Variability associated with a clay layer compacted in place was investigated using classical and geostatistical methods. Results showed that the geostatistical approach produces more useful information about the system studied. High variability among the flow parameters can more readily be explained using structural analysis and kriging. Observed decreases in intrinsic variability, as flows are averaged over larger volumes, suggest that reasonable approximations of Darcy velocity are possible in field situations. It is concluded that variability of flow data need to be considered when modeling or developing criteria for engineering structures.