The following abstract
has been submitted to the SEG for a talk at the 2009 convention. If
accepted, you can hear the whole talk/presentation there:
Using Horizon Binning and
2D GrAZ for horizon based fluid contact detection and anomaly
Horizon Binning and
2D GrAZ are horizon based methods for assessing the quality
of an HCI and for determining the location of a fluid contact, if any.
Horizon Binning is a technique for accumulating statistics as a function of
structure, beginning in the water leg of a prospective sand and working “up
structure.” A fluid contact is inferred when at a given time (or depth)
there is a consistent change from a water leg response to a “not-water”
response. This change should be approximately a step function. An
attribute that can be used with Horizon Binning to further enhance a
hydrocarbon effect is 2D GrAZ.
2D GrAZ is a horizon attribute computed as a dot product of
the gradient of the structure with the gradient of an attribute (e.g.,
amplitude). After horizon binning, a 2D GrAZ anomaly will appear as a peak, above background, at
the time (depth) of the fluid contact. In the water leg the 2D GrAZ signal should be near zero, indicating isotropy of
the seismic response with respect to structure.
After identifying an amplitude, AVO or other attribute anomaly, it is normal
to ask the question, “Does it fit structure?” Many anomalies can be
judged to “fit structure,” by inspection. In other cases, the fit might
be more tenuous, more debatable or simple corrupted by noise. In the less
straight-forward cases, it is helpful to have a set of tools to make simple
The purpose of both Horizon
Binning and 2D GrAZ are to assist in measurements around the question,
“Does it fit structure?” In both cases the inference is made by comparing
a down-dip water leg response to a response either up-dip or at the fluid
contact. These methods might best be thought of water-leg analysis tools,
with the existence hydrocarbons being inferred only at the confidence level of
asserting a “not water-leg” response.
Horizon Binning calculation
Horizon Binning is a
method of accumulating statistics as a function of structure, either time or
depth (the “Z” value). This is a horizon based calculation that requires
a fully interpreted horizon that contains both Z values and an extracted
horizon based attribute. Typical horizon based attributes might be the
maximum instantaneous amplitude near the picked event, RMS over a short window
or simply maximum negative trough amplitude. In the area of an anomaly of
interest, an enclosing polygon must be selected that encompasses both the
anomaly and a corresponding down-dip water-leg region. A sequence of bins
is constructed that covers the Z-range of the data. Within each Z-bin,
corresponding attribute values are collected. For each bin, the mean and
standard deviation are computed and then graphed (see Figure 1). In the
graph of Figure 1, down-structure is to the right (increasing Z). The
transition from the water-leg to the hydrocarbon-leg is seen as a step
function, in this example plot.
Figure 1: In a
Horizon Binning plot, the X-axis is time or depth and increases to the
right. The vertical axis is the horizon-attribute.
The mean is plotted
as a solid black line. The +/- 1 standard
deviations are plotted as error-bars with the dotted red lines.
2D GrAZ calculation
2D GrAZ is a method of
determining whether changes in the derived horizon attribute change abruptly in
the up-dip direction, at a specific Z value and, hopefully, due to a fluid
contact. To make this determination, the gradients of both the
interpreted Z values and of the extracted attributes are computed in 2D.
This produces 2 gradient vectors at each horizon grid location. A dot
product is formed using both vectors. The dot product can be put through a
Horizon Binning process to make a graph like the one seen in Figure 2, or can
be mapped directly, as shown in the example of Figure 3.
2D GrAZ relies on 2 assumptions. First, changes in the
water-leg response must be isotropic in map view, with respect to structural
dip. This means that locally the dot product with a structural gradient
vector should average to zero. Second, the assumption is made that a
change in the up-dip direction that occurs at a specific time or depth is
associated with a flat event, e.g. a fluid contact. This is the origin of
the peak of Figure
Figure 2: In an 2D GrAZ
plot after Horizon Binning, the X-axis is time or depth and increases to the
right. The vertical axis is the binned dot product of the gradient of the structure map
with the gradient of the seismic attribute. In the water-leg, isotropy
is inferred by a very small dot-product values to the right, down structure..
Figure 3: A 2D GrAZ dot product can be mapped. Shown is a 2D GrAZ calculation with contours over-lain. The location
of the anomaly (roughly north-sourth) corresponds
with an inferred fluid contact. Down-dip is to the north-east.
Note that the anomaly appears to break at likely fault locations.
size of the polygon area
Both Horizon Binning
and 2D GrAZ require enough points over a long enough Z-range to
accumulate statistics on a bin-by-bin basis. For instance, for 12 ms bins
and with at least 10 bins in the water-leg and 10 bins in the hydrocarbon-leg,
the mapped horizon must span 240 ms. Within each bin, one would prefer at
least 35 data points, for decent statistics. If the 3D bins are 110 ft x
110 ft, this means 9.7 ac per bin or about 100 acres for both the water-leg and
for the hydrocarbon-leg. The need for sufficient bins for statistics over
a Z range limits the utility of these tools for studying channel sands, low
relief structures and smaller fault-blocks.
Horizon Binning and
2D GrAZ are simple measurement techniques aimed at the question
of “Does the anomaly fit structure?” The calculations are simple, fast and
intellectually trivia. They do, however, introduce elements of rigor and
reproducibility in the anomaly analysis process. These techniques are not
“silver bullets,” but can logically augment a normal exploration and/or
development work-flow in many settings around the world.
The author would
like to thank OpenSpirit Corporation for permission to use the data analyzed in