Page 71 - Geology of Carbonate Reservoirs

52 CARBONATE RESERVOIR ROCK PROPERTIES

that it is important to know the limitations on some of the software applications.
Some of the programs can make calculations to depict up to fi ve mineral species in
the reservoir being studied; however, the analyst must choose which minerals are
most likely to occur in the formation and therefore focus the calculations on those
minerals only. Problems result from having to choose which minerals to exclude.
For example, consider the case where two petrophysicists are in competition during
litigation and have to make choices about which minerals to exclude from a list that
includes quartz, clay minerals, anhydrite, dolomite, and calcite. If the petrophysicists
exclude different minerals from their calculations, the two outcomes will be differ-
ent. At minimum, different results will cause uncertainty about the “ true ” lithology.
Consider a situation where reservoir quality is related to dolomite content of the
producing formation. The petrophysicist who minimized the importance of dolomite
by her choices of which minerals to include in the log calculations also minimized

her chances of success in reservoir characterization, flow unit mapping, and eco-
nomic evaluations.

Rider (1996) discusses multilog quantification of lithology and divides the
methods into two main categories: petrophysical multilog analysis and statistical
multilog analysis. The former method is “ essentially one of solving a number of
linked, simultaneous equations, for unknown volumes of chosen minerals or matri-

ces defined by pure, end - member (hypothetical) log responses. ” If the end members
are pure limestone, dolostone, and evaporites, the method works reasonably well
because the responses are usually linear. In the presence of shale, the results are
unpredictable, however (Rider, 1996 ). This problem can be tackled by generating
computer output of volume percent of each ideal, end - member component. This
generates a “ CPI ” log (computer processed interpretation). A complaint with the
method is that the ideal end members are artifi cially defined absolutes that have

little or no relation to fundamental rock properties (Rider, 1996 ). For example, CPI

logs define sandstones on the basis of their quartz content but do not represent
fundamental properties such as texture and sedimentary structures.
Statistical multilog analysis involves taking all log responses from a single
depth and combining them into a multidimensional set in n - dimensional space.
The sets are then subjected to multivariate statistical analyses to identify sets that
can be grouped into populations of numbers that have some internal similarity
and that can be differentiated from other populations of numbers. Think of cluster
analysis dendrograms. The next step is to try to relate the different number popu-
lations to rock types or synthetic lithofacies. The statistical populations are some-
times called “ statistical electrofacies, ” but as Rider (1996) points out, there is a
great distance between what geologists call facies and what is produced by mul-
tivariate statistical analysis of log values. In sand − shale sequences this may not
be such a problem because siliciclastic grains and shale are relatively easy to
distinguish on gamma ray and resistivity logs, among others. But in carbonates
where distinctions between constituent components, fabrics and textures, and pore
types are not readily distinguishable by borehole log measurements, it is a genuine
problem — a problem that can only be resolved with certainty by direct observa-
tion of the rocks.
A borehole log that offers great potential for geologists is the NMR (nuclear
magnetic resonance) log. The NMR log records the time required for the liquid in

a liquid - filled pore to change from an excited state to a relaxed state. This time