An investigation of Diatom physiology and its effect on silicon trapping using biogeochemical models

Silicon (Si) distribution in the world’s oceans is biologically controlled by silica-shelled phytoplankton called diatoms, which contribute 20% of global primary productivity. Diatoms decouple Si from other macronutrients that are upwelled in the Southern Ocean (SO), trapping Si and limiting the growth of siliceous algae elsewhere. This is caused by high diatom Si:N uptake ratios under Fe deficiency in combination with low attenuation through the water column and deep circulation back to the SO. The way diatom physiology is parameterised in biogeochemical models can lead to contrary responses to Fe fertilisation that occurred in past glaciations. It is important to understand the effect of diatom physiology on Si trapping to ensure models can address past climates and future changing oceans. This study firstly investigates the core mechanisms of SO Si trapping by using a simple 3-box model to represent the overturning circulation between the deep ocean SO and subtropical ocean. The model reproduced expected nutrient concentrations for P, Fe and Si as well as distributions of diatoms and non-diatoms. However, the addition of an Fe-dependent or Fe and Si-dependent diatom Si:N ratio led to near complete Si trapping and sensitivity testing showed that parameters and initial conditions required dramatic alteration to allow Si to escape to the subtropics. A simulation of Fe fertilisation produced no increase in Si leakage as the decrease in Si:N was overtaken by the growth of diatom productivity. In the second part of this study, two models with different representations of diatom physiology were tested against the observations of a series of SO nutrient addition experiments. A quota model which allowed for luxury uptake of Si and emergent Si:N stoichiometry outperformed a simple model using direct parameterisations of Si:N. The winning model was adapted to include an additional Chl state variable and then optimised using a genetic algorithm to fit the parameters. The algorithm was able to dramatically reduce the deviation between the model and the experiments. However, a sensitivity test that performed 120 optimisations found that many parameters were unconstrained by the data. This led to the adoption of a hybrid approach where only well-constrained parameters were fitted by the algorithm. The hybrid approach resulted in only a small reduction in the fit of the model to the observations while hopefully avoiding overtuning and retaining a higher performance at broader scales. When used to fit the initial proportion of diatoms in the model, the algorithm correlated with the results of pigment data implying the importance of community structure in addition to Fe and Si concentrations. Overall, this study suggests that raised Si:N ratios in SO diatoms can drastically reduce Si leakage even in the presence of Fe fertilisation and also presents a quota model approach to simulating diatom physiology, tuned to a powerful set of observations, which can be applied to other model frameworks