American Oil and Gas Reporter - July 2019 - 72

SpecialReport: Reservoir Characterization
FIGURE 2
Time-Domain and Phase-Domain Quality Control of Full-Azimuth Data

azimuth streamer dataset in the Keathley
Canyon area, the new algorithm is able
to update salt geometry to significantly
improve TOS/BOS and subsalt images.
The data was acquired using multiple
vessels in a staggered configuration.
Panel A in Figure 1 shows bathymetry
in the Keathley Canyon study area, while
Panel B shows the staggered receiver
acquisition geometry. The five colored
circles represent the source boats, with
blue lines denoting the leading receiver
boat and red lines denoting the tailing
receiver boat. The towed streamers had
variable depths ranging between 10 and
50 meters. The rose diagram in Panel C
shows that this acquisition configuration
provided full-azimuth coverage up to
10 kilometers and ultralong offsets up
to 18 kilometers (the inner circle marks
10 kilometer offset and the outer circle
denotes 18 kilometer offset). The color
bar represents the fold.
Figure 2 displays the time-domain and
phase-domain quality control of the data.
There are apparent signals down to 2.5
Hertz, thanks to the deep towing depth
of the streamers down to 50 meters, compared with ~4.0 Hz for a typical shallow
towed streamer dataset. The time-domain
QC at 2.5, 3.5 and 4.5 Hz are for one
shot gather from gun No. 1 (dark blue
circle in Panel B in Figure 1) with receivers
on one cable from the leading boat (blue
lines in Figure 1) and another cable from
the tailing boat (red lines in Figure 1).
The phase-domain QC is for the same
gun-cable pair at the same frequencies
with all the shots from one sequence.
Since time-lag FWI is still mostly
driven by diving-wave energy, the testing
area was deliberately identified with a
salt body at a shallow depth of four-five
kilometers, and therefore well sampled
by the diving waves of the dataset. The
initial velocity model for the FWI workflow is a smoothed version of the legacy
72 THE AMERICAN OIL & GAS REPORTER

model that was obtained after several iterations of ray-based reflection tomography and the previous diving-wave FWI
for the overburden, followed by typical
salt interpretation and subsalt velocity
updates.
Field Data Applications
Similar to most other FWI approaches,
the first step was determining the lowest
starting frequency. Based on the QC in
Figure 2 and past experience with previous
FWI approaches, we typically would
choose a starting frequency of 3.5 Hz
since the signal-to-noise ratio at 2.5 Hz
appears very low. Surprisingly, better results were obtained with the new FWI
algorithm when using a starting frequency
of 2.5 instead of 3.5 Hz. This indicates
that time-lag FWI has good tolerance for
low-frequency noise because of the frequency-dependent, time window traveltime cost function and the cross-correlation
coefficient-based weight function.
Panels A and B in Figure 3 show an
inline section with velocity overlaid on
the reverse-time migration (RTM) stack

image for the initial model (smoothed
version of the legacy model) and the
time-lag FWI-updated model (straight
output with minimal editing to mute updates below six kilometers that are not
well sampled by diving-wave energy),
respectively. The white circles and arrows
mark where the salt bodies were appreciably updated by time-lag FWI.
As expected, the velocity update from
the algorithm is very small in the shallow
sediment. There are some detailed changes
around the TOS, especially at the salt
finger marked by the white circle, and
the salt geometry is slightly reshaped.
The salt geometry changes were more
noticeable on the depth slice at about
three kilometers (panels C and D). Although slightly sharper than in the initial
model, the salt boundary after time-lag
FWI is smooth overall compared with
the typical salt models built with standard
velocity model building. This is partly
because the initial model had smooth salt
to begin with and FWI was run up only
to 6.3 Hz. Sharper salt boundaries are
expected at higher FWI frequencies.
Figure 4 shows an inline section of
the RTM stack image comparison between
the initial model (panels A and C) and
the time-lag FWI-updated model (panels
B and D). The first two panels compare
the models in a shallow window from
two-to-five kilometers. The small salt finger is poorly imaged with the initial
model, but well imaged with a clearly
defined overhang boundary after FWI
(yellow dashed lines). Both the TOS and
BOS are also better imaged using the
FWI model (yellow arrows).
The last two panels in Figure 4 compare

FIGURE 3
Inline Section/Depth Slice with Velocity Overlaid on RTM Stack
For Initial Model and FWI Straight Output Model
A

American Oil and Gas Reporter - July 2019

Table of Contents for the Digital Edition of American Oil and Gas Reporter - July 2019