined if the PALK method can be improved given the larger data set and the higher quality alkalinity data now available. Rubin and Key (2002) noted that slightly different regressions were obtained from the three oceans, but suggested that the simplicity of a single global regression out-weighted the minor differences. Investigations with the WOCE data show the same minor ocean to ocean differences in the PALK regressions. This is almost always the case for oceanographic regressions which are empirical rather than theory based: the smaller the region, the better the regression. Choosing regional fits, however, causes other significant problems; specifically, the regional functions have to be made consistent across region boundaries. Failure to eliminate boundary discontinuities can result in significant problems when using or interpreting the results (as noted in Key et al., 2004 regarding the Goyet et al. 2000, global alkalinity maps).

More interesting, and perhaps more significant depending upon the application, recent research by C. Sweeney and R. Key has shown that in Pacific deep and abyssal waters the PALK method produces bomb 14C estimates that show an unexpected pattern. If the PALK method was perfect one would expect the estimated bomb 14C values to decrease with depth to zero somewhere in the thermocline (based on tritium and CFC distributions) and scatter randomly about zero from that depth to the bottom (except in bottom water formation areas where one might expect small positive near-bottom values). In fact while the deep and abyssal values average to zero, there is a pattern. Estimated values near the oxygen minimum tend to be slightly negative while near bottom values are slightly positive. This pattern should not exist and certain types of inversion calculations (those based on individual sample results) that try to make use of all bomb 14C estimates may be biased as a result. This potential bias is not expected to have any significant impact of the inventory-based models we will be using. Regardless, we will continue to investigate the issue. Early tests indicate that including AOU in a modified PALK equation can remove the bias in the Pacific, but side effects of this change and the applicability to other oceans is yet to be determined.

The GEOSECS data are too sparse to map objectively. Furthermore, Peacock (2004) convincingly demonstrated that the zonal averaging technique used by Broecker et al. (1995) to obtain a global bomb radiocarbon inventory was biased high by 10 to 25% due to the locations of the GEOSECS stations. In that work GEOSECS (and some WOCE) data were extrapolated globally using MLR (in silicate, temperature, oxygen, salinity, depth and latitude) and the natural component was estimated from global silicate climatologies. We are now using GEOSECS 14C data along with the zonal averages of Peacock (2004) to represent the global distribution used re-assess the 1975 air-sea CO2 exchange rates, The WOCE-era exchange rates are being computed from the WOCE 14C dataset for 1994.

During the past year significant progress has been made toward accumulating sufficient new data to make a post-TTO evaluation of the North Atlantic bomb 14C inventory. In addition to those data which were included in GLODAP v.1.1, we have new results from two NOAA cruises (OACES93 on the A16N line and OACES98 on the A05/AR01 line) as well as German data from the A1 sections. Surface results have been obtained from the NOAA reoccupation of A16N in 2003 and we will very soon have surface data from the re-occupations of A20 and A22 (measured, but not yet reported by NOSAMS). We are actively trying to track down results from line AR18 (around Iceland), but have as yet been unable to contact the persons responsible for this dataset. We have added the surface results from R. Nydal’s group (posted at CDIAC) and a new atmospheric history record.

Re-mapping and integrating these values is extremely time consuming and labor intensive, therefore, this step will not be attempted until we have completed the “radiocarbon data search”. Please note, this data search is not limited to cruises with radiocarbon data. Eventually the air-sea gas exchange rate derived from this work will be most useful in the analysis of anthropogenic CO2. Therefore the data search bounds include all those cruises with carbon and/or tracer data. Thus far, 70 new cruises have been added to the database. Almost all of these are from European expeditions. Very few were included under the WOCE umbrella or otherwise reported to any national collection agency.
Figure 2 gives a hint of the information content of the combined data. On first inspection the data are simply “messy” with the only dominant signal being the significant increase between the 1950s and 1960s. Closer inspection and/or relatively simply analysis, however, shows previously identified and new trends in the data including:

1. The “M” shaped pattern with high values in the subtropical and subpolar gyres and lower values at the equator and high latitudes originally identified by the GEOSECS data for all measured decades
2. Significantly larger variance at any given latitude during the time frame of the maximum atmospheric signal (1960s and 1970s)
3. Significant longitudinal variance for a given latitude and time (WOCE data)
4. Muted signals at high southern latitudes
5. A steady decrease, on average, following the atmospheric maximum
6. Significantly more data than any previous compilation

Once these data are fully analyzed we expect to be able to re-analyze the currently accepted equilibration time (~10yrs) by examination of the lag time between the atmospheric and surface ocean values. The fact that we now have the beginning of a representative “time series” for the North Atlantic should place significant constraint on numerical ocean/atmospheric models.
Lateral Transport Model
Broecker et al., (1985) developed lateral transport models for the three major ocean basins to simulate the GEOSECS bomb 14C distribution in the water column. With the WOCE radiocarbon data, this same model can be used to study the effect of regionally varying air-sea CO2 exchange rates on the distribution of bomb 14C in the oceans. To begin this task, we needed to reconstruct the lateral transport model developed by Broecker et al., (1985). We have successfully completed this task, and we have reproduced model results published in the original paper. We have produced a MatLab code to facilitate doing sensitivity studies for the three major oceans basins.

The next step was to summarize the revised bomb 14C inventories based on separation of the natural and bomb 14C components according to PALK method (Rubin and Key, 2002). The CO2 exchange rates given in Broecker et al (1985) were based on bomb 14C inventories estimated from Tritium penetration depth and smooth natural 14C profiles in the upper ocean. This yields different inventories from those derived from PALK-natural 14C relationship as given in Rubin and Key (2002). The tasks completed in the 2nd year of the project included the re-evaluation of CO2 exchange rates for three major oceans based on this revised 1975 inventory by Rubin and Key (2002) and the evaluation based on 1995 inventory using WOCE radiocarbon data.

Broecker et al (1985) assumed a uniform exchange rate for each basin, but for the re-evaluation, we are determining the sensitivity of the bomb 14C inventories to specific CO2 exchange rates for each of the 13 boxes (5 in the Atlantic, 5 in the Pacific Basins and 3 in the Indian) used in the Broecker model. The estimates are based on averaging the monthly gas exchange rates for each 4ox5o bin within the region of interest. The gas exchange values are calculated using the same input values (SST, SSS, P, U10) as is used in the pCO2 climatology of Takahashi et al. (2002). Note, to account for the effect of variable winds on the quadratic relationship of gas exchange with wind speed the 2nd central moment of the monthly winds on 6-hour intervals from NCEP/NCAR re-analysis is used (Kalnay et al., 1996).

The three ocean basins have been divided into latitudinal zones with a prescribed circulation (Figure 3). Based on the gas exchange parameterization outlined above the estimated invasion rates are:

Atlantic    Antarctic        25.80 (mol m-2yr-1)
    South temperate    17.20
    Equatorial        11.67
    North temperate    12.40
    Arctic            21.32
Pacific    Antarctic zone        25.91
    South temperate    16.71
    Equatorial         8.84
    North temperate    13.74
    Arctic            22.40
Indian    Antarctic        29.76
    South temperate    19.26
    Equatorial        10.30

These air-sea CO2 exchange values are used in the model to produce bomb 14C distribution that are compared with those derived by Broecker et al. (1985, 1995). They used invasion rates of 22.3, 19.2 and 19.4 mol m-2yr-1 for the Atlantic, Pacific and Indian Ocean basins, respectively.

Using the prescribed model of oceanic meridional transport (Figure 3), estimates of the atmospheric bomb 14C time history for each hemisphere and the equator (Figure 4), revised estimates of the zonal areas and the newly evaluated invasion rates, bomb 14C inventories (Figure 5, left panel) and surface concentrations (Figure 5, right panel) have been calculated for both 1974 and 1994. It is seen that in general the basin wide constant invasion rate of Broecker et al. (1985) has a better fit with the inventory estimated by Peacock (2004) for 1974, while the re-determined invasion rates result in a better fit with inventory based on WOCE data for 1994. There are significant deviations of inventories from WOCE data for 1994 if a constant ocean basin wide invasion rate of Broecker et al.(1985) is used.
For year 1974, the surface bomb 14C concentrations predicted from using the constant invasion rates have a better fit to the observed data than those derived from the re-determined different zonal invasion rates. However, for the year 1994, significant deviations of model results from the WOCE data are observed both for constant invasion rates and re-determined zonal invasion rates. This is an indication that the physical parameters in this simplified transport model, such as advection fluxes and vertical mixing rates (diffusivities) are inconsistent with the ocean dynamics. Hence, adjustments to physical parameters of this model are still needed.

For illustration of the direction of adjustments required, model results with reduced vertical mixing rates, or with the reduced upwelling and downwelling fluxes for the Atlantic Ocean (Figure 5) were considered. The left hand two panels are for the inventories, using either the constant zonal invasion rate of 22.3 mol m-2 yr-1 of Broecker et al. (1985) or the re-determined different zonal invasion rates. Similarly, the surface bomb 14C concentrations are shown on the right hand side panels. It is shown that the model surface bomb 14C concentrations are significantly lower than the observed values, especially in 1994. One of reasons could be that the vertical mixing rates are too fast in the model.

To test the effects of vertical mixing rates, the diffusivity in zones 2, 3, and 4 were reduced by half (Figure 6). This gives results with reducing the surface 14C concentrations, but not significantly enough to match the observations. The reduced vertical mixing rates also reduce the inventory. Again, the reduced inventories do not significantly improve the model results to match with the observations in 1994. The effects of reduced upwelling flux (by 2 Sv) in zone 3 and downwelling flux in zone 4 are also shown. The surface concentration has increased slightly, but not enough to match the observations. However, the inventory has increased slightly in equatorial zone, but reduced in the north temperate zone. The magnitude of the changes is too small to explain the deviations of model results from the observations. Proper adjustments in vertical mixing rates and upwelling fluxes are necessary to assess the CO2 invasion rates in the world’s oceans using bomb 14C data. In the Pacific and Atlantic Basins the box inverse model will help assess the direction of these adjustments.

Inverse Modeling

The final focus within this project was the inclusion of 14C within existing inverse box models of the Atlantic and Pacific Basins. In the first year of the project, the background and bomb radio carbon data obtained from the GLODAP database were reformatted for use within the inverse system. The next step focused on the Pacific where most of the 14C measurements were obtained along the long lines which have been used in the inverse model. The original thought was to map these data both in the horizontal to fill in stations which did not include 14C measurements and in the vertical to place 14C on the 2 dbar pressure intervals used as input to the box models. The inverse models would then be run individually for the GEOSECS era and the WOCE era, making the assumption that the ocean circulation has not changed over that time period. However, large gaps in the data set make mapping problematic. Therefore, it was thought that a more useful direction, given the results of the lateral transport model, would be to use the circulation defined by the inverse box model in place of the lateral transport model. After some investigation, however, this route was also found to be problematic as the inverse model by definition includes mass imbalance in every box which integrated over time overwhelms the solution.
Given that the inverse portion of this project still has half it’s available funding left (~2 man months), we have decided to instead use the results of the Pacific inverse model (spreading the mass imbalance over the full basin) as the basis for the lateral transport model, then to perform a variety of runs to determine sensitivity to the underlying circulation. If this method proves useful, we will then, look to recent circulation results provided in the literature from WOCE and CLIVAR analyses to use as the base circulation for lateral transport model in the other basins, and again perform multiple runs to determine sensitivity.
The results will be compared to the present lateral transport model results with the focus on determining those aspects of the 3-dimensional circulation which are of most importance to the inventory and surface concentration calculations. This effort will take place in the coming 6 months.

Peng T.-H., R. Wanninkhof, R. Key and A. Macdonald. Assessment of the Air-Sea CO2 Exchange Rates in the World’s Oceans Using Bomb 14C Inventories, poster ICDC7, Boulder, Colorado, 2005.
A. Macdonald, T.-H. Peng, R. Wanninkhof and R. Key, Ocean Bomb Radio-Carbon (14C) GEOSECS-WOCE, WHOI seminar, Dec, 2005.
Web Links: (

Broecker, W. S., T.-H. Peng, H. G. Ostlund, M. Stuiver. The distribution of bomb radiocarbon in the ocean, Journal of Geophysical Research, 90, 6953-6970, 1985.

Broecker, W.S., S. Sutherland, W. Smethie, T.-H. Peng, and G. Ostlund, Oceanic radiocarbon: separation of the natural and bomb components, Global Biogeochem. Cycles, 9, 263-288, 1995.

Goyet, C., R. Healy and J. Ryan, Global distribution of total inorganic carbon and total alkalinity below the deepest winter mixed layer depths, 28pp., ORNL/CDIAC-127, NDP-076, Carbon Dioxide Information Analysis Center, Oak Ridge National Laboratory, Oak Ridge, TN., 2000.
Kalnay, E., and et. al., The NCEP/NCAR 40-year reanalysis project, Bull. Am. Meteorol. Soc., 77, 437-471, 1996.

Key, R.M., A. Kozyr, C.L. Sabine, K. Lee, R. Wanninkhof, J. Bullister, R.A. Feely, F.

Millero, C. Mordy, T.-H. Peng, A global ocean carbon climatology: Results from Global Data Analysis Project (GLODAP), Global Biogeochem. Cycles, 18, GB4031, doi:10.1029/2004GB002247, 2004.
Peacock, S. Debate over the ocean bomb radiocarbon sink: Closing the gap. Global Biogeochem. Cycles, 18, doi:10.1029/2003GB002211, 2004.

Rubin, S. and R. M. Key, Separating natural and bomb-produced radiocarbon in the ocean: The potential alkalinity method, Global Biogeoche. Cycles, 16(4), doi: 10.1029/2001GB001432, 2002.

Takahashi, T.; Sutherland, S. G.; Sweeney, C.; Poisson, A. P.; Metzl, N.; Tilbrook, B.; Bates, N. R.; Wanninkhof, R.; Feely, R. A.; Sabine, C. L.; Olafsson, J.; Nojjiri, Y., Global sea-air CO2 flux based on climatological surface ocean pCO2, and seasonal biological and temperature effects. Deep-Sea Res. II, 49, 1601-1622, 2002.

This project is a collaboration with NOAA-AOML PIs Tsung-Hung Peng and Rik Wanninkhof. Peng is the lead PI on the project and is for responsible for the lateral transport model results. I have worked with them in determining how the inverse model and it results can best be used to within the context of this project. I have converted the Peng’s Fortran Atlantic lateral transport model to run in the Matlab. In November 2006 I participated in the NOAA Global Climate Change review panel. I will be attending the NOAA GCC PI meeting this September in MD.

Last fall I attended the New England Board of Higher Education Science Network as a panel member and mentor. I participated in the Learning Inovations Focus Group in Feb. 2007 and attended the organizational meeting of Falmouth Public Schools’ Project SEpTeMber in December 2006. Beginning early this year, I developed a collaboration with a teacher at Falmouth Academy to begin a pilot program focused on having junior and senior high school students base their science fair projects on observations obtained from oceanographic cruises. I currently have 4 students: one who will obtain observations from a mooring cruise going out this September; two who will use data from a hydrographic cruise this October and another who will be taking advantage of a variety NOAA fishery’s cruises occurring this fall.

Last updated: August 19, 2008

whoi logo

Copyright ©2007 Woods Hole Oceanographic Institution, All Rights Reserved, Privacy Policy.
Problems or questions about the site, please contact