1:00 PM 7/12/2005

Biondi - Overview of the project

Dave Hale logistical support

Next generation 3D modeling
Window of opportunity for cooperative project
2.5D desired in Houston meetings
3D two-way acoustic main initial 3D focus, elastic a few years out
model design will be elastic, modeling not necessarily elastic.

Geological model working group task:

design 3D elastic model, with Vp rho for acoustic work
containing challenging and realistic structures (salt bodies, faults) 
velocity variations (shallow gas, low vel under salt, overpressure) to 
test the limits of present and future 3D seismic (acquisition and 
imaging) technologies
contains 2-3 reservoirs with complex and realistic architecture and that 
produce interesting "AVO" effects when computed in acoustic sense.
Is ready for fluid-flow simulation and computation of elastic/acoustic 
time-lapse surveys.

Identify in the 3D Elastic model one or two 2-D slices:
suitable for elastic 2.5D (single line of 3D point source elastic modeling)
suitable for non-seismic modeling/inversion, in particular for: gravity, 
active-source EM, MT

Stork: what about anisotropy? (acoustic)
Bednar: do aniso in 2.5D elastic (acoustic to elastic is 13x storage, 
compute 100x)
Stork: isotropic acoustic to elliptical aniso is cheap
Stefani: tensor rotation fills in to full 21 coefficients in presence of dip
Perhaps elliptical can provide some insight, however.
Biondi: VSP will have anisotropy
Bednar: Anderson says once you define horizons, structure, you can 
populate with more complex parameters.
Abriel: let's keep anisotropy on the issue table as part of the evolving 
project timeline
Stork: Why 2 years to get acoustic?  Why not shorter.
Bednar: model design tends to take a lot of time.
Biondi: there's a tradeoff between cost of computational resources and time
??: GOM model would be quickest to design.  World-wide expansion adds 
design time
Bednar: Input has been across the geologic spectrum

Determine parameters of 3D Elastic model that define a realistic 
computational task and capture important features of the model:
- determine maximum frequency needed for resolution
-  " shot aperture and recording time to capture all dips, including 
overturned events(?)
-  " spatial sampling
Define model representations, software tools, and appropriate interfaces

Calandra: 10-15 km deep structure controls are useful, but less 
important for data processing
Stork: imaging deep structural controls can be dicey, worth 
investigating.  let's do it right.
Calandra: changing frequency content in deeper modeling is not easy
Abriel: This is another issue for the modeling groups to deal with


SEG Convention workshop
- conceptual design of 3-D elastic model
- cross-validation of numerical code against Delft ZMAART physical model
- organizational structure and legal framework
- detailed plans for computations grid vendors, spare time on company 
clusters)
- initial plan for archival and distribution (including data compression)
- involved sister societies (SPE, AAPG?) and non-seismic SEG members
- Firm commitments by "founding members"

Objectives for today:
- agree on main elements of 3D elastic model (salt typology, reservoir 
stratigraphy, velocity anomalies)
- agree on main contributors to each of these elements
- agree on model representation and software tools
- agree on time-table between now and SEG
begin formation of numerical modeling working group
make progress towards finalizing structure of consortium
discuss nominations for project manager

1:43 PM 7/12/2005

Bednar: Planning Issues

nitty-gritty

the original idea

model design working group (good company base, lots of friction, however)
   numerical modeling and software working group (good base)
      model execution (contract?) (lots of issues)
         data distribution and archive (lots of issues)
            analysis working group

domain decomposition, large runtimes, data compression, 
OS/Hardware/Software compatibility
5-10 terabytes compressed (10:1), software


the emerging idea


planning issues


BP not (yet) on board.  they have a lot of internal strength
Abriel: Chevron compression code
   international geophysical archive initiative that SEG wants to get 
involved in

Stork: If ready to start distribution data in 2 years, the storage 
capacity of disk will be 10x.
bednarb@3dbee.com


2:00 PM 7/12/2005
Abriel: SEG Model Consortium

how do we thing it is to operate?
how does your company view the structural setup

umbrella is SEG operator administrator and executive committee 
supervising coordinating project manager

looking at about 75% of budget spent on execution (computation)

2/3rd majority agreements from participants (sole voting members)

advising participants: SEG representative, SEG research chair, EAGE advisor

Questions about when 2 yr proprietary clock should start, late participation

Academic groups & non-profit groups may make 1 page proposals to 
executive committee for projects and get license for data

Question about publication of sections versus release of data

Contract doesn't exist yet, but likely to be modeled on SMAART or 
DeepStar consortia

2:38 PM 7/12/2005

Joe Stefani

ZMAART recap

ZIGGY (SMAART II) coarse, complex structure macro stratigraphy
CSIRO does make thin-layered physical models

Too many features crammed into the model make it expensive and unwieldy

U of H has 2D extended Marmousi it may be willing to donate to the 
repository

2:56 PM 7/12/2005

Break

3:16 PM 7/12/2005

Graham Winbow, ExxonMobil (standing in for John Anderson)

Computations will take a long time
Current task is to define a suitable model

quick recap of SEG-EAGE salt model

elastic equivalent is about 144 times more compute.  Shear wave --> 2^4 
more expensive
3 component --> 3^2 more computations
data volume grows by factor of 3, more if sample rate is smaller

Desirable features:

1) smooth and rugose (both deep notches and chirp signal) top of salt
2) shallow salt with impedance match to give large P-S conversions
3) deeper salt matched P-P
4) multiple salt bodies
5) realistic compaction model for subsalt
6) overpressure
7) AVO
8) reservoir zones with compartmentalization
9) variations in the base of salt
10) subsalt sediments that truncate steeply against the BOS
11) realistic salt tectonics, including faulting and structures in the 
sediments corresponding to deep salt withdrawal and slip interfaces
12) deep carbonate with rift faults near bottom of section
13) components for calibrating images, e.g. deep horizontal reflector, 
isolated pt diffractors

pictures: allochthonous salt, salt welds

Realistic Plan

design elastic model, calculate acoustic response, then elastic 
multicomp, anisotropic elastic
subset collection
surface, OBC, VSP
both absorbing and reflecting surface bdy conds
zones of very dense sampling

Could equivalent physical acoustic (elastic!) models be done?

Stefani: 2D for stacked salt body study first...

Winbow: it's easy to make models so difficult they are impossible to image.

Further support from participants to ensure model design is realistic in 
terms of viable exploration targets.

Sub-basalt as part of a staged series of models, c.f. First Break July 2005.

Winbow: use budget to scope resources, size modeling to those resources.

Stork: Original SEG-EAGE salt model would cost about $50K these days

Possible nominees:  Bee Bednar, Bob Wiley (retired or between jobs is 
almost a necessity)

Bee Bednar overwhelmingly supported.  Looking for fallback candidate 
just in case.

4:01 PM 7/12/2005

Stefani

Notes and Opinions of acoustic and elastic structure and stratigraphy 
earth model and seismic modeling requirements and tradeoffs

list of notes of seismic requirements (minimums)

fully imaged geology:
elastic 5KM on a side (1 OCS block)
acoustic 10KM on a side

aperture: mild stratigraphic 3KM, complex salt 9KM

streamer: 6 to 8KM
total size:

strat: 15KMx11KM; 4KM deep
struct: 30KMx22KM; 8KM deep

frequency: 80Hz strat, 50Hz struct

cell size: strat 4M, struct 8M

nnodes: 4000x3000x1000 ~ 12 billion nodes

total runtime memory: strat ~ 400+Gb, struct ~200 Gb

Fits in 100 node cluster with 4Gb/node

double all frequencies: 1000 node 64-bit cluster

slide on tradeoffs between 1-way and 2-way, acoustic and elastic, 2D 
2.5D and 3D

stratigraphy folks want facies shapes, 2.5D does not really help

4:19 PM 7/12/2005

Stefani:  Chevron Perspective

1) given past SEG emphasis on "geometric" (container) imaging of 
structurally complex models with only weakly represented stratigraphy and
2) the growing need for better amplitude processing and seismic 
reservoir characterization,

we believe the SEG effort is worthwhile, and we particularly support a 
strat-flavored earth/seismic modeling exercise.

likely to need elastic modeling, and certain shortcuts & compromises 
might be necessary, depending on model details and required accuracy.

Q? can acoustic simulations provide enough value for strat objectives? 
(lose Vs effects on AVA, maintain strat scattering, ...) 3D vs. 2.5D.

4:23 PM 7/12/2005

Recipe for realistic seismic earth models: evolution of Earth/Strat Models

[talks too fast to keep up w/o shorthand ...]

Remark that 90% of sub-wavelength random features are lost at seismic 
scale, i.e. an infinite number of sub-wavelength models create basically 
the same data.

Biondi & Hale: higher frequency limited aperture recording over a 
shallow stratigraphic target and lower frequency over salt.

Amerada Hess, ExxonMobil, ConocoPhillips, Total?, Chevron on Geological Modeling Committee

Request for good ballpark on computational cost/time formulas

5:36 PM 7/12/2005

Adjourn