Improving the representation of the sea ice and snow heat conduction in models through the lens of the MOSAiC dataset

The parameterization of the heat conduction through sea ice and snow remains simple in state-of-the-art models. Specifically, it relies on prescribed conductivity parameters constant in time and space, therefore neglecting the substantial heterogeneity of these mediums down to the unresolved subgrid scale. This assumption clashes with robust observational evidence, which indicates that snow and ice conductivities can vary greatly depending on the environmental conditions and the history of the sea ice. The winter observations collected during the MOSAiC expedition are unique tools for advancing the quantitative understanding of heat conduction in sea ice and improving the realism of the thermodynamic parameterizations in models. Our investigation utilizes gridded helicopter-borne thermal infrared imaging, laser scanner (ALS) elevation observations, and meteorological measurements to assess the model bias and diagnose the importance of unresolved processes and topographic heterogeneity on heat conduction. We evidence different heat conduction regimes depending on the ice thickness, type (i.e., ridged or level ice), and snow patchiness. In light of these results, we will discuss strategies for an effective parametrization of these unresolved processes in sea ice models, and their harmonization with the preexisting model infrastructure. Furthermore, I will comment on the potential of emerging data-driven analysis techniques and machine learning in facilitating the formulation of parameterization at different stages of the development process.