HR: 1340h
AN: H53A-1233    [Abstracts]
TI: 3-D Source Identification of Borden Emplacement Site
AU: * Jin, X
EM: xjin@ncsu.edu
AF: North Carolina State Uiversity, Dept. of Civil Engineering, North Carolina State University, Raleigh, NC 27695-7908 United States
AU: Mahinthakumar, K
EM: gmkumar@ncsu.edu
AF: North Carolina State Uiversity, Dept. of Civil Engineering, North Carolina State University, Raleigh, NC 27695-7908 United States
AB: Finding the location and concentration of contaminant sources is an important step in groundwater remediation and management. This typically requires the solution of an inverse problem. In this study, an optimization-based approach is used for solving a three-dimensional field scale source identification inverse problem. Inverse modeling using optimization techniques requires many solutions of the corresponding forward problem and is computationally intensive. Fortunately, with the wide-spread use of parallel computing technology, solving inverse problems such as these are increasingly becoming feasible. We have developed a parallel hybrid optimization framework that can solve various compute-intensive groundwater inverse problems. In this study, the identification of the emplaced source at Borden site, Ontario, Canada, as described by Rivett, Feenstra and Cherry 2001 (RFC 2001), is carried out using this optimization framework by employing the supercomputing environment at North Carolina State University. The forward simulation model was first calibrated by selecting an appropriate heterogeneous hydraulic conductivity field that leads to a modeled contaminant plume that closely matches published field measurements reported by RFC2001. The randomly heterogeneous hydraulic conductivity field was generated by a turning bands algorithm using previously published statistical parameters for the Borden site (Sudicky 1985). The emplaced source location and concentration is then determined by using the hybrid optimization framework. The hybrid optimizer uses genetic algorithms (GAs) coupled with various local search approaches. In this study, real encoded GA (RGA), and two local search approaches, (i) Nelder-Mead simplex method, (ii) Fletcher and Reeves conjugate gradient method were used. The MPI (Message Passing Interface) communication library is used to implement parallelism in the optimizer and the simulators. Simulation results indicate that the combination of parallel computing and efficient optimization algorithms has enabled the solution of a field-scale three-dimensional source identification problem at the Borden site with over 90% accuracy in a timely manner.
DE: 3210 Modeling
DE: 3230 Numerical solutions
DE: 1831 Groundwater quality
DE: 1832 Groundwater transport
SC: Hydrology [H]
MN: 2004 AGU Fall Meeting