Interpretation
Interpreting geochemical data is probably the most important step in the process
and easily the most difficult. Understanding the limitations of the data is not easy
to convey or even describe. I wrote a paper for the APGE newsletter that helps
illustrate these limitations. Parenthetically, I haven't followed the
recommendations that I make in this paper in the survey section of this web site
because contouring of data is such an effective and satisfying display, however, a
pixel plot of your data along with a contour map will help you recognize the limits
of your data when interpreting.
and easily the most difficult. Understanding the limitations of the data is not easy
to convey or even describe. I wrote a paper for the APGE newsletter that helps
illustrate these limitations. Parenthetically, I haven't followed the
recommendations that I make in this paper in the survey section of this web site
because contouring of data is such an effective and satisfying display, however, a
pixel plot of your data along with a contour map will help you recognize the limits
of your data when interpreting.
Geochemical Survey Interpretation
Chuck Goudge, Graystone Exploration Labs, Apge Microseep Fall 1997 No. 57
Using geological contouring programs to present geochemical survey data,
although admittedly often helpful for data visualization, can be misleading and
misrepresent the precision of geochemical methods.
All Data Sets Are Not Created Equal
Subsurface data are appropriate to contour because of two important properties.
First, at each subsurface data point there exists an absolute depth to a particular
formation. Regardless of the number of times this measurement is made this
depth will not change. Second, subsurface data points are dependant upon each
other. This dependence makes the primary assumption of contouring true. This
assumption is that unknown points between datum lie along a smooth line of
transition, it fails when the distance between data points is large or when faulting
is present.
Geochemical data sets are more like the subsurface of Nevada than a typical
layer cake basin. A geologists would not input well data from Nevada into a
contouring program and expect a correct picture of the subsurface. The faulting
and the lack of well control destroys the value of the simplistic model used for
contouring. Similarly, geochemical data, which measures the variable flux of
highly volatile organic compounds, is the geological equivalent of a highly faulted
area. Geochemical measurements can and do jump between high to low over
distances measured in meters.
Geochemical Measurements Are Not Absolute
It can be stated with near certainty that no measurement of any geochemical
parameter can consistently or even occasionally yield exactly the same result.
Because most measurements depend on some form of quantitative analysis even
repeat measurements of the same sample will yield results that are no-better than
the precision of that particular analytical method. The geochemical and
environmental analytical industry only achieve a precision of plus or minus 20%
for many analysis and the best techniques rarely surpass plus or minus 5%.
However, the range of uncertainty for each data point is only partially
determined by this analytical precession.
Geochemical Data Areas
Each geochemical measurement is used to represent an area. This area can be
defined as a polygon stretching halfway to each adjacent data point. This data
area contains a large set of potential measurements. The density of sampling will
define the size of this area, and the larger the area the greater the probability that
the data set contained by the area will exhibit a wide distribution of potential
values, occasionally even mixtures of background and anomalous populations.
Clearly, as sample density decreases the uncertainty range for each measurement
increases. The total uncertainty for each point is the range of the distribution for
the data area, plus the analytical uncertainty. Because each data point represents
just a single sampling of this distribution, a certain percentage of the data points
will measure the tails of this distribution and reflect, if not technically incorrect
measurements, at least misleading measurements.
Geochemical Data Contouring
Based on the preceding discussion, geochemical data does not possess the
characteristics necessary to make the assumptions of contouring valid. Samples
with large uncertainties are being used to define "precise" intervening contours.
The precision of the pictures produced by contouring geochemical data appear
definitive, but in fact they are an exercise in speculation with only a small chance
of reflecting the actual seepage pattern.
Pixel Plots
Possibly the best presentation of geochemical data are pixel plots. The term pixel
has become familiar to many as it applies to computer graphics. If each
geochemical data area is considered a pixel and then assigned a color or pattern
representing it's value relative to the range of values of the survey area, the
information can be displayed without unfounded interpolations. Equally
important, the sample density and the "clarity" or "resolution" of the picture is
effectively communicated. The illusion of precise areas of anomaly produced by
contouring are eliminated and the display, although undoubtedly less satisfying,
is considerably more realistic.
Example - Implied Precision
The Friday Oil Field geochemical data is displayed below as a pixel plot and as a
contour map. The contour data is much more impressive but it exaggerates the
information actually available as displayed in the pixel plot.
Chuck Goudge, Graystone Exploration Labs, Apge Microseep Fall 1997 No. 57
Using geological contouring programs to present geochemical survey data,
although admittedly often helpful for data visualization, can be misleading and
misrepresent the precision of geochemical methods.
All Data Sets Are Not Created Equal
Subsurface data are appropriate to contour because of two important properties.
First, at each subsurface data point there exists an absolute depth to a particular
formation. Regardless of the number of times this measurement is made this
depth will not change. Second, subsurface data points are dependant upon each
other. This dependence makes the primary assumption of contouring true. This
assumption is that unknown points between datum lie along a smooth line of
transition, it fails when the distance between data points is large or when faulting
is present.
Geochemical data sets are more like the subsurface of Nevada than a typical
layer cake basin. A geologists would not input well data from Nevada into a
contouring program and expect a correct picture of the subsurface. The faulting
and the lack of well control destroys the value of the simplistic model used for
contouring. Similarly, geochemical data, which measures the variable flux of
highly volatile organic compounds, is the geological equivalent of a highly faulted
area. Geochemical measurements can and do jump between high to low over
distances measured in meters.
Geochemical Measurements Are Not Absolute
It can be stated with near certainty that no measurement of any geochemical
parameter can consistently or even occasionally yield exactly the same result.
Because most measurements depend on some form of quantitative analysis even
repeat measurements of the same sample will yield results that are no-better than
the precision of that particular analytical method. The geochemical and
environmental analytical industry only achieve a precision of plus or minus 20%
for many analysis and the best techniques rarely surpass plus or minus 5%.
However, the range of uncertainty for each data point is only partially
determined by this analytical precession.
Geochemical Data Areas
Each geochemical measurement is used to represent an area. This area can be
defined as a polygon stretching halfway to each adjacent data point. This data
area contains a large set of potential measurements. The density of sampling will
define the size of this area, and the larger the area the greater the probability that
the data set contained by the area will exhibit a wide distribution of potential
values, occasionally even mixtures of background and anomalous populations.
Clearly, as sample density decreases the uncertainty range for each measurement
increases. The total uncertainty for each point is the range of the distribution for
the data area, plus the analytical uncertainty. Because each data point represents
just a single sampling of this distribution, a certain percentage of the data points
will measure the tails of this distribution and reflect, if not technically incorrect
measurements, at least misleading measurements.
Geochemical Data Contouring
Based on the preceding discussion, geochemical data does not possess the
characteristics necessary to make the assumptions of contouring valid. Samples
with large uncertainties are being used to define "precise" intervening contours.
The precision of the pictures produced by contouring geochemical data appear
definitive, but in fact they are an exercise in speculation with only a small chance
of reflecting the actual seepage pattern.
Pixel Plots
Possibly the best presentation of geochemical data are pixel plots. The term pixel
has become familiar to many as it applies to computer graphics. If each
geochemical data area is considered a pixel and then assigned a color or pattern
representing it's value relative to the range of values of the survey area, the
information can be displayed without unfounded interpolations. Equally
important, the sample density and the "clarity" or "resolution" of the picture is
effectively communicated. The illusion of precise areas of anomaly produced by
contouring are eliminated and the display, although undoubtedly less satisfying,
is considerably more realistic.
Example - Implied Precision
The Friday Oil Field geochemical data is displayed below as a pixel plot and as a
contour map. The contour data is much more impressive but it exaggerates the
information actually available as displayed in the pixel plot.
Linear Transformation Iodine Data
Apical and Halo Anomalies
The Friday Field above is a good example of the problem of distinguishing
between apical (target) and halo anomalies. The northern well, the better of the
two, is clearly a halo. The southern well appears to be apical although there is
one low point at the well site. A majority of fields over the years have been apical
anomalies. Examples from the Survey section would include, Stoney Point,
Dolley, Jace, Second Wind, Nile, Plum Creek, Eland, East Boulder, Pamona,
Wehking and Veribest. A large minority of fields, however, have been halos.
Again from the Survey section this would include, Friday, Elbridge, State Line
and Weaver.
between apical (target) and halo anomalies. The northern well, the better of the
two, is clearly a halo. The southern well appears to be apical although there is
one low point at the well site. A majority of fields over the years have been apical
anomalies. Examples from the Survey section would include, Stoney Point,
Dolley, Jace, Second Wind, Nile, Plum Creek, Eland, East Boulder, Pamona,
Wehking and Veribest. A large minority of fields, however, have been halos.
Again from the Survey section this would include, Friday, Elbridge, State Line
and Weaver.
Almost every geochemical tool has exhibited apical and halo anomalies, the
mechanism responsible, however, has not yet been established. Weather you are
looking at a halo or a target when you map your data may not be knowable
without other data. Luckily almost no one is working completely in the dark.
Geological subsurface data, geophysical, magnetic or gravity surveys or a
geochemical survey over a nearby analog can all help with the interpretation of
your geochemical data.
mechanism responsible, however, has not yet been established. Weather you are
looking at a halo or a target when you map your data may not be knowable
without other data. Luckily almost no one is working completely in the dark.
Geological subsurface data, geophysical, magnetic or gravity surveys or a
geochemical survey over a nearby analog can all help with the interpretation of
your geochemical data.
What You See
The most important thing to remember is, what you see is what you get.
Geochemistry is exactly what it looks like. Keeping in mind the limitation that a
certain number of your points are unavoidably in error (see the discussion
above), your map is a "picture" of the seepage pattern of hydrocarbons in your
area.
Geochemistry is exactly what it looks like. Keeping in mind the limitation that a
certain number of your points are unavoidably in error (see the discussion
above), your map is a "picture" of the seepage pattern of hydrocarbons in your
area.
GrayStone Exploration Labs, Inc.
| How To |