Abstract

General circulation models have been amply used to quantify Arctic sea-ice predictability. While models share some common aspects of predictability loss with increasing forecast lead time, there is significant model spread in the magnitude and timing of predictability loss. Here we show that inte... Show moreGeneral circulation models have been amply used to quantify Arctic sea-ice predictability. While models share some common aspects of predictability loss with increasing forecast lead time, there is significant model spread in the magnitude and timing of predictability loss. Here we show that inter-model differences in predictability are linked to inter-model differences in the persistence timescales of sea-ice anomalies that are unique to each model, with models that exhibit longer persistence having higher potential predictability. Given this result and previous work showing that in a single model control simulation the magnitude of persistence fluctuates between multi-annual periods of high and low persistence, we assess whether initial-value predictability is dependent on the persistence state of the initial conditions. We find that predictability is not clearly impacted by the persistence state of the initial conditions, suggesting that predictability may be robust within a constant climate mean state. Show less