A Link Between CMIP5 Phytoplankton Phenology and Sea Ice in the Atlantic Southern Ocean

Abstract

Earth system models from the Climate Model Intercomparison Project phase 5 (CMIP5) are strongly biased in Southern Ocean phytoplankton phenology, especially in the marginal ice zone. In this study we describe the mechanisms driving CMIP5 models to misrepresent seasonal primary production in the Atlantic marginal ice zone during late winter. We link subsurface light availability during this period to simulated early growth, arguing that a combination of ice cover and deep winter mixing prevent biomass accumulation in the real ocean, while in models this combination of factors is not present. Furthermore, we find a statistically significant correlation across the model ensemble between vertical stratification and the location of the ice edge; whereby the more equatorward the ice edge is, the closer to the surface stratification occurs. We argue that models may be grouped according to how strongly they express two major controls on their phenology, namely, the location of the ice edge and the degree of stratification present in the water column in late winter. We find that models with small biases in just one of these controls (but large biases in the other) are able to simulate bloom initiation close to observations, while models with significant biases in both controls initiate growth 2–4 months early.