Sea ice parameters improvements

Sea ice concentration

The time series of SIC shown in Figure 3a reveal a significant variation during the year with more disagreements between solutions in wintertime and fairly consistent results during the time of the summer minimum. In particular, all model simulations match the observed summer extreme minimum SIC in the years 2007, 2012, and 2016. During the winter season, INTAROS-opt follows the satellite observations well, while the TOPAZ4 reanalysis underestimates SIC during the first four years and follows INTAROS-opt after that. The PIOMAS reanalysis also matches the observations well, but it slightly overestimates SIC in the winter season. In contrast, INTAROS-ctrl simulates more SIC than is observed.
To reveal the winter inter-model discrepancies in more detail, Figure 3b shows total SIC from January to March in the year 2007. The TOPAZ4 reanalysis clearly shows a seven-day signal caused by the artificial energy and mass input at the analysis step characteristic to every filter approach. The POIMAS reanalysis shows more SIC than all the other simulations. Discontinuities are not visible in the PIOMAS reanalysis because they used a mixed nudging and optimal interpolation method to alleviate discontinuities. Yet, the nudging term still adds artificial sources and sinks to the model simulation and violates model physics. The adjoint method brings the model-simulated sea ice cover closest to the observations by adjusting the atmospheric forcing and the initial state in the year 2007.