Abstract. This study focuses on an improved representation of the biological soft tissue pump in the global threedimensional biogeochemical ocean model PISCES. We compare three parameterizations of particle dynamics: (1) the model standard version including two particle size classes, aggregation-disaggregation and prescribed sinking speed; (2) an aggregation-disaggregation model with a particle size spectrum and prognostic sinking speed; (3) a mineral ballast parameterization with no size classes, but prognostic sinking speed. In addition, the model includes a description of surface sediments and organic carbon early diagenesis. Model output is compared to data or data based estimates of ocean productivity, pe-ratios, particle fluxes, surface sediment bulk composition and benthic O 2 fluxes. Model results suggest that different processes control POC fluxes at different depths. In the wind mixed layer turbulent particle coagulation appears as key process in controlling pe-ratios. Parameterization (2) yields simulated pe-ratios that compare well to observations. Below the wind mixed layer, POC fluxes are most sensitive to the intensity of zooplankton flux feeding, indicating the importance of zooplankton community composition. All model parameters being kept constant, the capability of the model to reproduce yearly mean POC fluxes below 2000 m and benthic oxygen demand does at first order not dependent on the resolution of the particle size spectrum. Aggregate formation appears essential to initiate an intense biological pump. At great depth the reported close to constant particle fluxes are most likely the result of the combined effect of aggregate formation and mineral ballasting.

Abstract. This study focuses on an improved representation of the biological soft tissue pump in the global three-dimensional biogeochemical ocean model PISCES. We compare three parameterizations of particle dynamics: (1) the model standard version including two particle size classes, aggregation-disaggregation and prescribed sinking speed; (2) an aggregation-disaggregation model with a particle size spectrum and prognostic sinking speed; (3) a mineral ballast parameterization with no size classes, but prognostic sinking speed. In addition, the model includes a description of surface sediments and organic carbon early diagenesis. The integrated representation of material fluxes from the productive surface ocean down to the sediment-water interface allows taking advantage of surface ocean observations, sediment trap data and exchange fluxes at the sediment-water interface. The capability of the model to reproduce yearly averaged particulate organic carbon fluxes and benthic oxygen demand does at first order not dependent on the resolution of the particle size spectrum. Model results obtained with the standard version and with the one including a particle size spectrum and prognostic sinking speed are not significantly different. Both model versions overestimate particulate organic carbon between 1000 and 2000 m, while deep fluxes are of the correct order of magnitude. Predicted benthic oxygen fluxes correspond with respect to their large scale distribution and magnitude to data based estimates. Modeled particulate organic C fluxes across the mesopelagos are most sensitive to the intensity of zooplankton flux feeding. An increase of the intensity of flux feeding in the standard version results in lower mid- and deep-water particulate organic carbon fluxes, shifting model results to an underestimation of particulate organic carbon fluxes in the deep. The corresponding benthic oxygen fluxes are too low. The model version including the mineral ballast parameterization yields an improved fit between modeled and observed particulate organic carbon fluxes below 2000 m and down to the sediment-water interface. Our results suggest that aggregate formation alone might not be sufficient to drive an intense biological pump. The later is most likely driven by the combined effect of aggregate formation and mineral ballasting.