Artificial Neural Networks Approximation of Density Dependent Saltwater Intrusion Process in Coastal Aquifers

Abstract

The flow and transport processes in a coastal aquifer are highly nonlinear, where both the flow and transport processes become density dependent. Therefore, numerical simulation of the saltwater intrusion process in such an aquifer is complex and time consuming. An approximate simulation of those complex flow and transport processes may be very useful, if sufficiently accurate, especially where repetitive simulations of these processes are necessary. A simulation methodology using a trained artificial neural network model (ANN) is developed to approximate the three-dimensional density dependent flow and transport processes in a coastal aquifer. The data required for initially training the ANN model is generated by using a numerical simulation model (FEMWATER). The simulated data consisting of corresponding sets of input and output patterns are used to train a multilayer perceptron using the back-propagation algorithm. The trained ANN predicts the concentration at specified observation locations at different times. The performance of the ANN as a simulator of the density dependent saltwater intrusion process in a coastal aquifer is evaluated using an illustrative study area. These evaluation results show that the ANN technique can be successfully used for approximating the three-dimensional flow and transport processes in coastal aquifers.