Model assessment and model-based data analyses of an ocean acidification mesocosm experiment

Ocean acidification (OA) has been dubbed as the “evil twin” of climate change. Studies suggest that OA has dramatic impacts on marine phytoplankton. Mesocosm facilities allow investigations on effects of changes in the carbonate chemistry of sea water on plankton communities in the vicinity of their natural habitats, e.g. Pelagic ecosystem CO2 enrichment (PeECE) studies. Marine ecosystem models serve as an efficient tool to analyse and interpret mesocosm data, as they use mathematical equations to describe processes controlling dynamics of planktonic ecosystems. The goal of this thesis is to investigate the effects of OA on phytoplankton growth dynamics by analysing data from an ocean acidification mesocosm experiment using different model approaches. To achieve this data assimilation (DA) methods are applied. These methods yield the optimised model solutions (with optimised parameter values) that maximize the likelihood probability of models explaining mesocosm data. In addition, DA methods estimate the ranges of uncertainty in optimised model parameter values. In the first study (Chapter 2), the performance of different metrics (cost functions) that maximize the predictive capability of a plankton model are evaluated. Next, an optimality-based model is applied to investigate the large observed variability in calcification and total alkalinity during the PeECE-I experiment (Chapter 3). The model considers an explicit CO2 dependency of calcification. Three model experiments are set up to simulate growth of bulk phytoplankton and coccolithophores in mesocosms with high, medium and low observed calcification rates. Skills of two plankton models (OBM and CN-REcoM) that differ in their mechanistic description of nutrient uptake and algal growth are assessed against mesocosm data in the last study (Chapter 4) of this thesis. In contrast to the calcification study, the plankton models that are applied in Chapter 4 do not resolve any CO2 effects on phytoplankton growth dynamics. The idea is to test whether this neglect of CO2 dependencies is revealed in differences of model parameter estimates between different CO2 treatments. According to DA results, the cost function that is derived from a probabilistic approach and accounts for changes in correlations between observations performs better as metric for model calibration than other types of cost functions (e.g. Root mean squared errors). The model-based data analysis of the PeECE-I experiment suggests that the large variability that was observed in calcification could have been generated due to small differences in initial abundance of coccolithophores during initialisation (filling) of mesocosms. A pattern is seen in the estimates of two physiological parameters, the potential carbon fixation rate (V C 0 ) and the subsistence quota (Qmin), between the CO2 treatments for the OBM. It predicts high estimates of V C 0 and Qmin for phytoplankton in mesocosms treated with high CO2 concentrations and vice versa for those in mesocosms with low CO2. The OBM seems to suggest that OA may enhance carbon fixation rates in phytoplankton, but at the cost of elevated metabolic stress. However, it is suggested to include mechanistic CO2 dependencies of nutrient uptake and phytoplankton growth in the OBM for future studies on OA.

Krishna S., 2018. Model assessment and model-based data analyses of an ocean acidification mesocosm experiment. PhD thesis, Christian-Albrechts-Universität. 143 p. Thesis.

  • Reset


OA-ICC Highlights

%d bloggers like this: