Skip to main content
OCADSAccess DataThe oceanic sink for anthropogenic CO2 from 1994 to 2007 - the data. NDP-100

NDP-100 (2019)

NCEI Accession 0186034 Data and Documentation Files

The oceanic sink for anthropogenic CO2 from 1994 to 2007 - the data (NCEI Accession 0186034)

Nicolas Gruber1, Dominic Clement1, Brendan R. Carter2, 17, Richard A. Feely2, Steven van Heuven3, Mario Hoppema4, Masao Ishii5, Robert M. Key6, Alex Kozyr7, Siv K. Lauvset8, 12, Claire Lo Monaco9, Jeremy T. Mathis10, Akihiko Murata11, Are Olsen12, Fiz F. Perez13, Christopher L. Sabine14, Toste Tanhua15, and Rik Wanninkhof16

1Environmental Physics, Institute of Biogeochemistry and Pollutant Dynamics, ETH Zurich, Zurich, Switzerland
2National Oceanic and Atmospheric Administration, Pacific Marine Environmental Laboratory, Seattle, USA.
3Centre for Isotope Research, Faculty of Science and Engineering, University of Groningen, the Netherlands.
4Alfred Wegener Institute, Helmholtz Centre for Polar and Marine Research, Bremerhaven, Germany.
5Meteorological Research Institute, Japan Meteorological Agency , Tsukuba, Japan.
6Atmospheric and Oceanic Sciences, Princeton University, Princeton, NJ, USA.
7NOAA National Centers for Environmental Information, Silver Spring, USA.
8NORCE Norwegian Research Centre, Bjerknes Centre for Climate Research, Bergen, Norway
9LOCEAN, CNRS, Sorbonne Université, Paris, France.
10National Oceanic and Atmospheric Administration, Arctic Research Program, Silver Spring, USA.
11Research and Development Center for Global Change, Japan Agency for Marine-Earth Science and Technology, Japan
12Geophysical Institute, University of Bergen and Bjerknes Centre for Climate Research, Norway.
13Instituto de Investigaciones Marinas, CSIC (IIM-CSIC), Vigo, Spain
14Department of Oceanography, University of Hawaii at Manoa, Honolulu, USA.
15GEOMAR Helmholtz Centre for Ocean Research Kiel, Kiel, Germany
16National Oceanic and Atmospheric Administration, Atlantic Oceanographic and Meteorological Laboratory, Miami, USA.
17Joint Institute for the Study of the Atmosphere and Ocean, University of Washington, Seattle, USA


Fig. 1 Vertical sections of the change in anthropogenic CO2 between the JGOFS–WOCE era (~1994) and the Repeat Hydrography–GO-SHIP era (~2007). Shown are the zonal mean sections in each ocean basin organized around the Southern Ocean in the center. The upper 500 m are expanded. From Gruber et al. (2019).

Please cite this data set as:
Gruber, Nicolas; Clement, Dominic; Carter, Brendan R.; Feely, Richard A.; van Heuven, Steven; Hoppema, Mario; Ishii, Masao; Key, Robert M.; Kozyr, Alex; Lauvset, Siv K.; Lo Monaco, Claire; Mathis, Jeremy T.; Murata, Akihiko; Olsen, Are; Perez, Fiz F.; Sabine, Christopher L.; Tanhua, Toste; Wanninkhof, Rik (2019). The oceanic sink for anthropogenic CO2 from 1994 to 2007 - the data (NCEI Accession 0186034). Version 1.1. National Oceanographic Data Center, NOAA. Dataset. doi: 10.25921/wdn2-pt10

Please cite the underlying study as:
Gruber et al. (2019). The oceanic sink for anthropogenic CO2 from 1994 to 2007, Science 363, 1193 - 1199 (2019), DOI: 10.1126/science.aau5153


This NCEI accession consists of the estimated changes in the ocean content of anthropogenic CO2 (Cant) between 1994 and 2007 as described in detail in Gruber et al. [2019] (Science). These estimates have been derived from the GLODAPv2 product [Olsen et al., 2016] using the eMLR(C*) method developed by Clement and Gruber [2018]. This method is based on the eMLR method [Friis et al., 2005], which determines the change in Cant on the basis of linear regression fits to data from two different time periods (here the JGOFS - WOCE era (~1994) and the Repeat Hydrography - GO-SHIP era (~2007)). The dataset contains in addition to the standard estimate also the estimates of 13 sensitivity cases, where different elements of the estimation procedure were changed to assess the robustness of the estimates. All estimates are given on 1° x 1° resolution. Two files are provided, i.e., one containing the full three-dimensional distribution of the change in Cant between 1994 and 2007 and one containing the vertically integrated values, i.e., the column inventories. These data provide strong constraints on the role of the ocean as a sink for anthropogenic CO2, and given the global nature of our assessment also constraints on the global carbon budget, specifically the magnitude of the land carbon sink. The estimates will prove also useful to assess ocean acidification and evaluate ocean models with regard to their carbon uptake and storage.

Further details are in Gruber, N. et al. (2019), The oceanic sink for anthropogenic CO2 from 1994 to 2007, Science (80-. )., (in press).


The observation-based estimates of the change in anthropogenic CO2 in the ocean between 1994 and 2007, ∆tCant were generated using the eMLR(C*) method introduced and described in detail by [Clement and Gruber, 2018]. Briefly, this method starts with dissolved inorganic carbon (DIC) observations from GLODAPv2 [Olsen et al., 2016] and then proceeds in three steps. In the first step, the semi- conservative tracer C* = DIC – rC:P PO43- – 1/2 (Alk + rN:P PO43-) is computed from the measured DIC, Alk and phosphate (PO43-) assuming a constant stoichiometric C:P and N:P ratio during photosynthesis and respiration/remineralization (rC:P = 116:1, rN:P = 16:1). In the second step, all data are allocated to either the first era (1981 through 1999) or the second one (2000-2014) and then adjusted to the mid-year of the respective era (1994 or 2007) using a transient steady state assumption. In the third step, the data from both eras are fitted using multiple linear regression (MLR) models and then combined to form the eMLR-based estimate of the change in ∆tCant. To this end, the data are first binned into ranges of neutral density, i.e., isoneutral slabs: 14 for the Atlantic and 12 for the combined Indo-Pacific ocean basins. Then separate multiple linear regression models are determined for each era and each bin. The regression coefficients from the MLR models for each era are then combined to generate the eMLR equation for each bin. These eMLR equations are then used together with the climatological distribution of the independent tracers to map out ∆tCant to the global ocean. A large suite of sensitivity studies were run in order to explore the robustness of the estimated ∆tCant to a range of subjective choices in the full estimation procedure (see Table 1 in the README file)

Description of NetCDF files

File 1:

This file contains the full 3D field of the standard estimate as well as those of the 13 other cases considered. Specifically, the netcdf file contains the grid coordinates. (see the file content in the README_DCant_eMLR(Cstar)_1994-2007.pdf file). All variables are provided on a 1° latitude by 1° longitude grid with 33 uneven depth levels. The DCant_XX variables contain the estimated ∆tCant using the version coding provided in the table below. All DCant variables have units of μmol kg-1. V01 is the standard case version, with the corresponding 1 sigma uncertainty from the Monte Carlo Simulation given by sigma_DCant_01. See the note for the use of the different versions.

File 2:

This file contains the vertically integrated changes in Cant between 1994 and 2007, i.e., the changes in the column inventory of anthropogenic CO2 (0-3000m). The file contains the estimates for the standard version (V101) as well as those for the 13 other cases considered. As file 1, the netcdf file contains the grid coordinates (see the file content in the README_DCant_eMLR(Cstar)_1994-2007.pdf file). All variables are provided on a 1° latitude by 1° longitude grid. The DCant_INVXX variables contain the vertically integrated changes in Cant using the version coding provided in the table below with units of mol m-2. V01 is the standard case version, while DCANT_INV_ENS_MEAN gives the mean across all 14 estimates and DCANT_INV_ENS_STD the associated standard deviation. See the note below for the use of the different versions.

Note for users

For most purposes, the standard version V01 is the best one to be used. However, the upper ocean estimates of this version exhibit some discontinuities in the high latitudes owing to the way the near surface changes in ∆tCant are estimated there, i.e., by extrapolating the interior ocean eMLR equation to the surface for each isoneutral slab. Thus, any user who would like to have a smooth upper ocean distribution, e.g., for model initialization, is likely better off by using the results from Version V12, where the entire surface ocean change is estimated on the basis of an equilibrium assumption.


Inquiries should be sent to Nicolas Gruber: The Matlab code required to generate the estimates is available upon request.


  • Clement, D., and N. Gruber (2018), The eMLR(C*) Method to Determine Decadal Changes in the Global Ocean Storage of Anthropogenic CO 2, Global Biogeochem. Cycles, 32(4), 654–679, doi:10.1002/2017GB005819.
  • Olsen, A. et al. (2016), The Global Ocean Data Analysis Project version 2 (GLODAPv2) – an internally consistent data product for the world ocean, Earth Syst. Sci. Data, 8(2), 297–323, doi:10.5194/essd-8-297-2016.
  • Last modified: 2021-03-17T18:30:28Z