Although the carbonate chemistry and physical aspects of ocean acidification are well constrained, its biological effects are not fully understood. Experimental research has shown large variability in responses to increased atmospheric CO2 input into the ocean, ranging from positive to zero and negative effects. Global models vary strongly in their results for particular parts of the global ocean, for instance in carbon export. This large inter-model variation and inconsistency with data reflect an insufficient understanding or representation of the biology, effectively reducing predictive capabilities. Sensitivity analyses and inter-model comparison allow for the identification of weaknesses in the parametrization and model structure.
We investigate the performance of an existing mesocosm-based ecosystem model developed by Schartau and co-workers (Biogeosciences, 2007), which allows for a decoupling of carbon and nitrogen dynamics. Our in-depth analysis demonstrates to what extent the model can be calibrated to a given dataset of measurements. Local sensitivity analyses enable us to identify important parameters that deserve more attention in experimental work. An overall evaluation of the precision of the model output, given the available data is provided by a global sensitivity analysis. Using datasets from different mesocom studies, the overall robustness of the model structure and parametrization is evaluated. At a later stage, this mesocosm-based ecosystem model could be integrated with a global ocean circulation model.