PAR-GWR-CARB: A monthly gridded surface carbonate system data product for the Pacific-Arctic Region from 1982-2021 (NCEI Accession 0292960)

The Pacific-Arctic Region (PAR) is highly sensitive to ocean acidification, yet data gaps in the carbonate system record persist due to the region's remote nature and environmental factors such as sea ice. To address these gaps, we utilized a geographically-weighted regression (GWR) to predict surface total alkalinity using salinity and temperature data at a monthly 1°x1° resolution. The GWR methodology, which allows regression coefficients to vary spatially, accounts for local factors such as riverine discharge, sea ice melt, primary productivity, and other influences on alkalinity. This approach achieved a high degree of accuracy, with an R² of 0.93, a mean absolute error (MAE) of 23 μmol/kg, and near-zero bias (mean bias error [MBE] of -0.01 μmol/kg), as validated by 10-fold cross-validation (N = 7,146). The derived regression coefficients were then applied to the Hadley EN4 salinity and OI SST temperature datasets to estimate total alkalinity. Subsequently, this total alkalinity dataset was integrated with the MPI-SOM-FFN surface pCO2 dataset to reconstruct the entire carbonate system (DIC, pH, [H⁺], Ω) from 1982 to 2021 at a monthly 1°x1° resolution.