Monte-Carlo analysis of uncertainty on reservoir performance is very expensive. It involves: first, the generation of multiple permeability fields at the scale of the measurements; second, the average of these permeabilities to obtain equivalent grid-block permeabilities at a scale that can be handled by current numerical reservoir simulators, and third, the simulation of flow on these scaled-up fields. Although the most expensive step is, generally, the third one, there is still time and effort invested in the first two steps. An approach is proposed for the fast generation of equivalent grid-block permeabilities directly from the measurements honoring both local information and structural information. The approach is based on a geostatistical model for the joint spatial variability of both measured permeabilities and grid-block permeabilities.