Page 6 - i1052-5173-27-2

was converted to depth using a simple
velocity model in MoveTM, described by
Equation (1):
(1)
where Z is the depth in meters, V0 is the
initial velocity (1500 ms−1), t is one-way
travel time, and k is the rate of change in
velocity with increasing depth (0.5). The
depth conversion located the bottom of the
section at 10.5 km depth. The depth con-
version was completed on a bitmap of the
seismic reflection image, which linearly
stretches the image. The result is a depth
section with apparently lower reflectivity
and contrast than the original TWT image
and 18% longer, due to this stretching.
With the exception of depth conversion,
both the TWT and depth images share
identical processing workflows. The actual
depth conversion method used is not
important for our experiment; it is the dif-
ference in image quality the process cre-
ates that concerns us.
IMAGE ANALYSIS
The image analysis undertaken focused
on the pixel intensity contrast and reflec-
tion continuity (referred to hereafter as
“contrast” and “continuity,” respectively)
of the TWT and depth sections (see Fig. S1
in GSA’s Supplemental Data Repository1).
For the image analysis, each seismic image
was subdivided into cells of 7.2 km
(length) × 1 km (depth) (1135 × 450 pixels).
The area encompassing the participants’
fault interpretations was subdivided into
smaller cells, 1.6 km × 0.4 km (216 × 163
pixels), in order to provide detailed image
analysis information in the area of interest. the analysis. Of these elements, 116 corre-
For ease of comparison of our results, the
seismic images are both shown with a ver-
tical scale in depth in all figures (except
Fig. 1A).
To analyze the image contrast we
extracted grayscale distributions for the
pixels in each cell for the two uninter-
preted images. The distributions range
from pixel number 0 (black) to 255 (white):
the wider these distributions are (i.e., the
more pixel values close to the extremes of
0 and 255), the more contrast the image
has; the narrower the pixel distribution, the in the TWT section (5 faults). The effect of
more similar the pixel values are and thus
1 GSA Supplemental Data Repository Item 2017031, image analysis methods, is online at http://www.geosociety.org/datarepository/2017/. If you have questions,
please email gsatoday@geosociety.org.
6 GSA Today | February 2017
the lower the contrast. The first and third
quartiles (Q1 and Q3) from these distribu-
tions were subtracted in order to calcu-
late the interquartile range (IQ) of the dis-
tributions. We use the interquartile range
as an analogue for visual contrast: the
wider the IQ of the cell, the higher the
contrast and vice versa. Each cell in
the images is colored according to its IQ
value in order to display graphically the
contrast analysis results.
To analyze the reflector continuity,
the images were first converted into a
binary, i.e., a black and white image. This
was performed using ImageJ software
(Schneider et al., 2012) by setting an auto-
matic threshold level based on the histo-
grams of the two images. This threshold
divides the pixel histogram in two halves,
assigning black or white color to all the
pixels. As a result, the seismic wave
reflections are separated into isolated
black bodies, corresponding to the posi-
tive amplitude reflections in this particu-
lar case, included in a white background.
A macro for the software ImageJ
(Heilbronner and Barrett, 2013) was used
to measure and analyze these resultant
bodies. In the analysis, the length of the
major axis of each reflection is calculated,
using a best-fit ellipse method, and each
reflection is then colored based on this
length value using a color scale.
INTERPRETATION OUTCOMES
Interpretations of the major discontinu-
ity of reflectors (faults) located in the mid- quartile follows a similar path to that of the
dle of the seismic images and related splay
faults (327 elements in total) were used in
spond to the interpretations of the TWT
image (Fig. 1C) and 211 to the interpreta-
tions of the depth seismic image (Fig. 1D).
In general, variability in fault placement
position (the spread in fault interpretations) variability and heterogeneity. The fault
increases with depth, and this observation
is more pronounced in those interpreta-
tions derived from the depth image. The
difference in fault placement spread
between the two images is at a maximum
at 5 km depth. Below this point, the amount
of interpreted faults dipping to the right is
greater in the depth section (23 faults) than
the difference in the populations of TWT
and depth interpretations was analyzed by
randomly selecting 70 of the depth inter-
pretations for comparison with the TWT
interpretation population of 70. Because
these were found to be similar to the full-
depth interpretation analysis, we conclude
that the population size had no effect on
the results.
Quantification of the variability in fault
placement for the interpretation popula-
tions were computed at nine depth markers
in each seismic image (Fig. 2). The four
quartile and outlier positions for the fault
interpretation populations were calculated
at each depth marker (results of the analy-
sis are shown in Fig. 2, overlying the
image analyses). The interquartile fault
range (the distance between the first and
third quartiles) provides a good estimation
of the fault placement spread within each
of the interpretation populations at a given
depth (continuous black lines in Fig. 2,
created from joining the quartiles between
depth markers). We use the interquartile
range of fault placement within each fault
interpretation population as an indicator of
fault placement uncertainty for each seis-
mic image. The interquartiles show that
fault spread remains similar in the upper
3.5 km. From 3.5 km downward, the inter-
quartile fault range in the depth image
increases until, at the base of the seismic
image, the interquartile width is twice that
observed in the TWT image. The increase
in fault spread defined by the interquartile
trend linearly increases in the TWT image
with depth. In the depth image, the first
TWT image, but the third quartile is more
heterogeneous (wavy) and is offset to the
right with respect to the third quartile line
in the TWT image. Meanwhile, the outliers
(dashed black lines in Fig. 2) show a simi-
lar general pattern with fault spread
increasing with depth, but with a greater
placement outliers for the fault interpreta-
tions of the TWT image show a convergent
trend down to 2 km in width at ~4 km
depth before the fault placement spread
increases to ~15 km at the base of the
image. The fault placement outliers from
the depth interpretation show a relatively
constant spread (~4.5 km width) down to
3 km depth. Below this point, fault spread