0km 17

1
Next Generation Numerical Modeling and
Imaging Marmousi-2, an elastic upgrade for AVO
calibration Gary Martin (University of
Houston/GX Technology), Kurt Marfurt (Allied
Geophysical Laboratory, UH)
Motivation and objective
Elastic Data Acquisition
Modifying the Marmousi model

• Data will be acquired using the E3D elastic
  modeling code at Lawrence Livermore National
  Laboratory.
• Computation is extremely intensive and will
  utilize massively parallel systems.
• Differences between the acquisition of the
  original Marmousi model and Marmousi-2 are
  outlined in the table below

• The original model has been recreated as a GXII
  layered model from interfaces extracted from the
  property grids.
• The original P-wave velocity definitions have
  been recreated and adjusted for the increased
  water depth.
• The model has been expanded to almost twice the
  width.
• Lithology has been assigned to each layer.
• Layer properties based upon original P-wave
  velocity and velocity transforms have been
  introduced.
• Additional hydrocarbon charged layers added.
• Additional stratigraphic features added.

• Synthetic seismic data are used by the
  exploration industry to aid the understanding of
  complex phenomena and to test and calibrate
  seismic processing algorithms.
• The original Marmousi data set has been the
  standard 2D imaging test. It was created in 1988
  at great computational cost, using acoustic
  methods. Advances in computational speed (and
  reduction of expense) have made it possible to
  perform elastic modeling over large 2D lines.
• Elastic model data are required over complex
  structures that contain hydrocarbons with AVO
  effects, and subtle stratigraphic targets, for
  the testing of modern processing algorithms.
• The model and dataset will be made publicly
  available for use in testing and calibrating
  algorithms. The data will be useful for velocity
  estimation, migration algorithm evaluation,
  multi-component and VSP processing evaluation,
  AVO calibration, evaluation of amplitude
  preservation during migration, multiple
  suppression, etc.

Original Marmousi
Marmousi-2
Original Marmousi Marmousi2
0 (km)
17
The Marmousi-2 Model
0 (km)
3.5
Transition layers at waterbottom
Water
Faults
Shales
Water wet sand
Water wet sand channels
Gas charged sand channel
Gas sand
Shales
Zero P-wave impedance sand

• 196 layers
• 3 large faults
• 1 shallow gas sand in simple area
• 1 relatively shallow oil sand in simple area
• 4 fault trap gas sands
• 3 fault trap oil sands
• Oil and gas trap on anticline
• Zero P-wave impedance layer stratigraphic feature

Oil charged sand channel
Oil sand
Unconformity
Marl
Shales
Salt
Gas and oil cap
Water wet sand
Hydrocarbons
Physical Properties
Preliminary Results

• 10 hydrocarbon charged layers created in
  structurally complex and structurally simple
  areas (see section above)
• Fluid substitution provides realistic properties,
  (useful for AVO calibration)
• Gas, heavy oil, and normal oil introduced
• Seismic responses verified using simplified
  models, ray-tracing, and wave equation modeling

• Elastic modeling requires three physical
  properties, compressional wave velocity (Vp),
  shear wave velocity (Vs), and density (r).
• Properties are constant in layer, or have
  vertical gradients within the layer.
• Vp from the original Marmousi model except for
• hydrocarbon charged layers,
• salt (given a constant velocity of 4500m/s)
• Vs from Greenburg and Castagnas Vp-Vs
  transforms (m/s)
• Sand Vs 0.804Vp 856
• Shale Vs 0.770Vp 867
• Limestone Vs 1.017Vp 0.055Vp2 1030
• Density from Castagnas Vp-r transforms (m/s and
  g/cm3)
• Sand r 0.2736Vp.261
• Shale r 0.2806Vp.265
• Limestone r 0.3170Vp.225
• Hills mixing laws used to calculate properties
  for the marl (30 limestone, 70 shale)

• Some test shots have been acquired using both
  acoustic and elastic methods.
• Snapshots show the complexity of wave propagation
  in the structurally complex areas.
• Multiple components and VSPs can be recorded.

Streamer pressure Waterbottom pressure