Defining misfit between biogeochemical models and data sets

Marine biogeochemistry deals with diverse variables, many of them (especially biological rates) difficult to measure. For parameter estimation and data assimilation, differences between predicted and measured values of these variables must be summarized by a single misfit number. A ‘correct’ measure of misfit incorporates choices about issues that include (1) the relative importance of deviations from a small and a large predicted value; (2) the relative importance of small and large deviations from a predicted value; (3) the relative importance of deviations above and below a predicted value; (4) comparing discrepancies measured in intrinsically different units, like concentrations and rates. Recent papers on estimating ecological parameters represent a variety of different choices. This paper uses a data set and model for the subtropical North Atlantic to demonstrate that the choices made for defining misfit can have a large influence on the ‘best’ estimate of biogeochemically important fluxes and concentrations.