Abstract. Observations by specialized sonars moored beneath drifting pack ice in the Beaufort Sea have been used to guide interpretation of SAR imagery under Arctic winter conditions. The sonars measured the draft, bottom-side topography and kinematic history of the ice field during the same time that SAR images were acquired by ERS-1 (satellite-based C -band) and STAR-2 (aircraft-based X -band) radar systems. Inter-comparison of the observations revealed that ridges crossing first-year pack ice cannot be detected reliably at the low incidence angles used by ERS-1. However, at the high incidence angles typical of airborne SAR, there is sufficient contrast to enable unambiguous discrimination of ridges and intervening floes. A strong correlation was observed between the spatial frequency of ridges and the average draft of the ice field. Therefore, a count of ridges intersecting a line across a SAR image can be used as a proxy for ice draft. A linear relation was observed between radar brightness and draft of ridges. This is probably a consequence of a progressive filling of the radar pixel by the ridge sail as the draft of the ridge increases. The relation between the frequency and maximum backscatter of ridges viewed by radiometrically calibrated SAR has potential as a surrogate for the keel frequency-draft function. Estimates of ridge frequency derived via SAR can, however, be biased, since both ridges and young leads appear as bright filaments in imagery. With ERS-1, there is a large variation in the microwave backscattering coefficient of sea ice during the early growth under cold conditions. This variation provides a clear indication of ice age and stage of development. For older pack ice, ERS-1 can provide only an aggregate measure of the backscatter from unresolved ridges and level floes in a selected area. The strength of this backscatter is a poor indicator of age or thickness because the relative contributions by ridges, frost flowers and smooth floes cannot be determined.