Understanding model spread in sea ice volume by attribution of model differences in seasonal ice growth and melt

Abstract. Arctic sea ice is declining rapidly, but predictions of its future loss are made difficult by the large spread both in present-day and area volume; hence, there a need to better understand drivers model state. Here we present framework for understanding differences between modelled simulations based on attributing seasonal growth melt differences. In method presented, net downward surface flux treated as principal driver melt. An energy balance approach used estimate pointwise effect key climate variables this hence We compare three models with very different historical simulations: HadGEM2-ES, HadGEM3-GC3.1 UKESM1.0. The largest these shown be summer (representing albedo feedback) thickness distribution winter (the thickness–growth feedback). Differences snow pond cover during early exert smaller melt, representing volume. particular, direct impacts differing parameterisations ponds small non-negligible.