(Iris Kriest, Markus Schartau)

Marine biogeochemical  (BGC) models are based on many assumptions concerning model structure, equations, and the associated constants (parameters), which determine the simulated responses of organisms to environmental or ecological changes. Because of the highly aggregated form of BGC model components and the flexibility of the organisms’ responses to environmental changes, the simulations are generally associated with large uncertainties. The specification and quantification of these uncertainties, and their reduction through model calibration, are investigated with different methods, while covering a wide range of spatial and temporal scales.

Model calibration and uncertainty quantification require the definition of metrics that assess and quantify model performance with regard to experimental or ocean data (obtained from, for example, remote sensing or in situ measurements). We are interested in the advantages and limitations of applying specific metrics for data-model comparison. Our work involves close collaboration with experimental and observational scientists, mathematicians and computer scientists.

Contributors:

Iris Kriest (Examination of global BGC model sensitivities to biogeochemical parameterisations; global BGC model assessment and parameter optimisation against various observations)

Markus Schartau (Parameter estimation for dynamical and diagnostic modelling approaches; non-parametric metrics for assessing model performance with respect to particle and plankton size spectra as well as biogeochemical tracers; exploring manifold of model solutions that can explain data equally well)

Mariana Hill (Calibration with evolutionary algorithm of individual-based fish model OSMOSE against observed biomass and landings to find optimal sets of parameters)

Volkmar Sauerland (Calibration of global models of ocean BGC; tailored variants for multiple objectives and stochastic parameters; development of metrics; derivative-free optimisation tools)

Giang Tran (Model-based assessments of climate engineering; uncertainty and sensitivity analysis using Gaussian Process emulation)