Data and Analysis Products for: Low-pressure storms drive nitrous oxide emissions in the Southern Ocean

This repository contains input datasets, analysis products, and figure data associated with the machine learning-based estimation of nitrous oxide (N₂O) air-sea fluxes in the Southern Ocean, as described in Kelly et al. (2025). Contents: Input Datasets: goshipdataset.parquet/.nc/.csv – GO-SHIP N₂O observations used for Random Forest model training, containing temperature, salinity, dissolved oxygen, nitrate, and partial pressure of N₂O (pN₂O) measurements argodataset.parquet/.nc/.csv – BGC-Argo float profiles (October 2024 snapshot) with interpolated ERA5 wind speed, sea level pressure, and sea ice data from the Southern Ocean fronts_Gray.mat – Southern Ocean frontal positions from Gray et al. (2018) Model Outputs: n2opredictions.parquet – Predicted pN₂O from BGC-Argo floats generated by applying four Random Forest models, including uncertainty estimates fluxtests/ (zip archive) – Air-sea N₂O flux calculations and Monte Carlo uncertainty propagation results Figure Data: fig1.csv through fig5column2.csv – Processed data underlying Figures 1-5 in Kelly et al. (2025) Code Repository: The Python code to reproduce these datasets and generate figures is available at: https://github.com/ckelly314/ml-argo-n2o Citation: If you use these data, please cite: Kelly, C.L., Chang, B.X., Emmanuelli, A., Park, E., Macdonald, A., & Nicholson, D.P. (2025). Low-pressure storms drive nitrous oxide emissions in the Southern Ocean. Research Square (Preprint). https://doi.org/10.21203/rs.3.rs-6378208/v1 File Formats: Data are provided in multiple formats for accessibility: .parquet – Apache Parquet (optimized for Python/pandas) .nc – NetCDF (CF-compliant, optimized for xarray) .csv – Comma-separated values (universal compatibility) .mat – MATLAB format Acknowledgements: Training data were collected and made publicly available by the U.S. Global Ship-based Hydrographic Investigations Program (U.S. GO-SHIP; https://usgoship.ucsd.edu/) and the programs that contribute to it. BGC-Argo data were assembled or collected and made available by the Global Ocean Biogeochemistry Array (GO-BGC) Project funded by the National Science Foundation (NSF grant #OCE-1946578). NCEP-NCAR Reanalysis 1 data were provided by the NOAA PSL, Boulder, Colorado, USA, from their website at https://psl.noaa.gov. Development of the BGC-Argo parquet dataset used in this work was supported by NSF grant #OAC-2311383. C.L. Kelly was supported by a Doherty Postdoctoral Scholarship funded by the Woods Hole Oceanographic Institution and a U.S. GO-SHIP Postdoctoral Fellowship funded by the National Science Foundation (NSF grant #OCE- 2023545). B.X. Chang was supported by NSF grant #OCE-2048518. D.P. Nicholson’s contribution to this work was sponsored by the National Science Foundation’s Global Ocean Biogeochemistry Array (GO-BGC) Project under NSF grant #OCE-1946578 with operational support from NSF grant #OCE-2110258. A.M. Macdonald acknowledges support from NSF grants #OCE-2023545 and #OCE-1923387. B.X. Chang and A.F. Emmanuelli were supported by the Cooperative Institute for Climate, Ocean, and Ecosystem Studies (CICOES) Summer Internship Program (NSF grant #OCE-2048518). License: These data are released under a Creative Commons Attribution 4.0 International License (CC BY 4.0). Contact: For questions about these data, please contact Colette Kelly ([e-mail]) or visit https://github.com/ckelly314.