MAPPING OF SUBSURFACE CONTAMINANT PROFILES BY NEURAL NETWORKS: Fid-Bau Portal

MAPPING OF SUBSURFACE CONTAMINANT PROFILES BY NEURAL NETWORKS

Tansel, Berrin Associate Professor / Jordahl, Claire Project Engineer / Tansel, Ibrahim Associate Professor

Determining the extent of contamination is a frequently encountered problem for private industry and governmental agencies. Numerous groundwater contaminant transport models have been developed involving mathematical relationships based on an understanding of the physical, chemical and microbiological processes that are thought to affect transport of contaminants in subsurface environments. This paper presents the application of neural networks for groundwater contaminant transport modeling. Recently, artificial neural networks (NN) have been successfully used in many different applications. The main advantages of NNs for contaminant transport modeling are their ability to develop contaminant profiles with limited data. The data used for the study were obtained from the monitoring studies conducted for the small potable water wells in Dade County, Florida. The water wells used in the study have less than 100,000 gpd capacity and are located within 1/4 mile of underground petroleum storage tank systems. Laboratory analysis by EPA method 524.2 resulted in the identification and quantification of volatile organic compounds in the wells sampled. An area contaminated from a specific underground storage tank was selected for this NN application. The locations of the contaminated wells in this area were mapped and a 2-dimensional grid system was developed for the area. Input to the neural network program for the learning set consisted of the coordinates of the contaminated wells and the contaminant concentration. The subsurface contaminant profiles for different types of petroleum organics were developed.