Modelling primary production: multitude of theories, or multitude of languages?

Marine primary production, converting approximately 50Gt of inorganic carbon into organic carbon per year, is an important component of the global carbon cycle, and a major determinant of past, present and future climate. Large-scale, long-term estimates of marine primary production rely primarily on two types of models: satellite-based models that make extensive use of remote-sensing data, and ecosystem models providing numerical simulation of ecological processes embedded in general ocean circulation models. Intercomparison exercises of model outputs (both within and across the two model types) have consistently revealed high discrepancies between estimated global ocean primary production, including divergent magnitudes and even opposite trends. Model-observation comparisons are also complex, differences in measurement techniques, and evolving methodologies could all lead to difficulties with the interpretation of results. These uncertainties limit the applications of primary production models (both satellite-based and ecosystem), especially in the climate context, where an important question is whether climate change will drive significant future changes in regional or global primary production. Both satellite-based and ecosystem models rely on a range of fixed model parameters, whose values need to be carefully estimated and tested. In this paper, we suggest that such model parameters represent an underappreciated but important source of inter-model differences. With the proliferation of both satellite and in situ observations of relevant variables at global scales, and the availability of powerful statistical tools such as data assimilation and machine learning, we argue that time is right to systematically examine model parameters, gaining both better insights into parameter values and how those values might vary in space and time. We argue that such spatio-temporal parameter variability can be theoretically justified for ecosystem models with complexity similar to those commonly used within Earth System Models (ESMs) in climate studies. The spatially and temporally varying parameter values could serve to unify models that are structurally different. An important aspect of this unification could be the ability to infer the spatio-temporal variability of parameters in the less complex models from the emergent behaviour of the more complex ones. This could include ecosystem model simulations of nutrients, temperature, phytoplankton classes, or vertical distributions informing satellite-based models. We conclude that better understanding of model parameter roles and integration (or inter-calibration) of different types of models could reduce discrepancies among the primary production models and improve the reliability of marine primary production projections.