American Oil and Gas Reporter - November 2016 - 46

SpecialReport: Oil & Gas Computing

FWI Key To Imaging Complex Structure
By Bing Bai,
Guo Yu
and Sunxi Wang
HOUSTON-With increased computing
power and advances in algorithm development, deriving a velocity model with
full waveform inversion (FWI) is becoming
more and more feasible. FWI generates
modeled synthetic data from an initial
velocity model, then inverts for an updated
velocity model by minimizing the difference between the synthetic and the recorded
data using a least-squares inversion method.
Unlike ray-based tomography, FWI
uses the full wave field instead of a single
ray-path travel time to derive the underlying velocity model. It can provide a
more stable solution with higher resolution
in the velocity model, especially in areas
of shallow, complex geology. However,
FWI has its limitations. Aside from its
high computational cost, an inadequate
initial velocity model can lead to cycle
skipping between the synthetic and recorded data, and can drive the least-squares
solutions into local minima.
To reduce the chance of cycle skipping,
low-frequency diving wave and refraction
energy are commonly used for FWI. As
a result, the maximum depth for FWI velocity updates is limited by the penetration
depth of diving waves, which is constrained by the maximum acquisition offset. Reflection data have better illumination
to a much greater depth than diving

waves. With a good initial velocity model,
reflection data can be used jointly with
diving waves in FWI to improve the velocity resolution and increase the maximum update depth.
A field in the East Breaks area in the
deepwater Gulf of Mexico faced seismic
imaging challenges caused by a masstransport complex (MTC) below the
seabed. The MTC is characterized by
moderate-to-strong seismic velocity and
amplitude variations caused by both structural and stratigraphic "inhomogeneities."
Traditional tomographic velocity inversion
methods could not provide sufficient resolution of the small-scale velocity anomalies within the MTC. As a result, seismic
events below the MTC appeared distorted
and out of focus.
Multiple iterations of high-frequency
FWI (up to 18 hertz) were able to derive
a detailed velocity model of the MTC,
thereby improving the seismic image beneath it. Detailed analysis at a deeper
reservoir level showed that event continuity, gather flatness, AVO response, and
prestack P-impedance inversion all were
improved after migration using the FWIderived velocity.
In this application, both diving wave
and reflection data were utilized to obtain
a velocity update in both the shallow and
the deep sections. We also tried to increase
the maximum frequency for the FWI update in order to gain additional resolution
in the velocity model. A more accurate

FIGURE 1
East Breaks Area Seismic Profile
Migrated with Tomography-Derived Velocity Model

1,600

2,000

2,400

2,800

3,200

46 THE AMERICAN OIL & GAS REPORTER

velocity model was expected, subsequently
improving the seismic image and revealing
structures at a higher resolution.
FWI Starting Point
The East Breaks dataset was acquired
by narrow-azimuth, flat-towed streamers
with a maximum offset of eight kilometers,
and the study area covered approximately
100 square kilometers where a roughly
one kilometer-thick MTC was observed
below the seabed. An MTC consists of
sediment that has been remobilized after
its original deposition. Its complex dips
and internal structures not only pose imaging difficulties for deeper events, but
may contain shallow drilling hazards.
Figure 1 shows a seismic profile from
the study area migrated with a tomography-derived velocity model. A low-amplitude background with isolated high-amplitude reflectors of various sizes and dips
was observed within the MTC. The red arrows indicate these internal structure/amplitude anomalies. The yellow arrows show
the distorted structures below the MTC
(the producing reservoir is located beneath
the MTC at a depth of 3.2 kilometers).
The blue arrow indicates a deep reservoir.
A good initial velocity model was
needed to reduce the likelihood of cycle
skipping. Two iterations of ray-based tomography were applied on an existing
tilted-transverse isotropic velocity model
as the starting point for FWI. The tomography corrected the background velocity trend and generated flat gathers in
most places. However, it failed to resolve
the small-scale velocity anomalies within
the MTC. As a result, events below the
MTC appeared distorted, namely "wavy"
in nature, and out of focus (as indicated
by the yellow arrows).
A ray-tracing analysis of diving-wave
energy showed that the maximum penetration depth of the recorded diving waves
was just above the reservoir. Reflection
signals needed to be included in the inversion in order to obtain a reasonable
update at the reservoir level. To determine
whether reflection data could help the
inversion, two sets of FWI were run up
to eight hertz with and without reflection
data, and the results were compared at
the well location.
Including the reflection data improved
the resolution of the model and generated
a more accurate update at the reservoir

American Oil and Gas Reporter - November 2016

Table of Contents for the Digital Edition of American Oil and Gas Reporter - November 2016