Fluid transport in the earth's crust is either extremely rapid, or extremel
y slow. Cracks, dikes and joints represent the former while tight crystalli
ne rocks and impermeable fault gouge/seals represent the latter. In many ca
ses, the local permeability can change instantaneously from one extreme to
the other. Instantaneous permeability changes can occur when pore pressures
increase to a level sufficient to induce hydro-fracture, or when slip duri
ng an earthquake ruptures a high fluid pressure compartment within a fault
zone. This 'toggle switch' permeability suggests that modeling approaches t
hat assume homogeneous permeability through the whole system may not captur
e the real processes occurring. An alternative approach to understanding pe
rmeability evolution, and modeling fluid pressure-controlled processes, inv
olves using local permeability rules to govern the fluid pressure evolution
of the system. Here we present a model based on the assumption that permea
bility is zero when a cell is below some failure condition, and very large
locally (e.g. nearest neighbors) when the failure condition is met. This to
ggle switch permeability assumption is incorporated into a cellular automat
on model driven by an internal fluid source. Fluid pressure increases (i.e.
from porosity reduction, dehydration, partial melt) induce a local hydro-f
racture that creates an internally connected network affecting only the reg
ions in the immediate neighborhood. The evolution, growth, and coalescence
of this internal network then determines how fluid ultimately flows out of
the system when an external (drained) boundary is breached. We show how the
fluid pressure state evolves in the system, and how networks of equal pore
pressure link on approach to a critical state. We fmd that the linking of
subnetworks marks the percolation threshold and the onset of a correlation
length in the model. Statistical distributions of cluster sizes show power
law statistics with an exponential tail at the percolation threshold, and p
ower laws when the system is at a critical state. The model provides insigh
ts into mechanisms that can establish long-range correlations in flow netwo
rks, with applications to earthquake mechanics, dehydration, and melting. (
C) 2000 Elsevier Science B.V. All rights reserved.