Climate change simulations indicate a polar amplification in temperature increase, presumably at least in part due to the well-known ice/snow-albedo feedback. However, proper modeling of this feedback requires an adequate modeling of the surface exchange processes and thus of the boundary layer. That this is not the case today is borne out in part by the very large inter-model scatter in climate scenarios for the Arctic, much larger than anywhere else on Earth.

In this paper, six state-of-the-art regional-scale models are compared to data from the Surface Heat Budget of the Arctic Ocean – SHEBA – experiment. All models where set up on the same moderate-sized domain over the western Arctic and run through the SHEBA-year, using six-hourly ECMWF analyses at the lateral boundaries. To further neutralize the different models cryospheres from the atmospheric models, sea and ice surface temperatures, as well as ice fraction, was prescribed from satellite data.

Although the actual surface temperatures were prescribed, the modeled 2-meter temperature and humidity are somewhat different in between the models. Many models have a problem describing the very coldest temperatures, indicating that they have problems with the long-lasting very stable surface inversions during the Arctic night. Conversely, some models have a slight problem with humidity and temperature also during summer, when the ocean water is near the freezing point of salt water, while the ice-surface temperature is near the melting point of fresh water - the melting snow on the ice. Different models also have different but tolerable wind-speed biases. Temperature and humidity bias profiles, from comparing to SHEBA soundings, show distinct differences between the models and also between the boundary layer and the free troposphere, usually with larger biases and more variability in the boundary layer.

There are also systematic biases in the momentum flux, consistent with the wind speed bias in each model. The real problem comes when comparing the components of the surface energy balance. While the radiative fluxes are similar between the models, but sometimes with large biases, the turbulent fluxes are problematic. For both sensible and latent heat flux there is little similarity between the models and even less between any of the models and the SHEBA observations. Cumulative errors in sensible and latent heat cancel to varying degree in some models, but the remaining error in the net heat flux is easily as large as typically observed net fluxes, or larger.