Surface hoar deposited on the snow surface represents, once buried by subsequent snowfall, one of the principal weak layers on which dry snow slab avalanches release. To predict instabilities caused by a buried surface hoar layer, its spatial extent needs to be known. Avalanche forecasting relies, among other things, on meteorological data from automatic stations. In principle, surface hoar formation can be predicted from these data. In order to study the spatial variation in surface hoar formation and destruction, daily observations were made during one winter at 23 locations of different aspect, slope inclination, and wind exposure within an area of about 3 km². Four automatic weather stations were located within the study area: one on level terrain and three across a ridge. Despite the good instrumentation the correlation between surface hoar growth and calculated sublimation rate was poor. Distinct spatial patterns of surface hoar growth were found. Surface hoar crystals were frequently larger at the ridge site than in the surroundings of the automatic weather station on level terrain. The variation in surface hoar formation was mainly due to different prevailing wind regimes during the formation periods. The surroundings of the automatic weather station on level terrain were under the influence of local katabatic winds that dried up the air so that growth conditions were locally less favorable. Our observations suggest that predicting surface hoar formation for complex alpine terrain on the basis of data from an automatic weather station, the standard procedure in avalanche forecasting, seems nearly impossible unless at least the local wind regime is known at high resolution (≤10 m). For both surface hoar formation and surface hoar destruction observations suggest wind conditions to be most crucial for spatial variation.