Global sensitivity analysis of a one-dimensional ocean biogeochemical model

Ocean biogeochemical (BGC) models are a powerful tool for investigating ocean biogeochemistry, marine ecosystem functioning, and the global carbon cycle. These models contain numerous uncertain parameters. The uncertainty of these parameters is substantial, and in turn, translates into possibly significant uncertainty in the model outputs. In this study, we perform a global sensitivity analysis (GSA) of an ocean BGC model to identify the parameters whose uncertainty has the largest impact on the variability of model outputs. We consider the BGC model Regulated Ecosystem Model 2 in a one-dimensional configuration at three locations of the global ocean. We compute variance-based Sobol's sensitivity indices and Shapley effects to assess of the most influential parameters for each location with respect to quantity of interest that are commonly considered for the calibration and validation of BGC models. The most important parameters for chl-a simulations are the parameters related to nitrogen uptake by phytoplankton, excretion of organic nitrogen by both phytoplankton and zooplankton and grazing by zooplankton, and for net primary production the most influential parameters are those related to photosynthesis, excretion of organic carbon by phytoplankton and grazing by zooplankton across all three sites. Export production and CO2 flux are influenced mainly by the remineralization of nutrients and grazing. Furthermore, the Shapely effects indicate that some parameters have mutual dependencies. Our results suggest that the implementation of multiple zooplankton function types in BGC models may be useful to improve chl-a and NPP prediction, provided that efforts are invested in estimating parameters characterizing the grazing in marine ecosystems. Our results also indicate that the implementation of heterotrophic bacteria explicitly in the model has the potential for better simulation of export production and CO2 fluxes. Despite the limitation of the one-dimensional model configuration, our application offers an objective list of the most important biogeochemical parameters that need to be quantified for future applications of a global configuration. The insight gained from the GSA will be broadly applicable in future BGC modeling case studies, and BGC model developments.