Mete and colleagues (2023, see reference below) used Machine Learning (ML) to predict the global distribution of oceanic barium (Ba). Models were first trained to predict [Ba] from standard oceanographic observations using GEOTRACES data from the Arctic, Atlantic, Pacific, and Southern oceans. Model predictions of [Ba] were then compared with actual [Ba] data from the Indian Ocean, with the best models achieving a mean absolute percentage error of just 6.0 %. This successful comparison allowed the authors to calculate the global distribution of [Ba], Ba*, and marine barite saturation using data from the World Ocean Atlas. This approach revealed four significant findings: 1) the ocean contains 122±7 Tmol of dissolved Ba; 2) the variability in the barium–silicon relationship is consistent with the biogeochemical characteristics of both elements; 3) marine barite saturation exhibits systematic spatial and vertical variations; 4) taken as a whole, the ocean below 1000 m is at equilibrium with respect to barite. These results have broad implications, both for the modern ocean and for interpreting paleo-records of barite. A data product, which includes a global grid of predictions and the ML model itself, is freely available from BCO-DMO: https://www.bco-dmo.org/dataset/885506.
Mete, Ö. Z., Subhas, A. V., Kim, H. H., Dunlea, A. G., Whitmore, L. M., Shiller, A. M., Gilbert, M., Leavitt, W. D., & Horner, T. J. (2023). Barium in seawater: dissolved distribution, relationship to silicon, and barite saturation state determined using machine learning. Earth System Science Data, 15, 4023–4045. Access the paper: 10.5194/essd-15-4023-2023