Abstract of Doctoral Thesis


Many areas of the world use groundwater as the main sources of fresh water supply. In order to cope with the increasing demand of water due to population expansion, and industrial and agricultural growth, the exploitation of groundwater becomes unavoidable in many parts of the world. Due to the unplanned exploitation of groundwater from the coastal aquifers, the saltwater from the sea or ocean enters into the freshwater aquifers near the seacoast. This intrusion of saltwater into the freshwater aquifers near the coastal region contaminates the freshwater aquifer and makes it unusable for further human utilization. The remedial cost of this contaminate aquifer is also very high and time consuming. Therefore, suitable management policies have to be taken in order to arrest further degradation of the aquifers due to saltwater intrusion.


Saltwater intrusion management models can be used to evolve optimal and efficient management strategies for controlling saltwater intrusion in coastal aquifers. In order to obtain physically meaningful optimal management strategies, the physical processes involved need to be simulated while deriving the management strategies. The heuristic search technique, Genetic Algorithm (GA) may be used as a tool for solving the optimum management model, because of its relative efficiency in identifying global optimal solutions especially for nonlinear non-convex problems.


The simulation of the flow and transport processes involved in coastal aquifers is difficult as the density dependent flow and transport processes are needed to be modeled. Incorporation of this simulation model within an optimization based management model is complex and difficult. However, as an alternative, it is possible to link the simulation model externally with an optimization based management model. The GA based optimization approach is especially suitable for externally linking the numerical simulation model within the optimization model. Further efficiency in computational procedure can be achieved for such a linked model, if the simulation process can be simplified by sufficiently accurate approximation, as very large number of iterations between the optimization and simulation model is generally necessary to evolve an optimal management strategy. A possible approach for approximating the simulation model is to use a trained Artificial Neural Networks (ANN) as the approximate simulator of the flow and transport processes in coastal aquifers. This trained ANN linked to a GA based optimization model can be useful in evolving management strategies for coastal aquifers. Therefore, a Neural Network-Genetic Algorithm (ANN-GA) based linked simulation optimization model is developed for evolving optimal management of saltwater intrusion in coastal aquifers, using spatially and temporally varying optimal pumping strategies.


In the first step, trained ANN model is developed as an approximate simulator of the three-dimensional density dependent flow and transport processes in the coastal aquifer. A linked simulation optimization model is then formulated to link the trained ANN with a GA based optimization model for solving saltwater management problems. Single and multiple objectives optimal saltwater intrusion management models are developed. Real coded GA incorporating elitism is utilized to solve the single objective optimization problem. The multi-objective optimization problem is solved using Non-dominating Sorting Genetic Algorithm II (NSGA II).


The performance of the ANN model, as simulator for flow and transport processes in coastal aquifer, is evaluated for illustrative study areas, and is found to be satisfactory. The developed ANN model is simple in concept and computationally less time consuming compared to other numerical simulation models for simulating flow and transport processes in coastal aquifers. The ANN model takes considerably less CPU time than other numerical methods.


The performance of the linked simulation optimization (ANN-GA) model is also evaluated using an illustrative study area comprising of a hypothetical coastal aquifer. The evaluation results show the potential applicability of the developed methodology using ANN and GA based linked simulation optimization model for optimal management of coastal aquifers. The formulated single and multiple objectives management models are also solved using the embedded optimization technique for comparison of these solution results with solution results obtained using the ANN-GA approach for a given study area. The nonlinear algorithm available in MINOS is used to solve the embedded optimization model. The Pareto-optimal solutions for the multiple objectives optimization problem are obtained using the e-constraint method. The evaluation shows that the solutions obtained using the ANN-GA model and the embedded optimization model are comparable.