Observation system simulation experiments in the Atlantic Ocean for enhanced surface ocean <i>p</i>CO<sub>2</sub> reconstructions

Abstract. To derive an optimal observation system for surface ocean pCO2 in the Atlantic Ocean and sector of Southern Ocean, 11 simulation experiments (OSSEs) were completed. Each OSSE is a feedforward neural network (FFNN) that based on different data distribution provides period 2008–2010 with 5 d time interval. Based geographical positions from three observational platforms, volunteering observing ships, Argo floats OceanSITES moorings, pseudo-observations constructed using outputs online-coupled physical–biogeochemical global model 0.25? nominal resolution. The aim this work was to find spatial observations supplement widely used Surface CO2 Atlas (SOCAT) improve accuracy reconstructions. OSSEs showed additional mooring stations improved coverage Hemisphere biogeochemical ARGO corresponding least 25 % density active (2008–2010) (OSSE 10) would significantly reconstruction reduce bias derived estimates sea–air fluxes by 74 compared outputs.