Evaluation of the Arctic Sea-Ice Simulation on SODA3 Datasets

This study evaluates the Arctic sea-ice simulation of the SODA3 dataset driven by different atmospheric forcing fields and explores the errors of the Arctic sea-ice simulation caused by the forcing field. We find that the SODA3 data driven by different forcing fields represent a significant systematical error in the simulation of Arctic sea-ice concentration, showing a low concentration of thick ice and a high concentration of thin ice. In terms of sea-ice extent, the SODA3 data from different versions well characterize the interannual variability and declining trend in the observed data, but they overestimate the overall Arctic sea-ice extent, which is related to excessive simulation of ice in the sea-ice margin. Compared to observations, all the chosen SODA3 reanalysis versions driven by different atmospheric forcing generally tend to underestimate the Arctic sea-ice thickness, especially for thick ice in the multi-year sea-ice regions. Inaccurate simulations of Arctic sea-ice transport may partly explain the error in SODA3 sea-ice thickness in multi-year sea-ice areas. The results of different SDOA3 versions differ greatly in the Beaufort Sea, the Fram Strait, and the Central Arctic Sea. The difference in sea-ice thickness among different SODA3 versions is primarily due to the thermodynamic contribution, which may come from the diversity of atmospheric forcing fields. Our work provides a reference for using SODA3 data to study Arctic sea ice.