Data supporting "A comprehensive analysis of air-sea CO2 flux uncertainties constructed from surface ocean data products"

Data included in this repository supports the manuscript "A comprehensive analysis of air-sea CO₂ flux uncertainties constructed from surface ocean data products".

Two files are present:

1. A Python config file used to run the software developed for the analysis (Ford et al., 2024)
2. A ZIP file containing the input, neural network, and output files for the analysis.

Within the ZIP file, multiple folders are present:

1. Decorrelation contains .csv files that contain the annual estimates of the decorrelation lengths for the parameters requiring these (SST, sea ice, wind, fCO₂ and fCO₂ network).
2. Flux contains the individual FluxEngine output files that provide all the flux calculations, and auxillary data to the flux calculations.
3. Fluxengine_input contains the input files to FluxEngine, which specifies the fCO_{2 (sw),} xCO_{2 (atm)} and the temperature, salinities for the skin and subskin layers.
4. Inputs contains all the monthly 1 degree input data used. Many of the data used are not native monthly 1 deg, and so these are generated from the higher resolution data. These are all combined into the neural_network_input.nc file, so a single file can be distributed with all the inputs used.
5. Networks contains the TensorFlow neural network (FNN) files, where each province has 10 folders (one for each ensemble).
6. Plots contains output plots for debugging and final plots of uncertainties
7. Scalars contains the scalars used to normalise the data before input into the neural network. These are saved as Python pickle files, as they are needed if the neural network is used on other data.
8. Unc_lut contains the look up tables to generate the parameter uncertainty as described in the manuscript. These are Python pickle files.
9. Validation contains a csv file with the independent test RMSD, along with Python Pickle files of the validation data.

In the main folder, three files are present:

1. Annual_flux.csv contains the annual air-sea CO₂ flux (or ocean sink estimate) estimated from the fCO_{2 (sw)} fields. This also contains the annual integrated uncertainties for each component in the uncertainty flow chart in the manuscript.
2. Output.nc contrains the gridded global fields of the fCO_{2 (sw)}, the air-sea CO₂ flux, and the uncertainties for all the individual components. Metadata within the file should provide all the information required.
3. Training.tsv contains the training/validation data alongside the input parameters for neural network training

Please contact Daniel J. Ford (d.ford@exeter.ac.uk) if you have any questions.

Acknowledgements

This work was funded by the Convex Seascape Survey (https://convexseascapesurvey.com/) and the European Union under grant agreement no. 101083922 (OceanICU; https://ocean-icu.eu/) and UK Research and Innovation (UKRI) under the UK government’s Horizon Europe funding guarantee [grant number 10054454, 10063673, 10064020, 10059241, 10079684, 10059012, 10048179]. The views, opinions and practices used to produce this dataset/software are however those of the author(s) only and do not necessarily reflect those of the European Union or European Research Executive Agency. Neither the European Union nor the granting authority can be held responsible for them.

The Surface Ocean CO₂ Atlas (SOCAT) is an international effort, endorsed by the International Ocean Carbon Coordination Project (IOCCP), the Surface Ocean Lower Atmosphere Study (SOLAS) and the Integrated Marine Biosphere Research (IMBeR) program, to deliver a uniformly quality-controlled surface ocean CO₂ database. The many researchers and funding agencies responsible for the collection of data and quality control are thanked for their contributions to SOCAT.

References

Ford, D. J., Blannin, J., Watts, J., Watson, A. J., Landschutzer, P., Jersild, A., & Shutler, J. D. (2024, June 30). OceanICU Neural Network Framework with per pixel uncertainty propagation (v1.1) (Version v1.1). Zenodo. https://doi.org/10.5281/ZENODO.12597803