Subsurface Characterization Using Artificial Neural Network and GIS

Abstract

A method for characterizing the subsurface is developed using an artificial neural network (ANN) and geographic information system (GIS). Data on the distribution of aquifer materials from monitoring well lithologic logs are used to train a multilayer perceptron using the back-propagation algorithm. The trained ANN predicts using an appropriate prediction scale, the subsurface formation materials at each point on a discretized grid of the model area. GIS is then used to develop subsurface profiles from the data generated using the ANN. These subsurface profiles are then compared with available geological sections to check the accuracy of the ANN-GIS generated profiles. This methodology is applied to determine the aquifer extent and calculate aquifer parameters for input to ground-water models for the multiaquifer system underlying the city of Bangkok, Thailand. A selected portion of the model domain is used for illustration. The integrated approach of ANN and GIS is shown to be a powerful tool for characterizing complex aquifer geometry, and for calculating aquifer parameters for ground-water flow modeling.