Identification of Groundwater Contaminant Sources and Release Histories Using Genetic Algorithms
Mustafa M. Aral and Jiabao Guan
Multimedia Environmental Simulation Laboratory
School of Civil and Environmental Engineering
Georgia Institute of Technology
Atlanta, Georgia 30332, USA
Abstract
Recovering the release histories and identifying the locations of contaminant sources in heterogeneous aquifers are important and also challenging problem in the field of contaminant geohydrology. In this study, the source identification problem was formulated as a nonlinear optimization model, in which contaminant source locations and release histories were defined as explicit unknown variables. The optimization model selected is the standard model, where residuals between the simulated and measured contaminant concentrations at observation sites are minimized. In this formulation, simulated concentrations at the observation locations are implicitly embedded into the optimization model through the solution of groundwater flow and contaminant fate and transport simulation models. It is well known that repeated solutions of these models, which are a necessary component of the optimization process, dominate the computational cost and adversely affect the efficiency of this approach. To simplify this computationally intensive process, a new combinatorial approach, identified as Progressive Genetic Algorithm (PGA), is proposed for the solution of the nonlinear optimization model. PGA is a subdomain method, which links the standard genetic algorithm (GA) with groundwater simulation models within a subdomain. In this approach the optimization process is divided into iteration and search stages. During the iteration stage, groundwater simulation models are run to generate a subdomain optimization model, and in the search stage, GA is used to search for the optimal solution in the subdomain. During this iterative cycle, the position and the size of the subdomain continuously changes, yielding an efficient solution process. Numerical results show that the proposed approach provides a robust tool for the solution of groundwater contaminant source identification problem. The proposed approach may also be used in the solution of other nonlinear optimization problems, which are computationally intensive. These application areas include optimal groundwater remediation design, optimal monitoring well system design, optimal management of groundwater resources and surface water groundwater interaction and optimal management of water supply systems.