Evaluation and Attribution of a Warm Winter Bias Over Arctic Sea Ice in a Climate Model

Biases of the winter near-surface temperature over Arctic sea ice have been reported in climate models, increasing uncertainties in future sea ice and Arctic climate projections. Mitigating these biases in future model versions requires proper evaluation and understanding of their origin. To progress on these matters, we focus on the near surface air temperature simulated in the atmosphere-only and coupled configurations of the IPSL-CM6A-LR climate model. To establish a reliable baseline for evaluating simulations, we identified a linear relationship between the mean surface air temperature from the European Centre for Medium-Range Weather Forecasts 5th generation reanalysis (ERA) and their bias relative to in situ observations from Soviet North Pole drifting stations. This relationship is then used to correct the ERA5 data. We find the winter near-surface temperature bias in the atmosphere-only IPSL-CM6A-LR configuration turns from cold to warm once ERA5 is linearly corrected, reaching +2.2°C over Arctic multiyear ice. The bias increases to +4.8°C in the fully-coupled configuration. Using a pan-Arctic energy budget evaluation, the warm bias in IPSL models is explained by an excessive poleward atmospheric heat transport. In the coupled configuration, the warm bias is increased by the too small sea ice extent, which also acts to reduce the overestimated atmospheric heat transport and leads to a too small poleward oceanic heat transport and surface energy budget. The methods developed here could be used in multi-model evaluations to further progress in understanding and reducing biases in climate models.