144
Statistical Methods for Estimating Petroleum Resources
Beaverhill Lake oil play is shown in Chapter 2 (Fig. 2.9). The horizontal
axis indicates the discovery sequence for the wells drilled, and the gaps
in the sequence represent the occurrence of dry holes. The upper vertical
axis indicates the individual discovered in-place pool sizes, whereas
the lower vertical axis indicates oil fl ow rates obtained from drill stem
tests. These pool sizes and drill stem test recoveries are the basic input
data required for resource assessment.
Oil or gas occurrences in a specifi c exploratory well can range in
magnitude from a discovery of commercial size to the show of oil
droplets or gas bubbles. Each occurrence can be considered, by defi -
nition, as a pool. In practice, however, an oil or gas accumulation is
considered to be a pool only if it is of commercial value at the time of
discovery. Imposing such a restricted defi nition on the underlying pool
population has a severe impact on the validity of the resource estimate,
because small pools in the population will be underrepresented and the
amount of information needed to determine the total number of pools
within a play will not be suffi cient.
It is essential, therefore, to examine all possible potential pools that
were not reported at the time of assessment. Although time- consuming
and tedious, this extensive collecting of data is rewarding. It is much
better to have an adequate data set for an assessment than to attempt
to model the economic truncation problem from ill-defi ned statistical
models. This is illustrated by the Beaverhill Lake example in Chapter 3.
Step 3: Validating Mixed Populations or
Lognormal Assumptions
Having collected all the pool data for a play, two aspects must be
validated: (1) the possible mixed populations and (2) the assumption of
lognormality if LDSCV is used. A logarithmic probability plot such as
that shown in Chapter 2 (Fig. 2.12) can be used to check whether these
two attributes exist.
If the discoveries are thought of as a single population, then the
empirical distribution function should exhibit an almost straight line
on the plot. Also, if the discoveries obey a lognormal distribution, they
should exhibit a straight line on the same plot. However, the statistical
assumption required by the logarithmic probability plot is that the discoveries are a random sample from their population. This assumption,
as we know, is not valid. The conclusion obtained from the plot is that
there is no evidence to negate the hypothesis. Further Q–Q tests must
be executed using the output derived by NDSCV.
Statistical Methods for Estimating Petroleum Resources
Beaverhill Lake oil play is shown in Chapter 2 (Fig. 2.9). The horizontal
axis indicates the discovery sequence for the wells drilled, and the gaps
in the sequence represent the occurrence of dry holes. The upper vertical
axis indicates the individual discovered in-place pool sizes, whereas
the lower vertical axis indicates oil fl ow rates obtained from drill stem
tests. These pool sizes and drill stem test recoveries are the basic input
data required for resource assessment.
Oil or gas occurrences in a specifi c exploratory well can range in
magnitude from a discovery of commercial size to the show of oil
droplets or gas bubbles. Each occurrence can be considered, by defi -
nition, as a pool. In practice, however, an oil or gas accumulation is
considered to be a pool only if it is of commercial value at the time of
discovery. Imposing such a restricted defi nition on the underlying pool
population has a severe impact on the validity of the resource estimate,
because small pools in the population will be underrepresented and the
amount of information needed to determine the total number of pools
within a play will not be suffi cient.
It is essential, therefore, to examine all possible potential pools that
were not reported at the time of assessment. Although time- consuming
and tedious, this extensive collecting of data is rewarding. It is much
better to have an adequate data set for an assessment than to attempt
to model the economic truncation problem from ill-defi ned statistical
models. This is illustrated by the Beaverhill Lake example in Chapter 3.
Step 3: Validating Mixed Populations or
Lognormal Assumptions
Having collected all the pool data for a play, two aspects must be
validated: (1) the possible mixed populations and (2) the assumption of
lognormality if LDSCV is used. A logarithmic probability plot such as
that shown in Chapter 2 (Fig. 2.12) can be used to check whether these
two attributes exist.
If the discoveries are thought of as a single population, then the
empirical distribution function should exhibit an almost straight line
on the plot. Also, if the discoveries obey a lognormal distribution, they
should exhibit a straight line on the same plot. However, the statistical
assumption required by the logarithmic probability plot is that the discoveries are a random sample from their population. This assumption,
as we know, is not valid. The conclusion obtained from the plot is that
there is no evidence to negate the hypothesis. Further Q–Q tests must
be executed using the output derived by NDSCV.