Comparison of two carbon-nitrogen regulatory models calibrated with mesocosm data

Marine phytoplankton can regulate their stoichiometric composition in response to variations in the availability of nutrients, light and the pH of seawater. Varying elemental composition of photoautotrophs affects several important ecological and biogeochemical processes, e.g., primary and export production, nutrient cycling, calcification, and grazing. Here we compare two plankton ecosystem models that consider regulatory mechanisms of cellular carbon and nitrogen, driving the physiological acclimation of photoautotrophs. The Carbon:Nitrogen Regulated Ecosystem Model (CN-REcoM) and the optimality-based model (OBM) differ in their representation of phytoplankton dynamics, i.e. nutrient acquisition, synthesis of chlorophyll a, and growth. All other model compartments (zooplankton, detritus, dissolved inorganic and organic matter) and processes (grazing, aggregation, remineralisation) remain identical in both models. We assess the skills of the two models against data from an ocean acidification mesocosm experiment with three CO2 treatments. Neither model accounts for any carbon dioxide (CO2) effects explicitly. Instead, we assimilate data of the different CO2 treatments separately into the models. Thereby we aim at identifying optimal model parameter values that might correlate with differences in CO2 conditions. For the OBM, optimal parameter estimates of Q(min) (subsistence N:C ratio) and V-C(0) (maximum potential photosynthesis rate of photoautotrophs) turned out to be higher for mesocosms exposed to high CO2 compared to those with low CO2 concentrations. By contrast, a similar correlation is not observed for the CN-REcoM. A possible physiological interpretation of higher estimates of Q(min), and V-C(0) according to the OBM is that phytoplankton may experience environmental stress under more acidic conditions, and hence must invest more energy/resources for maintaining basic cellular functions. Our data assimilation reveals that the parameters of the OBM are better constrained by the data than those of the CN-REcoM. Furthermore, the OBM is better able than CN-REcoM to reproduce data that were not used for parameter optimization.