Global climate models results show a substantial spread among different models at high latitudes. Possibly, the spread is caused by the high climate variability, alternatively it can also be caused by the inability of today's climate model to accurately reproduce the climate processes in the Arctic. The feedback between ice albedo and radiation is an example of a process that is highly sensitive, and small errors in the parameterizations can have large consequences for the resulting climate.

We have studied the interaction between radiation and clouds with a state of the art regional climate model that has been set up for the Arctic. The model results are compared against observations from the ARM SHEBA site and against satellite data. The simulated SW and LW radiation agree fairly well with observations, but the factors that strongly influence the radiative transfer don't. In particular the annual cycle of the cloud cover is not reproduced well.

Possibly, the fair result for radiation is due to compensating errors, and improved parameterizations may lead to worse result despite a better representation of the physics.

We investigate the impact from a separation of cloud phase, and parameterizations of the effective radius and surface albedo. However, changing the parameterizations within the radiation scheme is not sufficient to reproduce the observed variability. A statistical cloud scheme is tested to see if the exisiting cloud cover parameterization can be improved. Finally, updated parameterizations of key processes are then used in a coupled version of the regional climate model to study the feedback between sea-ice and atmosphere in the Arctic.