Key points are not available for this paper at this time.
Abstract. The Southern Ocean plays a crucial role in regulating atmospheric carbon dioxide (CO2) concentrations and modulating the global oceanic carbon cycle, thereby substantially mitigating the effects of anthropogenic climate change. However, due to the region's challenging environment and sparse observational coverage, large uncertainties remain regarding the magnitude and mechanisms of carbon uptake in the Southern Ocean. In recent decades, the deployment of Argo float arrays has facilitated autonomous and continuous profiling of hydrographic and biogeochemical properties from the surface to depths of up to 6000 m, complementing traditional ship-based observations. Nevertheless, high-resolution, integrated datasets that combine ship-based and Argo-derived observations remain rare, partly due to the challenges of data harmonization, quality control, and uncertainty estimation, as well as the indirect nature of carbonate system parameter retrievals from Argo measurements. Here, 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 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 (0–5600 m), 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 cover 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. The gridded products, collectively termed SOCOML (Southern Ocean CO2 Machine Learning products, https://doi.org/10.17632/xzr59ngmpz.2, Zhong et al., 2025a; https://doi.org/10.25921/8c29-rv75, Zhong et al., 2025b), are freely available for downloaded and are expected to support future studies of Southern Ocean carbon cycle.
Zhong et al. (Mon,) studied this question.