Mechanistic marine ecosystem models serve to synthesize diverse information about the interacting organisms and processes shaping marine ecosystems. However, using ecosystem models to provide mechanistic, quantitative explanations for observed patterns is made difficult due to uncertainty in model parameters and structure, the complex interdependencies between model elements, and the sparseness of data. These challenges are exacerbated when the model must match observations across a broad range of ecosystems.

To develop a globally robust representation of marine ecosystem dynamics, results of a size-structured marine ecosystem model are compared to global trends in: 1) phytoplankton biomass ; 2) phytoplankton turnover rates; 3) phytoplankton size-structure; 4) bacterial biomass; 5) bacterial turnover rates; and 6) zooplankton biomass. Steady-state results within idealized euphotic zones are first considered. Basic observed trends are apparent in the model solutions using mean parameter values, but notable biases also exist. The impacts of ~40 parameter and structural sensitivities within the model are thus diagnosed. Perturbation analyses are used to identify key parameters and processes and group those with similar dynamical impacts. Two largely orthogonal groups are revealed. The first group, which includes parameters describing zooplankton ingestion and phytoplankton nutrient uptake, modulates the phytoplankton size structure, phytoplankton turnover rates, and the ratio of bacterial to phytoplankton carbon. The second group, which includes parameters governing the balance between recycling and export, imposes correlated changes in steady-state biomass across functional groups, wherein biomass increases with more efficient recycling. A step-wise calibration using maximum likelihood is then performed by sequentially adding critical parameters identified by the sensitivity analysis. The largest model biases are removed after tuning 5-7 parameters, with diminishing improvement from tuning additional parameters thereafter. Finally, the calibrated model serves as a tool to diagnose the relative impacts of size-structure and temperature on the ratio of production to export.