The Influence of Partly Unobservable Areas on Fracture Network Characterization
Zeeb, Conny 1; Blum, Philipp 1; Gomez-Rivas,
Enrique 2; Stark, Lena 2; Grabowski, David 2;
Bons, Paul D.2
(1)Institute for Applied Geosciences, Karlsruhe
Institute of Technology, Karlsruhe, Germany. (2) Department of Geosciences,
Eberhard Karls University, Tübingen, Germany.
A common method to evaluate
the degree of fracturing in the
subsurface is the sampling of fracture characteristics at analogue outcrops or
from well cores. They allow the generation of artificial discrete fracture
networks (DFN), which can then be used to predict transport through a fractured
rock mass by means of numerical simulations. Key parameters for the generation
of DFN are density (number of fractures per unit area), length (e.g. fractal
dimension) and orientation. The term discontinuity is used here to describe
various kinds of mechanical defects, such as fractures, joints, veins, etc.
However, outcrops are often covered or well sections can be damaged, so that
discontinuities are difficult to impossible to identify. The presence of
vegetation, debris, or damaged parts of a well core, prevents a complete
sampling of discontinuities and thus increases the degree of sampling bias. The
term cover is used to account for all factors that render an outcrop or well
core partly unobservable. Our aim is to investigate
how
, and to which extent,
cover influences sampling bias and causes deviations of the estimated key
parameters from the true values.
Before investigating natural discontinuity systems we quantify the
effect of cover using artificial 2D discontinuity networks with known input
parameters. The percentage of cover is increased stepwise. We compare
the
results by applying several standard sampling methods: 1) window sampling, 2)
scanline sampling, and 3) circular scanlines. These methods are affected
differently by sampling bias, and thus by cover. Window sampling is mainly
affected by censoring, whereas scanline sampling is strongly affected by
truncation, since shorter discontinuities have a lower chance of being
intersected by the scanline than the longer ones. Circular scanlines and window
sampling are not subjected to sampling bias, since these are maximum likelihood
estimators.
In addition, we also investigated the degree of uncertainty in
density and length distribution estimates due to sampling bias. Knowing the
efficiency, limitations and possible corrections of each method has allowed us
to determine the best sampling technique depending on the outcrop situation and
to optimize the time required to adequately capture the properties of a
discontinuity network. We show how
the performance of the different methods
changes with increasing percentages of cover and apply this knowledge to
different examples of natural discontinuity systems.
AAPG Search and Discovery Article #90135©2011 AAPG International Conference and Exhibition, Milan, Italy, 23-26 October 2011.