Understanding Arctic Sea Ice Variability and Predictability With Regional-to-Global-Scale Process-Oriented Evaluation

Abstract:

Sea ice is a key source of climate predictability on weekly to decadal timescales. The ability of models to represent important sea-ice processes that give rise to sea-ice predictability is influenced by model bias. Yet a long-standing feature of sea-ice predictions is the very large intermodel spread that is pervasive on many time scales. Part of this spread appears to be irreducible, but a considerable portion is due to model errors in the sea-ice physics and coupled interactions of the sea ice, atmosphere and ocean. We are analyzing CMIP5 and -6 simulations and observations to identify errors that are critical to simulating sea-ice variability with a focus on quantifying model processes that influence persistence, with the assumption that excessive persistence leads to overestimating predictability. We find that anomalies in the Arctic sea-ice cover (local concentration and area/extent by region) in CMIP models are too persistent from year to year and month to month compared to observations. Such excessive persistence means that the anomalies last too long, drive variance higher and create bias. We also find that the standard deviation of monthly sea-ice area is too high in nearly all models. Unsurprisingly, sea-surface temperature anomalies are similarly overly persistent in the Arctic in CMIP models. Ocean boundary layer depths tend to be too deep in models, but we find little association between these depths and sea ice persistence. We will present results of new metrics that associate sea-ice variability and persistence to oceanic and atmospheric heat transport, ocean boundary layer depths, sea-ice albedo feedbacks, and teleconnections from the midlatitudes and tropics.