Product name | OceanSODA-ETHZ-HR | |
SOCOM-style name | OSETHZ-CarbNN | |
Product version | v2023.prerelease | |
Coverage | 1982-01-01 — 2022-12-31 | Global ice free ocean |
Resolution | 8-daily | 0.25° x 0.25° |
Contact | nicolas.gruber@env.ethz.ch | OR luke.gregor@usys.ethz.ch |
DOI | available on publication |
Each netCDF contains only one variable. The metadata also has
a start and end time as well as min/max lat/lon.
The information about the settings used in PyCO2SYS is also
included in each file’s metadata.
Folder | Variable name | Units | Description |
---|---|---|---|
fgco2 |
FCO2 | mmolC m−2 day−1 | Sea-air CO2 flux |
fgco2_global |
globally integrated FCO2 | PgC yr−1 | Surface - atmosphere ocean fugacity of CO2 |
dfco2 |
ΔfCO2 | μatm | Surface - atmosphere ocean fugacity of CO2 |
sfco2 |
fCO2sea | μatm | Surface ocean fugacity of CO2 |
kw |
kw | cm hr−1 | Gas transfer velocity |
sol |
K0 | mol L−1 atm−1 | Solubility of CO2 in seawater |
ice |
ice | fraction | Sea-ice fraction |
The folder structure is
./<variable>/<year>/<single_netcdf_for_each_time_step>.nc
.
You can see all the variables on the HTTPS server – includes
temperature and salinity.
The data runs from 1982 to the end of 2022 with 46 time steps
for each year. The size for each variable (1982-2022) is between
2.5 and 3.5 GB.
This dataset contains a global gridded dataset of sea-air carbon dioxide fluxes (FCO2) for subseasonal to decadal studies. The product is available from 1982 through 2022 and is at an 8-daily by 0.25° by 0.25° resolution. FCO2 is calculated using the bulk formulation of flux:
(fCO2sea−fCO2atm) ⋅ kw ⋅ K0 ⋅ (1−ice)
where (fCO2sea − fCO2atm) is the sea-air difference in fugacity of carbon dioxide, kw is the gas transfer velocity, K0 is the solubility of CO2 in seawater, and ice is the sea-ice fraction.
The surface ocean fugacity of carbon dioxide (fCO2sea) is calculated using the surface Ocean Carbon dioxide Neural Network (OceanCarbNN). This two-step approach first estimats an 8-daily climatology of fCO2sea based on climatological drivers. The second step uses the output of the first step as a prior estimate of fCO2sea which is then adjusted to a time-varying product. An important novelty of this method is that the first step uses Gradient Boosted Trees (low bias, low variance, but unable to extrapolate) while the second step uses Neural Networks (able to extrapolate, but higher variance). We use 35 ensemble members to make these estimates. The average of the ensemble members is used as the final estimate of fCO2sea. Further, the standard deviation of the ensemble members can be scaled (× 5.7) to be used as an analog of the uncertainty. For more information see Gregor et al. (202X).
The atmospheric fugacity of carbon dioxide (fCO2atm) is calculated using the NOAA ESRL marine boundary layer product, ERA5 sea level pressure. kw is calculated using the Wanninkhof (2014) formulation, but we adjust the coefficient of gas transfer for the ERA5 wind reanalysis product (0.271 from Fay and Gregor et al., 2021). K0 is calculated using the Weiss (1974) formulation. For temperature and salinity data, please see the Gantt Chart below: