Evaluation of underground contamination is very important to select the proper location for housing and to evaluate the potential risks of water resources. In this paper, a self-contained Underground Contamination Visualization Package (UCVP) is proposed to estimate underground contamination by using neural networks. In the paper, the accuracy of the backpropagation-type neural network is evaluated and the structure of the package is presented. The UCVP is capable of establishing a model automatically from the data of randomly distributed test wells and to estimate the concentration of contaminants at the site without requiring any additional information. The user may check the concentration by moving the cursor to any desired location on the map or XY, YZ and ZX cross-sections of the site are simultaneously displayed.
