Understanding the Intermodel Spread of Simulated Arctic September Sea-Ice Sensitivity

We investigate the reasons for the intermodel spread of simulated Arctic September sea-ice sensitivity. Previous studies have found that Arctic September sea-ice area declines linearly with cumulative CO2 emissions both in observations and climate-model simulations. However, the models’ sensitivity differs substantially, with the models generally underestimating the sensitivity of sea-ice area to CO2 emissions. We here examine the reasons for the large intermodel spread in order to be also able to understand the general underestimation.We identify a chain of processes contributing to the overall sea-ice sensitivity and investigate the simulation of each sub-process separately in each CMIP6 model. The process chain considers the global-mean temperature response to CO2 increase, Arctic amplification, the increase in incoming longwave radiation, the total non-shortwave heat flux in the Arctic, and the resulting sea-ice loss. In addition, we separately examine the impact of the simulated incoming longwave radiation for the spread of sea-ice sensitivity. Doing so, we find that clouds play a minor role for the spread of simulated incoming longwave radiation but that temperature rise and water vapour content in the Arctic are relevant.Based on these analyses, we identify three processes whose different representation in climate models likely is the main cause for the intermodel spread of simulated sea-ice sensitivity, and which need to be improved to improve the modeled sensitivity of Arctic sea ice: firstly the global-mean temperature response to CO2 increase, secondly the Arctic amplification and thirdly local sea-ice processes. The first two factors highly impact the evolution of temperature in the Arctic which affects the incoming longwave radiation and thus the evolution of sea ice.