From a modelling point of view, plankton modelling is a data-poor subject. For even simple models, such as NPZ, there is little or no data for key processes and rates:    mortality,    zooplankton grazing,    remineralisation rates,    higher predator effects. Many ecosystem models have considerably more complexity than NPZ. Do they fit the data because they “get it right”, or because they have more degrees of freedom for “tuning”?