During a recent study of the katabatic boundary layer at Dumont d'Urville, a triaxial sodar ran continuously for two consecutive years, 1993 and 1994. This study aims to evaluate the capacity of sodar to measure the local vertical velocity, by comparison with the analyses and re-analysis of the European Center, and a basic evaluation using local soundings. It is found that the results of the operational analyses are distorted by an important bias. At the surface and coastal area the large scale vertical velocity is of the order of 1 cms-1, and can reach values as large as 5 cms-1. In the re-analysis the difference between vertical velocity at the surface and 650 hPa is weak. At the surface inside the continent, in the area of Dome C, vertical velocity is weak, lower than 1 cms-1. The vertical velocity on a synoptic scale shows a seasonal cycle. ¿ 2001 American Geophysical Union