Exploring the links between model biases in Southern Ocean sea ice and deep ocean circulation in glacial simulations

The Southern Ocean represents a carbon sink and also one of the few regions where deep water formation is triggered, because of cold temperatures and brine release consequent to sea-ice formation. Several paleoclimate studies have underlined its crucial influence to explain a shoaled AMOC (Shin et al., 2003, Klockmann et al., 2016, Marzocchi and Jansen, 2017) and a lower CO₂ concentration (Ferrari et al., 2014, Stein et al., 2020) at the Last Glacial Maximum, when the Southern Ocean sea ice was more extensive and seasonal (Gersonde et al., 2005, Roche et al., 2012). Considering that a majority of PMIP models simulate a too deep and intense AMOC at the LGM (Muglia and Schmittner, 2015, Sherriff-Tadano and Klockmann, 2021), in contrast with paleotracer reconstructions, Marzocchi and Jansen (2017) suggest largely attributing this discrepancy and the large intermodel spread to "differing (and likely insufficient) Antarctic sea-ice formation". In Lhardy et al. (2021), we show that the iLOVECLIM model of intermediate complexity produces at the LGM a very deep and intense NADW overturning cell, in addition to model-data disagreements in the Southern Ocean sea ice (such as an underestimated seasonal range).

We propose investigating the link between these biases thanks to sensitivity tests with a modified wind stress and a parameterisation of the sinking of brines (Bouttes et al., 2010) in the iLOVECLIM model. Wind stress, convection in the Southern Ocean, and sea-ice seasonality seem broadly related in the simulations, with reduced wind stress leading to less convection and an enhanced sea-ice seasonality in the Southern Ocean, in better agreement with proxy data. However, experiments with a modified wind stress do not lead to a water mass distribution in good match with δ¹³C data (Peterson et al., 2014), contrary to simulations with a parameterised sinking of brines. These results thus do not support the systematic attribution of the deep and intense AMOC to an insufficient sea-ice cover in the Southern Ocean, but rather underline the importance of model representation of convection processes.