Groundwater
Contaminant Transport Dr.
Darrell Pepper , Dr. David Carrington
Overview GWADAPT is a finite-element model which incorporates mesh
(h-) adaptation to calculate ground-water flow and pollutant transport. The
formulation is based on the equations for conservation of mass, Darcy's law
for an anisotropic medium, and the time-dependent species transport equation.
Modifications are used in the finite-element formulation to enhance computational
speed and reduce storage; Petrov-Galerkin weighting of the advection terms
provides numerical stability. An explicit time marching scheme is used to
solve the transient equations. By utilizing unstructured adaptive meshing,
species concentration and location of steep fronts are accurately resolved,
even though one begins with a coarse mesh. The algorithm is written in C/C++,
and runs on PCs under WINDOWS; versions are also available for workstations
and a Cray super-computers.
Introduction
Groundwater contamination has become an important environmental issue which
poses a serious threat to drinking water quality. Such problems involve complicated
physics, chemistry, and multi phase flow phenomena. The simulation of contaminant
transport through the subsurface is necessary in order to effectively design
mitigation methods for cleanup and prevention of the deterioration of ground
water. The transport of contaminants in groundwater systems requires reliable
predictions to assess potential hazards to the public. Analytical methods
have limited capability to accurately predict complex dispersion patterns
where geometry and soil conditions vary over a wide range. Numerical models
are more flexible in solving such complex processes and yield realistic solutions,
but the cost and amount of time involved in the computation of these multi
dimensional problems can be considerable.
Current Results
An adapting finite element model, GWADAPT, has been developed for calculating
contaminant transport and ground water flow. Simple modifications to the basic
Galerkin formulations are used to enhance speed and reduce storage; a Petrov-Galerkin
weighting scheme is used for the advection terms. Mesh adaptation is achieved
using interpolation based commands and averaging to refine/unrefine the mesh.
An explicit, second-order accurate in time Runge-Kutta method is employed
to advance the transient solutions; the solution times are fast on PCs, depending
on the mesh density, and indicate the potential for significant payoff in
overall solution time for three-dimensional problems. GWADAPT is fast enough
in 2-D to allow transient solutions to be generated on enhanced 486 and Pentium
class PCs without significant waiting, providing moderately sized meshes are
used. GWADAPT has been written in C/C++ to provide ease of portability to
other machines, and quickly display graphical output of the model results,
as well as mesh changes. GWADAPT also runs quickly on workstation class machines
(which allow easy interface to real time graphics capability). Savings in
computer time and storage amount to several factors using locally adapting
meshes. Preliminary tests of the model show promise for execution on vector
super-computers as well as massively parallel computers. Extension of the
two-dimensional version of GWADAPT to three dimensions has been undertaken.
Testing of the 3-D version is still being conducted. The ability to perform
3-D adaptation with speed and prevent CPU memory overload (as a result of
new nodal numbering and bandwidth increases) is rather tricky, especially
when running on small machines. The 3-D version of GWADAPT should be available
in 1997, and will be primarily aimed at the workstation level (although a
reduced version will be available for the PC).