Geoscience Reference
In-Depth Information
Carbonates (mixed pore type)
North Sea Rotliegendes Fm
Crevasse splay sandstone
Shallow marine rippled micaceous sst
Fluvial lateral accretion sst
Distributary/tidal channel Etive sst
Beach/stacked tidal Etive Fm.
Very heterogeneous
Heterolithic channel fill
Shallow marine HCS
Shallow marine high contrast lamination
Shallow marine Lochaline sst
Shallow marine Rannoch Fm
Aeolian interdune
Shallow marine SCS
Heterogeneous
Large scale cross-bed channels
Mixed aeolian wind ripple/grainflow
Fluvial trough-cross beds
Fluvial trough-cross beds
Shallow marine low contrast lamination
Homogeneous
Aeolian grainflow
Aeolian wind ripple
Homogeneous core plugs
Synthetic core plugs
0
1
2
3
4
Coefficient of Variation, C v
Fig. 3.15 Reservoir heterogeneity for a large range of reservoir and outcrop permeability datasets ranked by the
Coefficient of Variation, C v , (Redrawn from Corbett and Jensen 1992 )
The C v can be estimated from a sample by:
how confident one can be given a limited dataset.
The N o statistic indicates the sample number
required to estimate the true mean to with a
20 % tolerance (at a 95 % confidence level) as a
function of the Coefficient of Variation, C v :
N o ¼
C V ˃
ðÞ
p
p
ð
3
:
16
Þ
where
(p) and p are the standard deviation and
mean of the sample.
Corbett and Jensen ( 1992 ) proposed a simple
classification of C v values using a large selection
of permeability data from petroleum reservoirs
and outcrop analogues (Fig. 3.15 ):
￿C v <
˃
2
Þ
If the actual sample number is significantly
less than N o then a clear inference can be made
that the sample is insufficient and that the sample
statistics must be treated with extreme caution.
For practical purposes we can use N o as rule
of thumb to indicate data sufficiency (e.g.
Table 3.2 ). This simple approach assumes
a Gaussian distribution and statistical
representivity of the sample, so the approach is
only intended as a first approximation. More
precise estimation of the error associated with
ð
10 C v
Þ
ð
3
:
17
0.5 implies an effectively homogeneous
dataset
￿ 0.5
<
C v <
1 is termed heterogeneous
￿C v >
1 is termed very heterogeneous
The N-zero (N o ) statistic (Hurst and Rosvoll
1991 )
captures
these underlying statistical
theories into a practical guideline for deciding
Search WWH ::




Custom Search