We compare the formulation and emergent dynamics of 11 CMIP6 IPCC marine biogeochemical models. We find that the largest source of uncertainty across model simulations of marine carbon cycling is grazing pressure (i.e. the phytoplankton specific loss rate to grazing). Variability in grazing pressure is driven by large differences in zooplankton specific grazing rates, which are not sufficiently compensated for by offsetting differences in zooplankton specific mortality rates. Models instead must tune the turnover rate of the phytoplankton population to balance large differences in top-down grazing pressure and constrain net primary production. We then run a controlled sensitivity experiment in a global, coupled ocean-biogeochemistry model to test the sensitivity of marine carbon cycling to this uncertainty and find that even when tuned to identical net primary production, export and secondary production remain extremely sensitive to grazing, likely biasing predictions of future climate states and food security.

Data was collected and processed for the project: "Assessment and communication of risks to Tasmanian aquaculture and fisheries from marine heatwaves". Observational data is from NOAA OISST v2.1 (1982-2020), and model data is from 25 CMIP6 models over the historical period (from 1982-2014), with SSP1-2.6 and SSP5-8.5 extensions (out to 2100). Raw sea surface temperature data truncated to the Tasmanian region: 138-155E, 49-35S. Time-series of subdomain area averages are also provided, along with calendar corrections, mean-bias corrections, and seasonal bias corrections for the model data. Further details are provided in Kajtar, J.B. and Holbrook, N.J. (2021): "Future projections of marine heatwave hazards to aquaculture and fisheries in Tasmania", Institute for Marine and Antarctic Studies, University of Tasmania, Australia. 36pp. ISBN: 978-1-922708-06-9. http://ecite.utas.edu.au/147866.