Southern Ocean CO2 Machine Learning products (SOCOML)
Description
We present a comprehensive, quality-controlled reconstruction of key carbonate system parameters in the Southern Ocean interior—including total alkalinity (TA), dissolved inorganic carbon (DIC), pH (total scale), nitrate (NO3), phosphate (PO4), silicate (SiO4), anthropogenic carbon (Cₐₙₜ), and aragonite saturation (Ωₐᵣ)—by leveraging machine learning techniques (ESPER_NN) and integrating all available Argo float profiles with ship-based survey data. The resulting datasets are gridded at 1°×1° horizontal resolution and 84 vertical pressure levels (0-5,600 dbar), and are provided as distinct climatological products: the Float Grid (using all Argo float profiles) and the All-Data Grid (integrating all available Argo and ship-based observations). The Float Grid is further separated into the Non-O₂-Float Grid (limited to Core Argo floats) and O₂-Float Grid (limited to oxygen-measured Biogeochemical Argo floats). Each gridded product is accompanied by uncertainty estimates. The climatological products covers nearly the whole Sothern Ocean based on direct measurements instead of applying interpolating mapping methods, thereby providing a more robust result. Model performance is assessed through cross-comparison of Argo and shipboard measurements.