Page 75 - Well Logging and Formation Evaluation

Saturation/Height Analysis 65

Having derived a and b, you are now in the same position as you were
at the end of the P c -curve averaging exercise. Generic saturation/height
functions can be derived in an identical manner as before. The curves
should be compared with those derived from P c , or “cap,” curves and an
attempt made to explain any differences. Differences may arise from the
following:

• The core measurements, particularly if performed using air/mercury,
may be unrepresentative of the reservoir as seen by the logs.
• There may be a discrepancy in the permeability transform used for the
core plugs compared with that used on the logs.
• The assumed position of the FWL may be wrong.
• The logs may be influenced by other effects, making either the porosi-
ties or the saturations calculated erroneous.
• Depletion may have occurred prior to logging.

In cases where the discrepancies cannot be explained, it may be con-
sidered reasonable to take the more pessimistic set of functions as a “low”
and the optimistic ones as a “high” and construct a baseline lying halfway
between.
Having decided on a baseline function, the logs should be reevaluated
using the J function in reverse to recalculate saturations. Considering that
the function will use only the curves for height above FWL and poros-
ity/k as input (all derived from the density log), the results may prove sur-
prising. Having applied this technique on more than 20 reservoirs, I have
always been amazed at how reasonable the results look, and often how
well they compare with the resistivity-derived S w values, except in the thin
beds, where resistivity-derived S w is usually too high. The density log will
typically have a vertical resolution of 1ft, far superior to the deep induc-
tion or laterolog. The resulting increase in average oil or gas saturation
will typically be on the order of 20%.
I believe that a properly calibrated Leverett J function, reconciling reli-
able log saturations with core data, represents the best possible way to
calculate saturations in a reservoir and to transfer this information into
static and dynamic models. It should also be recognized that where core
data are reliable but the position of the FWL is unknown, the function
may be used to predict the depth of the FWL from the log saturations. In
general I have not had much success in applying this technique, but it may
merit further investigation.