This study isolates the influence of sea ice mean state on; pre-industrial climate and transient 1850-2100 climate change within a; fully coupled global model: The Community Earth System Model version 2; (CESM2). The CESM2 sea ice model is modified to increase surface albedo,; reduce surface sea ice melt, and increase Arctic sea ice thickness and; late summer cover. Importantly, increased Arctic sea ice in the modified; model reduces a well-known present-day late-summer ice cover bias. Of; interest to coupled model development, this bias reduction is realized; without degrading the global model simulation including; top-of-atmosphere energy imbalance, surface temperature, surface; precipitation, and major modes of climate variability. The influence of; sea ice mean state on transient 1850-2100 climate change is compared; within a large initial condition ensemble framework. Despite similar; global warming, the modified model with thicker Arctic sea ice than; CESM2 has a delayed and more realistic transition to a seasonally ice; free Arctic Ocean. Differences in transient climate change between the; modified model and CESM2 are challenging to detect due to large; internally generated climate variability. In particular, two common sea; ice benchmarks - sea ice sensitivity and sea ice trends - are of limited; value for contrasting model performance. More broadly, these results; show the importance of a reasonable initial Arctic sea ice state when; simulating the transition to an ice-free Arctic Ocean in a warming; world. Additionally, this work highlights the need for and value of; large initial condition ensembles for credible model-to-model and; observation-model comparisons.
