Opening the Black Box of Anisotropic Seismic Processing

Morgan Brown

Tenax Geoscience, LLC

PBGS Annual Exploration Meeting

May 7, 2015

Seismic Velocity Anisotropy

• Seismic Velocity Anisotropy ”Anisotropy”

• Seismic waves travel at different speeds through a rock volume, depending on direction

• This presentation: P-waves only

Formative Anisotropy Experience

Gonna

be a

long 7

years!

“Jon, which student will work on anisotropy?”…

“What about Morgan?”…

…“I don’t know, Tariq, anisotropy is hard.”

…”No, it has to be one of the smart ones.”

Tariq Alkhalifah

Jon Claerbout

The Need For This Talk

Even after you master the mathematics of

anisotropy, can you answer these questions?

• What rocks have it? How much?

• How does it affect real data?

• Has processing handled it correctly?

• Will it change an interpretation?

http://en.wikipedia.org/wiki/Seismic_anisotropy

Anisotropy Pitfall: Lateral Shifts

Surface Location

Imaged reflector

with Anisotropic

Data

Imaged reflector

with Isotropic Data

Surface Location

Tim

e

11o dip error

Anisotropy Pitfall: Well/PSDM Mistie

Z

X

Pie

rre S

ha

le

Top Niobara from seismic

Top Niobrara from well tops

Anisotropy Pitfall: Hockey Sticks

Z

Angle

(deg)

These events

should be flat

But this event is

“unflattenable”,

due to anisotropy

Hockey sticks impair:

• Image quality

• Interpretation

• Velocity analysis

• Fracture analysis

• AVO analysis

Anisotropy Pitfall: AVO Errors

14% difference

in offset

50o

Surface LocationSurface Location

Depth

Ray traced in

anisotropic earth

Ray traced in

isotropic earth

If you convert offset to

angle for AVO analysis

without correctly handling

anisotropy, you’ll screw up

the AVO gradient.

Anisotropy Pitfall: Natural Fractures?

E

N

Azimuthal Anisotropy is measured to infer potential

natural fractures. Maps are produced and

interpreted without full appreciation of the meaning

of the measurements and their ambiguity.

Azimuthal Anisotropy isn’t simple!

Anisotropy Causes

1) http://www.fredonia.edu/department/geosciences/lash/fractures/capseal/capseal.htm

2) http://www.earth.ox.ac.uk/~oesis/micro/

3) http://www.rupestreweb.info/tacitas.html

Rock FabricShale Sandstone

Slow

Fast

~0-5% anisotropy~10-20% anisotropy

Weak layer

Thin Layering of isotropic materials

Weak layer

The aggregate is “squishy” when squeezed

vertically, but stiff when squeezed horizontally.

Thus, vertical velocity < horizontal velocity

Slow

FastWood

Wood

Sponge

Sponge1) 2)

Open fracturesx

y

Similarly, when rocks have a dominant open fracture set,

they are stiffer (faster velocity) in the fracture direction

3)

Vertical or Bedding

Plane-oriented Anisotropy

Horizontal or Azimuthal Anisotropy

Isotropic Wavefront

X

Z

Simple imaging experiment highlights basic anisotropy concepts

Fire a source in an isotropic, constant velocity medium.

The wavefront is semicircular

Elliptical VTI Wavefront

Slo

w

Fast

Elliptical Vertical Transverse Isotropy (VTI)

Summary: The simplest type of vertical anisotropy

Effect: Wavefront is stretched horizontally.

Math: One parameter, Thomsen d, describes the stretch.

Causes

Anelliptic VTI Wavefront

Slo

w

Fast

Anelliptic Vertical Transverse Isotropy (VTI)

Summary: The most common type of vertical anisotropy

Math: Two parameters, d and h, describe the stretch.

Causes

L=1

Anelliptic TTI Wavefront

Anelliptic Tilted Transverse Isotropy (TTI)

Summary: Tilted VTI (here: 30o)

Real rocks: Parameter ambiguity; Tilt/azimuth may not follow dip!

Math: 4 parameters, d, h, tilt, azimuth, describe the stretch

Causes

Anelliptic HTI Wavefront

X

Y

Anelliptic Horizontal Transverse Isotropy (HTI)

Summary: Horizontal VTI, aka “Azimuthal anisotropy”

Real rocks: 5-10x weaker than VTI; fractures must be vertical

Cause

Zero Offset Seismic ExperimentD

epth

Surface Location

Seismic energy is sent out from the surface to the buried targets. Some of the energy is reflected and

recorded by the geophone at the surface.

Zero Offset Seismic Experiment

Surface Location

Tim

eThe recorded signal is called a “seismic trace”. Notice how the

apparent location of the reflection on the zero offset seismic

panel is displaced from the actual reflection point.

Depth

Zero Offset Seismic Experiment

Surface Location

Tim

e

By taking zero offset seismic experiments across the spread, we record a multitude of seismic reflections.

The events are “stacked” on top of each other, inconsistent with the real “geology”. Migration is an

automated procedure to correctly position events which are mis-positioned due to geologic dip.

Zero Offset Migration

Surface Location

The reflection happened at an unknown location along the

semicircle. “Spray” the energy to all possible locations.

Tim

e

Zero Offset Migration

Surface Location

Tim

e

Repeat the process for many adjacent traces. I’ve made the lines transparent. When multiple lines

overlay one another, the color saturation increases. A geophysicist would interpret a reflection as

occurring at this location.

Zero Offset Migration

Surface Location

Tim

e

Migration accurately recovers the known target geology. Ah, the magic!

Zero Offset Migration (elliptical VTI data)

Surface Location

Tim

e

4o dip error In the following slides, we repeat the “migration”

exercise, using data modeled in an anisotropic

medium.

For elliptical VTI (d=0.05), we note time shifts and

mild dip changes, which worsen as dip increases

Zero Offset Migration (anelliptic VTI data)

Surface Location

Tim

e

11o dip error

3o dip error

1o dip error

Switching from elliptical VTI to anelliptic VTI

(d=0.05, h=0.1), the time shifts and dip errors

worsen.

Zero Offset Migration (TTI data #1)

Surface Location

Tim

e

19o dip error

8o dip error

3o dip error

When the medium is TTI (d=0.05, h=0.1, tilt=-30o),

the time shifts and dip errors are severe. Large

lateral reflector shifts are apparent, and even the

flat reflector is mis-positioned!

This could change an interpretation…

Zero Offset Migration (TTI data #2)

Surface Location

Tim

e

2o dip error When the tilt is rotated (d=0.05, h=0.1, tilt=30o), the

dip errors and shifts are less on the steep dips than

even for elliptical VTI! But the flat reflector is still

mispositioned

This could change an interpretation…

Isotropic versus TTI RTM

De

pth

X TTI RTMIsotropic RTM

Try as I could to beat TTI, it simply made the imaging better in

every test. This geology is tailor-made for a TTI test – how do

we build TTI models in “real” examples?

TTI – it seems to work, but what’s the workflow?

2D line shot near Rocky Flats, Colorado. Trivial h=0.05, d=0.0, tilt=41o (where beds dip, 0o elsewhere)

Anisotropy Pitfall: Well/Seismic Mistie

• VTI Vfocusing > Vvertical

• PSDM events positioned too deep

• Depth mistie controlled almost completely by d

Z

X

Pie

rre S

ha

le

Top Niobara from seismic

Top Niobrara from well tops

Anisotropy Pitfall: False Anisotropy

Time processor 1 Time processor 2

Refraction statics has a long wavelength and “datum” component. If not handled

consistently, the data may suffer a large “bulk shift”.

Here, two processed versions of the same data have a 40-60 ms time shift

Static shifts masquerade as large, shallow Thomsen d!

How to Cheat with Thomsen d

Well

Mis

tie

(%)

-10

12

3D

ep

th

X

d volume

Z

X

After isotropic PSDM with a sloppy velocity

How to Cheat with Thomsen d

Well

Mis

tie

(%)

-10

12

3D

ep

th

X

d volume

Z

X

After isotropic PSDM with a sensible velocity

Pie

rre S

hale

Thomsen d is a “magic” PSDM well tie fudge factor – it shifts PSDM events vertically. An

image may appear to tie wells perfectly, but clients should ask how the image got there.

With anisotropy, apply Occam’s razor – it is likely “smooth”, so misties should be spatially

consistent. Scott MacKay nicely describes this issue.

If Thomsen d isn’t “geologic”, then the dips between the wells are immediately suspect!

Always QC Thomsen d! Peer into the black box!

PSDM Velocity Update

X

Z

PSDM Gathers with different velocity

Offset

Z Z Z

Too Fast Correct Too Slow

Angle Gathers: PSTM Isotropic Velocity

X

Z

Y Angle

(deg)

Angle Gathers: Updated Isotropic Velocity

X

Z

Y Angle

(deg)

These (isotropic) angle gathers look pretty flat,

but let’s zoom in to the top Niobrara reflection…

Anisotropy Pitfall: Hockey Sticks

• Events should be flat with angle

• Two velocities apparently needed to flatten this gather

• The VTI h parameter controls these two velocities in a systematic way

Z

Angle

(deg)

V(q)

dominated

by d

V(q) affected

by d and e

Anelliptic VTI Velocity

qhqd

qdhqd

qdeqd

qeqqd

qeqqdq

422

0

422

0

422

0

4222

0

4222

0

2

sinsin1

sin21sin1

sinsin1

sinsin1sin1

sincossin1)(

P

P

P

P

P

V

V

V

V

VV

Thomsen parameters

In real rocks, h > 0

V(q) dominated

by d

V(q) affected

by d and e

V(q

) (f

t/se

c)

Incidence Angle, q (deg)

Isotropic

d=0.05,

h=0.1

Anellipticity parameter, hed/12d

Refraction Velocities for PSDM?

300 ft

8,700 ft

Here, we show a long offset head wave arrival.

d210 PNMO VV e 10PREFR VV

Velocities derived from reflections and refractions are contaminated by

anisotropy…in an inconsistent way!

Near-surface models must be corrected for anisotropy before being

merged with PSDM velocities

A Robust VTI PSDM Workflow

• Isotropic PSDM velocity updates (0-30o only)

• Build Thomsen d model from well/seismic misties• Ensure statistical “goodness” of misties (MacKay)

• Estimate h from large angles (30-50o)• Either multi-parameter analysis or brute-force scans

• Fix d and h, run VTI PSDM velocity updates (0-50o)

• Final mistie correction

A Right Way to do VTI PSDM

Z

Angle (deg)

Isotropic PSDM angle

gather, muted to 30 deg

incidence angle. Well top

shown in purple

Isotropic PSDM angle

gather, no mute applied.

Hockey sticks and vertical

misties readily apparent

A competent, for-purpose

VTI workflow can flatten

hockey sticks and correct

for well/seismic misties

0 15 30 45

A Wrong Way to do VTI PSDM

• The best approach is to first estimate the “near angle” isotropic velocity, measure dfrom misties, h from hockey sticks, and finally perform VTI PSDMs

• If we use all the angles in the isotropic updates, we obtain a d which is too large. We can estimate an h which helps flatten the gathers, but we are unlikely to ever get gathers as flat as in the first approach

Well top Well toph

d

Angle gather from

isotropic velocity obtained

by flattening 0-30o.

Thomsen d causes

seismic to mistie well top.

h causes hockey sticking.

Angle gather from

isotropic velocity obtained

by flattening 0-50o.

This imaging velocity is

faster than the 0-30o

velocity. It reduces the

apparent h and increases

the apparent d

Good Bad

VTI versus HTI

• The VTI effect causes:

• Depth misties between well tops and seismic

• “Hockey sticks” on migrated gathers

• Idealized HTI causes “sinusoid” vs azimuth angle

• From the best-fitting sinusoid, we can infer:

• Fracture/stress orientation

• Fracture/stress magnitude

Azimuth

Angle (deg)

Fast Slow

Incidence

Angle (deg)

NIOB

VTI HTI

OVT or Azimuth/Incidence Angle GathersD

ep

th (

ft)

East

Incidence

Angle

(deg)

Azimuth

Angle

(deg)

North

q=10o

f=

90

o

f=

-90

o

f=0o

q=10o

q=10o

q=10o

q=50o

HTI Analysis Workflow

• Stack WEM or RTM incidence/azimuth angle gathers from ~25-50 deg incidence angle to compute azimuth gathers

• For each event on each angle gather, compute best fitting sinusoid (in azimuth)

• For each sinusoid, tabulate “amplitude” of sinusoid (fracture intensity), “phase” of sinusoid (fracture orientation), and quality of fit

• Generate 3D volumes for fracture intensity, fracture orientation, and quality

• Pick horizon(s) of interest

• Flatten fracture volumes to horizon(s)

• Map fracture attributes on horizon slices

amplitude

phase

What Does HTI Really Mean?

• Perform “raytracing” in a simple 3-layer HTI model

• Background velocity = 12K fps

• Vary HTI d and orientation in each layer, analyze azimuthal traveltime variations at base of layer 3

• Shoot 45o ray up from base of layer 3

Layer 1

Layer 2

Layer 3

5000 ft

1000 ft

1000 ftAnalysis

Depth

What Does HTI Really Mean?

• ~1000 ft thick fractured layer produces measurable HTI

• Is d=0.05 too strong?

d=0.0f=N/A

d=0.0; f=N/A

d=0.05; f=45o

Layer 3

What Does HTI Really Mean?

• Detectability threshold test

• d=0.005 over the entire section is barely detectable

• Yet, typical real data measurements indicate this magnitude of d is common

d=0.005f=45o

d=0.005; f=45o

d=0.005; f=45o

Layer 1/2/3

What Does HTI Really Mean?

• Opposing orientations in layers 2/3

• HTI (Nearly) cancels out

• But notice a nonzero deviation from NMO time – will look like a pull-up!

d=0.0f=N/A

d=0.05; f=135o

d=0.05; f=45o

Layer 3

Layer 2

What Does HTI Really Mean?

• “Fracture unconformity”

• Net HTI effect relatively weak

• Apparent orientation differs from either interval orientation

d=0.01f=90o

d=0.03; f=45o

d=0.03; f=45o

Layer 2/3 Layer 1

Apparent f=65o

HTI “Fracture” Map

E

N~0.1%

~0.3%

FMI

This is a typical “unconventional resources” HTI response (DJ

Basin, Niobrara) – weak apparent fracture magnitude, in spite of

observed fractures in FMI and azimuthal anisotropy in sonic

scanners.

From the above slides, it is clear that layer stripped HTI solutions

are key. However, the odds are small that “cancelling” interval HTI

responses consistently produce the observed (lack of response).

More likely, the in situ P-wave HTI effect is small in most quiet

basins. We saw that a fairly strong 0.5%, top-to-bottom, HTI

effect barely produces an observable sinusoidal response.

HTI technology seems more applicable in more tectonically

complex geologies. Unfortunately, a dipping HTI effect becomes a

TTI effect, making parameter estimation ambiguous. Triclinic?!?

HTI – Caveat Emptor!

Conclusions

• Goals for the Audience

• Basic concepts

• Real data effects

• Contract with confidence (e.g., VTI PSDM)

• Anisotropy Pitfalls

• Positioning errors

• False anisotropy / d cheating

• AVO errors

• HTI is tricky at best

Acknowledgements

• Fidelity E&P

• Dave List, Chris Lang, Patrick Rutty

• Wave Imaging Technology/GeoCenter

• Randy Ray, Ned Sterne, Jim Applegate