NUMERICAL MODELING OF MASS-TRANSPORT IN HYDROGEOLOGIC ENVIRONMENTS - PERFORMANCE COMPARISON OF THE LAPLACE TRANSFORM GALERKIN AND ARNOLDI MODAL REDUCTION SCHEMES
Da. Farrell et al., NUMERICAL MODELING OF MASS-TRANSPORT IN HYDROGEOLOGIC ENVIRONMENTS - PERFORMANCE COMPARISON OF THE LAPLACE TRANSFORM GALERKIN AND ARNOLDI MODAL REDUCTION SCHEMES, Advances in water resources, 21(3), 1998, pp. 217-235
The Laplace Transform Galerkin (LTG) method and the Arnoldi modal redu
ction method (AMRM) have been implemented in finite element schemes de
signed to solve mass transport problems in porous media by Sudicky [Su
dicky, E.A., Water Resour. Res., 25(8) (1989) 1833-46] and Woodbury et
al [Woodbury, A.D., Dunbar, W.S., & Nour-Omid, B., Water Resour. Res.
, 26(10) (1990) 2579-90]. In this work, a comparative analysis of the
two methods is performed with attention focused on efficiency and accu
racy. The analysis is performed over one- and two-dimensional domains
composed of homogeneous and heterogeneous material properties. The res
ults obtained using homogeneous material properties indicate that for
a given mesh design the LTG method maintains a higher degree of accura
cy than does the AMRM. However, in terms of efficiency, the Arnoldi at
tains a pre-defined level of accuracy faster than does the LTG method.
It is also shown that for problems involving homogeneous material pro
perties the solution obtained using the LTG method on a coarse mesh is
comparable in terms of solution time and accuracy to that obtained us
ing the AMRM on a fine mesh. Comparisons similar to those performed us
ing homogeneous material properties are also performed for the case wh
ere the hydraulic conductivity field is heterogeneous. For this case,
the level of accuracy achieved by the AMRM and the LTG method are simi
lar. However, as with the analysis involving homogeneous material prop
erties, the AMRM is found to be more efficient than the LTG method. It
is also shown that for heterogeneous material properties, use of the
LTG method under high grid Peclet conditions can be potentially proble
matic. Copyright (C) 1997 Elsevier Science Limited.