Toward Unanimous Projections for Sea Ice Using CMIP5 Multi-model Simulations

Abstract:

Coupled global climate models have been used to provide future climate projections as major objective tools based on physical laws that govern the dynamics and thermodynamics of the climate system. However, while climate models in general predict declines in Arctic sea ice cover (i.e., ice extent and volume) from late 20^th century through the next decades in response to increase of anthropogenic forcing, the model simulated Arctic sea ice demonstrates considerable biases in both the mean and the declining trend in comparison with the observations over the satellite era (1979-present). The models also show wide inter-model spread in hindcast and projected sea ice decline, raising the question of uncertainty in model predicted polar climate.

In order to address the model uncertainty in the Arctic sea ice projection, we analyze the Arctic sea ice extent under the context of surface air temperature (SAT) as simulated in the historical, RCP4.5 and RCP8.5 experiments by 27 CMIP5 models. These 27 models are all we could obtain from the CMIP5 archive with sufficient gird information for processing the sea ice data. Unlike many previous studies in which only limited number of models were selected based on metrics of modeled sea ice characteristics for getting projected ice with reduced uncertainty, our analysis is applied to all model simulations with no discrimination. It is found that the changes in total Arctic sea ice in various seasons from one model are closely related to the changes in global mean SAT in the corresponding model. This relationship appears very similar in all models and agrees well with that in the observational data. In particular, the ratio of the total Arctic sea ice changes in March, September and annual mean with respect to the baseline climatology (1979-2008) are seen to linearly correlate to the global mean annual SAT anomaly, suggesting unanimous projection of the sea ice extent may be possible with this relationship. Further analysis is underway to understand how the patterns of the projected sea ice changes are related to the model biases, and whether there are common characteristics in these patterns among models.