Correcting Seasonal Feedbacks for Physically Constrained Sea Ice Reanalysis

Anthropogenic climate change is reducing Arctic sea ice during all seasons. The strongest decline takes place during summer and while under current global warming projections most climate models project the Arctic to become ice-free before 2060, the spread among the models is huge. The reduction in sea ice matters because it is having, and is expected to continue to have, a dominant impact on local climate, ecosystems and northern communities that rely on the sea ice as a platform for travel and subsistence hunting as well as protection from waves during storms. Impacts, however, extend well beyond the Arctic, as enhanced warming within the Arctic (Arctic amplification) alters large-scale weather patterns and the loss of sea ice influences ocean stratification and circulation. Satellites have been key tools in our ability to document the reductions in areal sea ice extent over the last 45 years, but until recently, information on how thick that ice was were lacking. With the launch of new satellites, e.g. ICESat-2 (IS2) and CryoSat-2 (CS2) our observational capacity has improved to monitor ice thickness year-round, and, combined with field experiments such as those made during the year-long MOSAiC expedition, we have a better understanding of the limitations of current satellite data products. We have discovered two fundamental differences between model simulations and satellite observations: (1) Sea ice recovery during winter (strength of the negative sea ice growth - sea ice thickness feedback) and (2) Timing of the melt season. In the model simulations, the ice growth rate is around 3 times higher for thinner ice (below 1m) than for thick ice (above 3 m, see Fig. 2 in Vision). This means that after a warm summer with increased ice melt, the thinner ice will grow quicker and recover. In the observations, this dependence on ice thickness is hardly evident. Further, we find the modelled pan-Arctic melt rate in August is more than twice as strong as that in estimates from CS2 (Fig.1 in Vision). As a result, we hypothesise that the recovery of ice during winter is too strong in climate models and this is compensated by increased summer melt. Known model issues include representation of lead fraction and thin ice during winter (areas with maximum ice growth) and the accumulation/distribution of snow: uniform snow reduces net ice growth more than piled-up snow behind pressure ridges. Recent observationally-based thin ice fraction and snow products allow us to improve upon these key areas in the models, and enable a more realistic ice growth and melt simulation for the correct reasons. Year-round MOSAiC observations will be used to verify the feedbacks and to examine uncertainties in the CS2 data products that may help to quantify mismatches in observed vs. simulated feedback processes. With the improved sea ice model and the improved sea ice data we will create a sea ice reanalysis for the CS2 period (2010 to present) which provides a physically constrained estimate not only of the sea ice state but also of the processes contributing to the sea ice volume budget: sea ice melt and freeze, transport, and deformation. This project will provide a data set to validate the Arctic sea ice mass budget from climate simulations and will improve feedback processes in climate models, a necessary step to improve upon future projections.