A model-based assessment of observational methods to estimate anthropogenic CO2
Three methods to estimate anthropogenic CO2 in the ocean (deltaC* method, TTD (transit time distribution) method and taking the difference of two DIC equilibrium concentrations), which are widely used in observations, were adapted to a coarse resolution (4/3°) ocean general circulation model of the North Atlantic. This is done for testing the methods and comparing them to anthropogenic CO2 modelled directly in this model. Furthermore two methods for modelling anthropogenic CO2 are compared in a climatological forced model run. Modelling anthropogenic CO2 as a passive tracer with a surface flux containing the information of anthropogenic pCO2 perturbation leads to 15% to 20% lower concentrations than the “true” anthropogenic CO2 of the model. This is computed as the difference of DIC concentrations from a model run with realistically prescribed atmospheric pCO2 concentrations minus a model run with a constant preindustrial atmospheric pCO2 concentration, where in both cases a NPZD (nutrient, phytoplankton, zooplankton, detritus) model is coupled to the physical model. For correcting this underestimation by the passive anthropogenic CO2 tracer an improvement to model this tracer is suggested. Adapting a method which uses an age-spectrum (TTD method) or directly modelled water age (difference of two DIC equilibrium concentrations) of a water patch for estimating anthropogenic CO2, leads to concentrations, which are close to the “true” modelled anthropogenic CO2 concentrations. In contrast to that, estimating anthropogenic CO2 by using a method which needs a single chlorofluorocarbon (CFC) derived water age (deltaC* method), leads to high overestimation of the “true” anthropogenic CO2 of the model, especially in the deep ocean. One assumption of the deltaC* method is that anthropogenic CO2 is transported predominantly along isopycnal surfaces. To test this, a sensitivity model run excluding isopycnal mixing was performed. Especially estimating anthropogenic CO2 with the deltaC* method using the output of this model run should fail. However, the results in comparison to the modelled anthropogenic CO2 are differing in nearly the same magnitude as for the model run including isopycnal mixing. The same is the case for the other two methods (TTD method and taking the difference of two DIC equilibrium concentrations) for estimating anthropogenic CO2. Another sensitivity run forced by inter-annual variable surface fluxes was performed in order to determine the influence of inter-annual variability on the estimates of anthropogenic CO2 by using the different methods. In this case the inter-annual variability of anthropogenic CO2 by using the deltaC* method is of the same magnitude as the difference to the “true” modelled anthropogenic CO2 itself. The inter-annual variability of the total amount of anthropogenic CO2 deviates by just about 1% from the climatological reference experiment. Locally variations of up to 10% compared to the climatological mean occur.