21st century projections of ocean ecology and productivity across the CMIP5 models: contrasting the Southern Ocean and the Arctic

We use the newest generation of fully-coupled earth system models to study and contrast the response of Southern Ocean and Arctic phytoplankton productivity and biomass to 21^st century climate change. The Arctic is warming at a rate much higher than the Southern high latitudes. Despite fundamental differences between the physical responses to climate in these regions, CMIP5 modes predict small increases in biological production both in the SO and the Arctic, partially compensating for the severe loss of primary production in the rest of the oceans.

South of 40ºS, the CMIP5 models predict a complex, zonally-banded pattern of phytoplankton abundance and productivity changes driven by shifts in light and iron availability with future warming, in agreement with patterns and mechanisms emerging from a satellite trend analysis and other recent observational work. Increased SAM plays a major role in these projections, as increased Southern Ocean westerlies act both to increase mixing and phytoplankton light limitation in a band around 50S, and to modify iron supply to the surface in regions where phytoplankton are iron limited. By contrast, phytoplankton are light and nitrate co-limited in the strongly stratified Arctic ocean. Here we find that over 100 years, release of light limitation due to sea ice retreat is counter-balanced by an increase in nitrogen limitation due to increased stratification. While most models predict net increases in Arctic production by the end of the 21^st century, the different strengths of nutrient-light co-limitation among models result in different magnitude changes in production across models.

We assess the multi-model 100-year trend significance using a novel technique based on bootstrap combined with a weighting scheme based in similarity across models.

We find that model uncertainty in ecological and biogeochemical parameters is higher than for the physical parameters. Additionally, the spread in model predictions is smaller than the spread in historical ecological and biogeochemical parameters.

More energy should be focused on improving simulations of ocean-atmosphere dynamics, biogeochemistry and ecology within the high latitudes, as the uncertainties and spread among models are larger in these regions compared to lower latitudes.