Arctic sea ice thickness and surface morphology obtained by means of helicopter-borne electromagnetic induction sounding and laser altimetry have been investigated in order to improve radar ice type classification. Simultaneously acquired Synthetic Aperture Radar (SAR) images are available for many of the flight tracks. Since ice thickness measurements are considerably more difficult to accomplish than surface measurements, it is important to improve techniques for estimating thickness from surface characteristics by means of remote sensing. Radar signatures are dependent on ice surface topography and ice volume properties, but ice thickness cannot be measured directly by means of radar. The surface and thickness profiles were analysed in order to improve understanding of the relation between surface roughness and ice thickness. The stochastic properties of the surface profiles have been analysed and parameters have been extracted to characterize the roughness. Based on the available thickness information, profiles have been grouped into thickness classes, and the roughness parameters for the different groups have been analysed. In addition, normalized backscatter coefficients obtained from SAR images have been classified into groups and compared to the roughness parameters. Independently, a clustering algorithm has been applied to the roughness parameters, and the resulting roughness classes have been compared to the ice thickness classes previously obtained.