Arctic sea ice in Earth system models: decadal hindcast skill and constraints of long-term projections

Abstract

The recent retreat of summer Arctic sea ice has critical implications for climate worldwide. This thesis investigates the potential of observations to improve Earth system model simulations of Arctic climate. It is shown that model initialization with observations can improve decadal hindcasts in the North Atlantic, Greenland and Beaufort Sea regions. The thesis also identifies ways to improve decadal climate predictions. Furthermore, a relatively new observation-based analysis method is used to constrain uncertainties in multi-model projections of 21st-century Arctic sea ice extent, by which model uncertainty could be reduced by up to 50%. However, irreducible internal variability is too large for exact predictions of, for example, the timing of a first disappearance of summer Arctic sea ice. The thesis concludes that mitigation strategies to reduce Arctic warming need to be intensified and that it becomes increasingly crucial to further improve the understanding of the Arctic climate system.