A mechanistic particle flux model applied to the oceanic phosphorus cycle

Abstract. The sinking and decomposition of particulate organic matter are critical processes in the ocean's biological pump, but are poorly understood and crudely represented in biogeochemical models. Here we present a mechanistic particle remineralization and sinking model (PRiSM) that solves the evolution of the particle size distribution with depth. The model can represent a wide range of particle flux profiles, depending on the surface particle size distribution, the relationships between particle size, mass and sinking velocity, and the rate of particle mass loss during decomposition. The particle flux model is embedded in a data-constrained ocean circulation and biogeochemical model with a simple P cycle. Surface particle size distributions are derived from satellite remote sensing, and the remaining uncertain parameters governing particle dynamics are tuned to achieve an optimal fit to the global distribution of phosphate. The resolution of spatially variable particle sizes has a significant effect on modeled organic matter production rates, increasing production in oligotrophic regions and decreasing production in eutrophic regions compared to a model that assumes spatially uniform particle sizes and sinking speeds. The mechanistic particle model can reproduce global nutrient distributions better than, and sediment trap fluxes as well as, other commonly used empirical formulas. However, these two independent data constraints cannot be simultaneously matched in a closed P budget commonly assumed in ocean models. Through a systematic addition of model processes, we show that the apparent discrepancy between particle flux and nutrient data can be resolved through P burial, but only if that burial is associated with a slowly decaying component of organic matter such as might be achieved through protection by ballast minerals. Moreover, the model solution that best matches both data sets requires a larger rate of P burial (and compensating inputs) than have been previously estimated. Our results imply a marine P inventory with a residence time of a few thousand years, similar to that of the dynamic N cycle.