Ms. Edwards et P. Grindrod, A CHANNEL NETWORK MODEL FOR CHEMICAL MIGRATION IN SUBSURFACE MEDIA, Mathematical models and methods in applied sciences, 5(5), 1995, pp. 641-657
We consider the migration of chemical species through saturated hetero
geneous media, where the scale of the flow path geometry is large enou
gh to invalidate a Fickian based, representative elementary volumes, a
pproach to dispersal. Channel network models combine the effects of ch
anneling over small spatial scales with network mixing which gives ris
e to an equivalent dispersion effect over large spatial scales. We ext
end equations and solutions given previously for ideal nonabsorbing tr
acers, to include the retardation of chemically active species. The pr
imary aim is to derive explicit solutions for future calibration purpo
ses, and illustrate scale dependent dispersive behavior with and witho
ut chemical absorption processes. In this paper we develop a twin chan
nel model allowing for various specification of the retardation proces
s. This allows the retardation to be either positively or negatively c
orrelated with pathway apertures (which are in turn related directly t
o flow rates of the respective channels). We show that if absorption i
s controlled by specific surface area, then a scaling argument infers
that highly transmissive paths are also less retarding. We illustrate
the model's applicability to such cases. We shall discuss the scale de
pendence of calculated (observed) equivalent dispersivities obtained b
y analyzing the moments of breakthrough curves at various distances. F
or a range of parameter values we exhibit scaling behavior similar to
that observed in the field and laboratory. This does not necessarily i
mply any validity to the approach: it is a direct consequence of defer
ring the network - mixing effect to large length scales. An important
point of consideration is that wherever conceptual model choices must
be made, in modeling natural processes such as chemical migration, the
re is a possibility of biasing calculations by sole reliance on a part
icular approach. Channel network models marry channeling models and cl
assical Fickian models. Thus in calibrating a model such as the Twin C
hannel model from given experimental behavior, we can immediately esti
mate over which scales channeling or Fickian dispersive effects are li
kely to be dominant.