Abstract: Risk Assessment in Seismic Direct Hydrocarbon Detection
Truxillo, Stanton G. - Amoco Corporation
Fifteen years of industry experience
with AVO has produced numerous successes, but the early rule of thumb associating
positive AVO with hydrocarbons led to occasional disappointments. Experience
showed that hydrocarbon AVO signatures vary with hydrocarbon type, age,
depth, pressure, nature of contact between seal and reservoir, thickness,
and other variables. Several technological developments improved success
over time. Longer offsets permit better estimates of gradient values, including
anisotropic effects. 3D data permit integration of AVO analysis
into stratigraphic
and structural interpretation. Processing workstations permit interactive
testing of processing parameters, especially moveout or migration
velocity
,
leading to better tracking of amplitudes across CDP or CRP
gathers
. Pre-stack
migration better-positions energy from dipping events. Dipole sonic logs
provide measured shear data in place of generic estimators. Rock properties
databases permit more-sophisticated AVO modeling. The resulting technical
diversity created pressure for a systematic approach to assessing DHI-related
prospects: what confidence should we give to different approaches to DHI
analysis
,
using
different amounts and quality of data, in different parts
of the world? In this study, a datum point is defined as a horizon for
which a pre-drill AVO or stack amplitude
analysis
had been done, and the
results of the first well to penetrate that horizon. Subsequent wells in
the same reservoir are excluded from the database. By these criteria a
well may be a commercial success but score as multiple AVO failures if
the pay predictions for individual reservoirs were incorrect. Prediction
accuracy may be analyzed in terms of presence of hydrocarbons, or of minimum
economic case hydrocarbons. Figure 1 shows the results of Bayesian
analysis
of economic results for AVO.
The simple analysis
above does not reflect differences
in confidence due to more or better data. To quantify this difference,
a model-based rating system for AVO analyses was developed which mimics
the mental checklist an experienced geophysicist might follow, assigns
values to "correct" answers, and predicts the confidence of this
analysis
in the exploration decision process. This algorithm met several design
criteria:
- systematic and objective, applicable anywhere in the world;
- independent of particular computational attributes used;
- recognize uncertainty: more or better data should increase confidence;
- scientifically sound, comparing the match between predictions and observations;
- geologically consistent: the anomaly should fit the trap, and downdip should conform to structure.
The database records age, depth, pressure, etc., in addition to drilling results. Confidence numbers are scaled from +1 to -1. Positive values predict that hydrocarbons are present, negative values predict that hydrocarbons are not present. Increasing absolute value indicates increasing confidence in that prediction.
To date there are 66 drilled AVO cases in the database. Figure 2 shows the correlation of the prediction algorithm with drilling results.
AAPG Search and Discovery Article #90933©1998 ABGP/AAPG International Conference and Exhibition, Rio de Janeiro, Brazil