Conductivity estimates Kp, from packer tests are the normal basis for sto-
chastic continuum (SC) models of sparsely fractured rock. The scale of the tests is often
much smaller than the scale on which the rock acts as an equivalent porous medium
(EPM). Therefore extrapolation of test results to the scale of blocks in a SC model, based
on continuum assumptions, is questionable. This paper demonstrates an alternative
method for extrapolation of Kp, data, based on discrete fracture network (DFN) model-
ing. 3-D rock blocks were simulated based on fracture geometric and packer test data
from Finnsjtn, Sweden. Constant-pressure packer tests were simulated in these blocks,
and the results were compared with field data to validate the fracture model. Effective
block-scale conductivities Kb, were estimated by modeling steady-state flow through the
blocks. Probabilistic estimates were obtained of the relationship between Kp, estimates
and Kb, for block scales from 15 to 50 m. The DFN model predicts that the correlations
between Kp, and Kb, are poor, even for the smaller blocks.
Predictive modeling of water flow in sparsely fractured, low-permeability rock is
required for design of high-level nuclear waste repositories and other types of under-
ground facilities such as gas storage caverns. Problems in modeling this rock arise from
the tendency for the flow to be concentrated within fractures. Interconnections among
fractures to form flow paths can be highly irregular on scales of concern for site model-
ing. This is evident from the high variability of heads seen in the vicinity of tunnels in
tight, sparsely fractured rock.
To model this rock, its strongly heterogeneous and effectively random nature must be
represented. Two types of models which do this are:
-Discrete fracture network (DFN) models.
-Stochasticontinuum (SC) models.
DFN models represent the discrete flow paths explicitly, whereas SC models represent
blocks of the rock mass in terms of an "equivalent porous medium" (EPM). The more
detailed representation used in DFN models has thus far restricted use of these models to
rock volumes smaller than (200 m)3. SC models are viewed as being more tractable for
site-scale modeling.
Unfortunately, the properties of rock blocks on the scale of the constituent blocks of
SC models cannot be measured directly in the field. Therefore it is difficult to obtain
answers to two critical questions:
-Does the rock behave as an EPM on the block scale?
-If so, then how should the block-scale conductivities be estimated from packer tests?
The first question has been treated extensively in the literature (e.g., Long et al., 1982;
Khaleel, 1989). This paper focuses primarily on the second question, which has not thus
far been examined using a discrete fracture approach.