Tuning ocean biogeochemistry for an Earth system model: approaches, challenges and impact on model performance

Abstract. The parameterisation of biogeochemical models, when simulated within global ocean models, poses many challenges, among them those related to the calibration of rate constants (parameters) to the different functional groups. Global coupled general circulation models are computationally expensive to run, which limits the number of experiments and increases the risk of issues such as unrealistic tracer distributions. Efficient "surrogate'' circulations and methods to accelerate the spin up of global models have now become available, and provide the possibility to simulate a biogeochemical model globally with only moderate computational resources.  When coupled to an optimisation algorithm these surrogate models even allow the calibration of model parameters against observations in a coherent and systematic way. We here investigate whether biogeochemical model parameters, that were objectively calibrated (optimised) in an efficient surrogate circulation against a wide range of observations, can be transferred to the same biogeochemical model embedded in an Earth system model (ESM), without loosing too much of the model's improvement through optimisation. Comparison of biogeochemical results obtained from two circulation model environments shows that insolation as well as circulation can play a role especially for simulated primary production, and, regionally, surface concentrations such as chlorophyll; however, comparison to an earlier biogeochemical setup of the ESM shows that biogeochemical model parameters play an even larger role on certain biogeochemical processes. Export production is mainly affected by circulation, in line with earlier model studies. Deep particle flux is, however, mostly affected by the biogeochemical model parameters, especially the particle flux length scale. Given the relatively large role of biogeochemical model parameters, their optimisation in a surrogate circulation and subsequent transfer to the ESM leads to an improved performance of the ESM. This suggests that prior calibration of parameters in "light-weight'' circulations can support model development, before performing computationally expensive simulations with ESMs. Model improvement with respect to organic matter, for which only sparse observations exist, becomes most visible when applying non-parametric metrics, that avoid interferences from data patchiness and episodic events. These metrics, which also allow for a clearer distinction between the performance of different biogeochemical models, merit further exploitation with regard to model assessment and calibration.