This study describes an inverse approach for efficiently identifying the spatial shapes of zones of low (or high) permeability using the level set method, given a set of spatially distributed head measurements. By this method, the boundaries of zones are characterized by a level set function. From an initial setting, the unknown regions of zones are determined by evolving the boundaries in artificial time using a pseudo velocity field that is related to the sensitivity of head to permeability and the residual between the measured head and modeled head at the current time. A synthetic example presented to illustrate the method.