Optimal control and system limitation in a Southern Ocean ecosystem model

Simple models of coupled nutrient-biological cycles are useful for understanding and evaluating biogeochemical processes in the marine environment. Indeed simple models offer advantages over more complex formulations in that they have fewer uncertain parameters. A common approach to calibrating such models is to optimise parameter values by minimising some measure of the difference between model outputs and observational data. Optimised parameters are usually invariant in a given simulation, but some recent implementations allow parameter values to vary as a function of time. These latter approaches are arguably more realistic in that they capture changes in key rate processes associated with changes in assemblage structure, such as seasonal succession in phytoplankton communities. We adopt an optimal control approach in which one model parameter is viewed as a time-varying function. By examining the conditions under which our 'controlled' model is able to realistically reproduce observed dynamics in remotely sensed surface chlorophyll, we gain insights into other critical parameters that influence model behaviour. This approach has advantages over conventional optimisation in which the relative importance of different modelled processes is potentially unclear. We find that zooplankton control of phytoplankton biomass through grazing is critical to our model's ability to realistically capture seasonal phytoplankton dynamics for two locations along a Southern Ocean transect at 140°E; the middle of Sub-Antarctic Front (52°S) and the middle of the Polar Front (58°S).