Despite differences in the sea ice thickness, mean winter temperature and proximity to the ice edge, the N-ICE2015 and SHEBA datasets both have a bimodal distribution of the net longwave radiative flux (NetLW) for January-February, with modal values of -40 W m-2 and 0 W m-2. These correspond to the radiatively clear and opaquely cloudy atmospheric winter states, respectively. The recent N-ICE2015 observations indicate that these modal values of NetLW are representative of the wider Arctic. The ability to simulate such a NetLW distribution is therefore suggested as a benchmark for climate models in the Arctic.
We further compare N-ICE2015 and SHEBA data with the ERA-Interim reanalysis (ERA-I) and output from the coupled Arctic regional climate model HIRHAM-NAOSIM. ERA-I simulates two Arctic winter states and the captures the timing of transitions from the radiatively clear to opaquely cloudy state well, despite underestimating the cloud liquid water path. HIRHAM-NAOSIM permits cloud liquid water at lower temperatures compared with ERA-I, and the winter mean NetLW flux is in closer agreement with the observations. However, HIRHAM-NAOSIM has a Gaussian rather than bimodal NetLW distribution, and thus does not simulate the two states.
We identify a strong correlation between the number of opaquely cloudy days and mean winter temperature at the N-ICE2015 site in ERA-I, coupled with a positive trend in the number of opaquely cloudy days and surface temperature from 1979-2015. Our analyses indicate that an increase in the occurrence of the opaquely cloudy state is an important contributor to recent Arctic warming. This means that climate models must simulate the two winter states accurately to give robust estimates of future and recent Arctic warming.