Investigating the Biases in the Antarctic Sea Ice - Ocean System of Climate Models using Process-oriented Diagnostics

Abstract:

Most analyses of Antarctic sea ice in simulations of the CMIP5 archive have so far been oriented towards the quantification of the disagreement between model results and sea ice observations only. Since the decomposition of those biases into distinct physical components is necessary to understand their origins, we propose here an ocean-sea ice-atmosphere integrated and process-oriented approach. Not only the biases in variables essential to the sea ice seasonal evolution are estimated regionally with regard to observations, but their contributions to the sea ice concentration budget are estimated.

Following a previously developed method, the sea ice concentration balance over the autumn-winter seasons is decomposed into four terms, including the sea ice concentration change during the period of interest, advection, divergence and a residual accounting for the net contribution of thermodynamics and ice deformation. Concurrently, correlations between trends in ocean temperature at depth and trends in ice concentration are calculated directly from various model output fields (including CMIP5 models) to disentangle the role of ice-ocean interactions.

Results show that the geographical patterns of all mean sea ice concentration budget terms over 1992-2005 are in qualitative agreement with the observed ones. Sea ice thermodynamic growth is maintained by horizontal divergence near the continent and in the central ice pack, whereas melting close to the ice edge is led by sea ice advection. However, significant errors in all budget terms are observed due to ice velocities that tend to be overestimated all around Antarctica in several models, leading to a relatively weak divergence in the inner ice pack and to an excessive advection in the marginal ice zone. Biases in ice drift speed and direction are ultimately related to biases in winds in all models.

This method paves the way for a systematic assessment of forthcoming CMIP6 sea ice model outputs in the Southern Hemisphere.