Given an established probabilistic description of hydraulic conductivity, realizations of the hydraulic conductivity field can be generated using a lattice sampling technique, a special case of Latin hypercube sampling, and subsequently input to a groundwater flow and transport model. Realizations of dependent variables such as contaminant concentration are obtained as the output of the model and their corresponding statistics can be calculated. Compared with the other three alternative methods for generating a random hydraulic conductivity field, the suggested method is more efficient in terms of computational effort because it needs fewer samples to achieve the same statistical accuracy.