Reconstruction of δ13CDIC in the Atlantic Ocean: A Probabilistic Machine Learning Approach for Filling Historical Data Gaps

The data description paper is submitted to the journal Earth System Science Data. Gao, H., Wu, Z., Sun, Z., Cai, D., Jin, M., and Cai, W.-J.: Reconstruction of δ13CDIC in the Atlantic Ocean: A Probabilistic Machine Learning Approach for Filling Historical Data Gaps, Earth Syst. Sci. Data Discuss. [preprint], https://doi.org/10.5194/essd-2025-517, in review, 2025. OverviewThis Zenodo archive contains two complementary datasets related to oceanic dissolved inorganic carbon isotope (δ¹³C_DIC) measurements and reconstructions in the Atlantic Ocean. The datasets serve distinct purposes, retain different origins, and align with community standards (e.g., GLODAPv2) to ensure interoperability and usability. Below is a detailed description of each dataset, their structures, variables, and intended applications. Dataset 1: GPR-Reconstructed δ¹³C_DIC DatasetName:GLODAPv2.2023_Atlantic_Ocean_with_Reconstructed_d13C Origin and Purpose:This dataset is derived from the GLODAPv2.2023 Atlantic Ocean subset, with the primary goal of supplementing the original dataset with machine learning (GPR)-reconstructed δ¹³C_DIC values for sparse or missing observations. It is designed to complement GLODAPv2’s existing basin-specific products and facilitate direct comparison with the original δ¹³C_DIC measurements. Structure and Compatibility:The dataset retains all original variable names, units, and data structures of GLODAPv2.2023’s Atlantic subset, ensuring full interoperability with GLODAPv2-standard tools and workflows. No modifications are made to the original GLODAPv2 variables. Key Added Variables:ReC13: Reconstructed δ¹³C_DIC value (unit: ‰, relative to VPDB).ReC13f: Quality flag for the reconstructed δ¹³C_DIC (consistent with GLODAPv2 flag conventions: 2 = good data, 3 = questionable data, 9 = missing data). Dataset 2: Atlantic Observational Compilation DatasetName:Atlantic_cruises_with_c13 Origin and Purpose:This dataset is a comprehensive compilation of raw and reprocessed δ¹³C_DIC observational data from 51 Atlantic cruises (see Table 1 in the associated manuscript). It includes data from both GLODAPv2-included cruises and non-GLODAPv2 cruises, serving as a supporting dataset for the GPR reconstruction (not a final reconstructed product). Structure and Variables:Unlike Dataset 1, this dataset is not strictly constrained to GLODAPv2 conventions, as it integrates non-GLODAPv2 cruises with limited ancillary data. For non-GLODAPv2 cruises, key biogeochemical variables (e.g., apparent oxygen utilization [AOU], pH) were calculated using the seawater, GSW, and CO2SYS toolboxes. As a result, some variable names and units may differ slightly from GLODAPv2 standards. Key Unique Variable:adjusted_C13: Reprocessed δ¹³C_DIC value (unit: ‰, relative to VPDB) after applying inter-cruise adjustments (e.g., instrument drift corrections) and secondary quality control. Citation GuidelinesTo ensure proper attribution, users must cite the appropriate sources based on the data used:Lauvset, S. K., Lange, N., Tanhua, T., Bittig, H. C., Olsen, A., Kozyr, A., Álvarez, M., Azetsu-Scott, K., Brown, P. J., Carter, B. R., Cotrim Da Cunha, L., Hoppema, M., Humphreys, M. P., Ishii, M., Jeansson, E., Murata, A., Müller, J. D., Pérez, F. F., Schirnick, C., Steinfeldt, R., Suzuki, T., Ulfsbo, A., Velo, A., Woosley, R. J., and Key, R. M.: The annual update GLODAPv2.2023: the global interior ocean biogeochemical data product, Earth Syst. Sci. Data, 16, 2047–2072, https://doi.org/10.5194/essd-16-2047-2024, 2024.Gao, H., Wu, Z., Sun, Z., Cai, D., Jin, M., & Cai, W.-J.: Reconstruction of δ¹³CDIC in the Atlantic Ocean [Data set]. Zenodo. https://doi.org/10.5281/zenodo.16907402, 2025. Contact InformationFor questions regarding dataset structure, variables, or usage, please contact:[e-mail]