Hydraulic tomography is an emerging field and modeling method that provides a continuous hydraulic conductivity (K) distribution for an investigated region. Characterization approaches that rely on interpolation between one-dimensional (1D) profiles have limited ability to accurately identify high-K channels, juxtapositions of lenses with high K contrast, and breaches in layers or channels between such profiles. However, locating these features is especially important for groundwater flow and transport modeling, and for design and operation of in situ remediation in complex hydrogeologic environments. We use transient hydraulic tomography to estimate 3D K in a volume of 15-m diameter by 20-m saturated thickness in a highly heterogeneous unconfined alluvial (clay to sand-and-gravel) aquifer with a K range of approximately seven orders of magnitude at an active industrial site in Assemini, Sardinia, Italy. A modified Levenberg-Marquardt algorithm was used for geostatistical inversion to deal with the nonlinear nature of the highly heterogeneous system. The imaging results are validated with pumping tests not used in the tomographic inversion. These tests were conducted from three of five clusters of continuous multichannel tubing (CMTs) installed for observation in the tomographic testing. Locations of high-K continuity and discontinuity, juxtaposition of very high-K and very low-K lenses, and low-K "plugs" are evident in regions of the investigated volume where they likely would not have been identified with interpolation from 1D profiles at the positions of the pumping well and five CMT clusters. Quality assessment methods identified a suspect high-K feature between the tested volume and a lateral boundary of the model.