Processing and Interpretation Considerations for Full Waveform Inversion of PSDM
Velocity
Fields
Abstract
Abstract
Full waveform inversion (FWI) technology offers the potential reward of subsurface velocity
resolution and PSDM imaging quality never before attainable within the industry. It is performed on prestack
data
, and prestack processing and interpretive experience with modern datasets are valuable assets in conducting production FWI projects. FWI is a nonlinear inversion and, as such, does not guarantee convergence to a geologically reasonable solution. In practical applications, without continuous monitoring of intermediate
data
and model updates, the output of the inversion process can diverge and produce non-geological output models and inversion artifacts.
As a new technology, FWI is often presented in the language of mathematicians, which can preclude input by non-FWI experts. This is regrettable because theoretical knowledge is desirable, but often not as important as familiarity with data
and expertise of processing and interpretation tools. This paper describes the FWI workflow in terms to which a seismic professional can relate. Also discussed are workflow steps that, at first glance, might appear unfamiliar or disconcerting to FWI novices. These lessons learned are exercised on a synthetic marine case study, which simulates a multitude of challenges encountered during real-life FWI projects. The
velocity
field has a variety of exploration-relevant
velocity
features, including shallow channels, low- and high-
velocity
lenses, layered sequences, and near-vertical
velocity
discontinuities. In the discussed simulated “blind” test, it was determined that detailed understanding of the input
data
set,
velocity
estimation techniques, and imaging artifacts are paramount to a successful inversion project. The FWI case study began by first evaluating three different input
velocity
models suitable for FWI. Next, prestack
data
was prepared to generate differential
data
to be carefully analyzed for optimal processing parameters using a new condensed differential
data
QC technique. FWI gradients were conditioned based on geologic considerations and an understanding of imaging artifacts. Offset ranges, mutes, and frequency ranges were carefully adjusted based on condensed differential
data
analysis
. After 59 increasingly expensive FWI iterations, a good correlation of
velocity
to geology and a crisp image with clear fault patterns were obtained.
AAPG Datapages/Search and Discovery Article #90260 © 2016 AAPG/SEG International Conference & Exhibition, Cancun, Mexico, September 6-9, 2016