Large-ensemble analysis of Antarctic sea ice model sensitivity to parameter uncertainty

Abstract

Simulated Antarctic sea ice sensitivity to parameter variation is tested in a global coupled climate model. Parameter changes affect areal coverage, thickness and seasonality of sea ice beyond the bounds of internal variability, with a tendency for several parameters to substantially impact sea ice retreat. The largest sea ice response occurs with changes to the sea ice turning angle parameter: enhanced northward transport of ice leads to a thinner ice cover that more readily melts, while lower turning angles permit greater convergence in the interior, resulting in a thicker, more persistent ice pack. Setting snow and ice albedo above the default value increased mean sea ice year-round, including at the sea ice minimum, while other parameter experiments did not increase minimum sea ice beyond internal variability. Low values of heat exchange between the ocean and base of the sea ice (ocean–ice heat flux parameter) sharply reduce bottom melt and slow sea ice retreat. High-density sampling in both the control scenario (identical parameters but differing initial conditions) and the parameter tests shows large simulated internal variability, suggesting that the signal of the sea ice response could be overwhelmed in a low-density sample test. Large-ensemble testing is a valuable tool for testing sensitivity in coupled systems where variability and feedbacks are more complex than in stand-alone sea ice models.