A regional atmospheric climate model has been applied to simulate the Arctic climate north of 65 ¿N at a 50 km horizontal resolution for January and July 1990. The monthly mean components of the surface radiative balance and clouds using ECHAM3 parameterization have been described and compared with data. The model overestimates the incoming shortwave flux, overestimates clouds during January, and underestimates clouds during July. The high-resolution simulations show many regional details due to regionally different surface temperatures, clouds, surface albedo, and snow depths. The bias in the model simulation can be reduced by using the ECHAM4 parameterization. The use of the ECHAM4 radiation parameterization clearly improves the results, due to the better description of atmospheric absorption. Sensitivity experiments with different cloud parameters (drop size, droplet concentration, cloud phase) have been described. The simulated changes due to these cloud parameter changes occur in a very small spatial scale and are of similar order as the model bias itself. ¿ 1997 American Geophysical Union