Search for a command to run...
The Multidisciplinary drifting Observatory for the Study of Arctic Climate (MOSAiC) expedition was implemented with one year observation of atmosphere, ocean and sea ice, giving us an opportunity to understand the sea ice processes. Due to the missing observation during the expedition, ERA5 atmospheric reanalysis along the MOSAiC drift trajectory, after its validation, is used to force a column sea ice model Icepack, commonly used in coupled climate models. We compare sea ice thickness (SIT) simulations against MOSAiC observation to understand the reasons of SIT simulation misfits fordifferent combinations of two melt pond schemes and three snow redistribution configurations. The three snow redistribution configurations are bulk scheme, snwITDrag scheme and one simulation selection without snow redistribution. In both bulk and snwITDrdg snow redistribution schemes, snow can be lost to leads and open water. In the bulk scheme, snow from level ice can be lost to leads or open water. In snwITDrdg scheme, snow is distributed to different sea ice categories and the scheme also allows wind-driven snow compaction and erosion. The two melt pond schemes are TOPO scheme and LVL scheme, which differ in the distribution of melt water. The results show that Icepack can reproduce sea ice growth in the winter and spring periods of MOSAiC expedition. Icepack without snow redistribution scheme simulates excessive snow ice formation and its contribution to sea ice mass balance, resulting in thicker SIT simulation than the observation in spring. Applying snow redistribution schemes in Icepack reduces snow-ice formation while enhancing congelation rate. The bulk snow redistribution scheme improves the SIT simulation in winter and spring, while the bias is larger in simulations using the snwITDrdg scheme. During summer time, Icepack underestimates the sea ice surface albedo, resulting in an underestimation of SIT at the end of simulation. The simulations with TOPO scheme present a more reasonable melt pond evolution than the LVL scheme, resulting in a smaller bias in SIT simulation. Sensitivity analysis and parameter estimation are required to improve sea ice thickness simulation. Some earlier results using adjoint model to improve sea ice simulation will also be presented.