Evaluating multi-source uncertainties in marine biogeochemical modelling: a case study on the Eco3M-Med model

International audience ; Coupled with ocean and atmosphere models, marine biogeochemical models are key tools for studying marine ecosystems, in particular the dynamics of planktonic food webs and biogeochemical cycles. They also provide possible long-term trajectories of the evolution of marine ecosystems under SSP–RCP scenarios. At the same time, they provide forcing in terms of available prey for high trophic level models and for fisheries dynamics. The evolution of biogeochemical models from early NPZD-type versions to more complex structures that incorporate plankton functional groups and variable stoichiometry within organisms has led to a rapid increase in the number of state variables and parameters, as well as to higher computational cost. This limits the ability to explore a large number of scenarios and to quantify model uncertainties, given the large number of ensemble simulations required.In this study, we propose a framework for “application-targeted model simplification” and for “assessing the parameter uncertainties introduced before and after model simplification.” The simplification is carried out using a semi-automatic iterative simplification framework, which we apply to the most recent version of the biogeochemical model Eco3M-MED in a 1DV configuration. We first perform a local sensitivity analysis by perturbing only a subset of independent parameters and by accounting for the dependencies between parameters, which allows us to identify parameters and processes with low impact on the targeted model outputs. We then simplify either the model structure or the identified processes. At each step, the fidelity of the simplified model relative to the reference model is evaluated using the Kling–Gupta Efficiency metric, and we ultimately obtain two simplified models corresponding to two different simplification objectives.After obtaining the simplified models, we perform a parameter uncertainty analysis to test whether simplification significantly changes the uncertainty range of key marine ...