Monthly gridded sea surface fugacity of carbon dioxide (fCO2), total alkalinity (TA), dissolved inorganic carbon (DIC), pH, and aragonite (Ωarag) in Mid-Atlantic Bight and South Atlantic Bight from 1982 to 2015 (NCEI Accession 0208346)

This dataset contains monthly interpolated sea surface fugacity of carbon dioxide (fCO2) from 1982 through 2015 is estimated from monthly atmospheric CO2, temperature, and salinity. Firstly, the disequilibrium between the observed seawater fCO2 and fCO2air (i.e., ΔfCO2; ΔfCO2 = fCO2 – fCO2air) is estimated using a bayesian-neural-network approach for each 1°×1° grid box in the MAB and the SAB. The input parameters are latitude, longitude, SST, and SSS. The output parameter is ΔfCO2. The feedforward backpropagation network is constructed by two hidden layers with tanh activation functions. The neural network is trained with the Levenberg-Marquardt algorithm (Hagan and Menhaj, 1994). Then fCO2 is calculated using ΔfCO2 and fCO2air. Next, continuous monthly SST and SSS data from 1982 to 2015 are used to calculate output with the trained network to fill in SOCAT data gaps. Monthly pH, DIC, and Ωarag are calculated from fCO2 and salinity-derived TA using CO2SYS (Van Heuven et al., 2009) with the first and second dissociation constants of carbonic acid in seawater (K1 and K2) from Lueker et al. (2000) and borate-to-salinity ratio determined by Uppström (1974). TA is derived from salinity using their linear relationships in the MAB and the SAB (Cai et al., 2010).