Vous êtes sur la page 1sur 36

Seismic inversion

Seismic inversion course, UPPA, 2008


Course Outline
General aspects of seismic inversion
ƒ Concept and purpose
ƒ Petrophysical basis
ƒ Data requirements
ƒ Overview of methods and basic inversion process

Poststack inversion – practical workflow


ƒ Data QC
ƒ Wavelet extraction
ƒ A priori model building
ƒ Inversion parameterization and QC of results

Prestack inversion – practical workflow


ƒ Data QC
ƒ Wavelet extraction
ƒ A priori model building
ƒ Inversion parameterization and QC of results

Advanced inversion techniques


ƒ Geostatistical inversion
ƒ Joint PP/PS inversion
ƒ 4D inversion

2 - Seismic inversion course, UPPA, 2008


Various inversion methods

Sparse Spike: algorithm to find an impedance model matching the seismic bandwith,
but with the smallest number of non zero reflectivities (limited number of
homogeneous layers)

Model-based: uses an initial low-frequency impedance model which is perturbed

Layer based: similar to model based, except that the model is layered and inverted
both for thicknesses of layers and impedances

Stochastic: multiple realizations of impedance models that can reproduce the seismic
amplitudes within a given error

These four can be post-stack (inversion of a full stack) or pre-stack (inversion of several
sub-stacks)

3 - Seismic inversion course, UPPA, 2008


Petrophysics in shallow siliciclastics

Discrimination between the


various lithologies/fluid
content requires two
petroelastic parameters:
P impedance and Poisson’s
ratio
Water
2
1 ⎛ Vp ⎞
⎜ ⎟ −1
2 ⎝ Vs ⎠
Shales ν =
Sands
2
⎛ Vp ⎞
⎟ −1
Poisson’s ratio


⎝ Vs ⎠
Brine
HC
Poisson’s ratio or S
Sandstones
impedance can be obtained
Class III II I by inverting pre-stack data
P-wave Impedance at several angles of
Sa ist

incidence (sub-stacks):
nd og
h

/ s ram

Pre-stack inversion
ha s
le

4 - Seismic inversion course, UPPA, 2008


Prestack inversion workflow

Data QC
ƒ Substack definition
ƒ QC of homogeneity between substacks
ƒ Time alignment between substacks

Seismic to well tie and wavelet extraction


ƒ For each angle stack

A-priori model building


ƒ 3 parameters: Ip, Is, Rho

Inversion parameterization

Interpretation of the inversion results

5 - Seismic inversion course, UPPA, 2008


Required input data

Well logs: Vp, Vs, Rhob, Phie, caliper, GR, ….

Check-shots, VSP or T=f(depth)

Seismic data: angle stacks and full stack


ƒ Minimum of 3 angle stacks, more are better
ƒ Depends on angle range and fold of the data

Stacking or migration velocities

Smoothed seismic interpretation (horizons, faults)

Notions of depositional mode within units

Quality control of all input data is


paramount!

6 - Seismic inversion course, UPPA, 2008


Simultaneous inversion of angle stacks
Seismic 0 - α1 Seismic α2 - α3 Seismic α1 - α2

Simultaneous inversion
of substacks to derive
Ip, Is and Rho volumes

Other elastic
Wavelet Wavelet Wavelet parameters can be
derived from the
AI Logs EI Logs EI Logs
outputs
ƒ Vp/Vs, Poisson’s ratio
Logs

Constraints Subsequent lithology


and Low
Frequency and fluid classification

Horizons
P-Impedance
S_Impedance
Density

Vp/Vs MuRho
LambdaRho

7 - Seismic inversion course, UPPA, 2008


Prestack inversion

Data QC
Wavelet extraction
A priori model building
Inversion parameterization and QC of
results

8 - Seismic inversion course, UPPA, 2008


Angle stack definition
Minimum Offset Maximum Offset
Available 150 m Available 3050 m degrees
degrees

Θ min@ 150 m-1900ms(x,y)

Scale step 0.5deg.


Scale step 0.02deg.

ttop 1900 ms

Θ max@ 3050m-2300ms(x,y)

tbase 2300 ms

Isolines every 0.1deg. Isolines every 4deg.


Maps of Min/Max angles reached @Min/Max offsets at Top and Base of Target interval

9 - Seismic inversion course, UPPA, 2008


Preliminary QC’s of seismic substacks

Homogeneity of frequency spectra


NEAR OFFSET TRACES SUB-STACK FAR OFFSET TRACES SUB-STACK

NOISE SIGNAL NOISE SIGNAL

AUTOCORRELATION AUTOCORRELATION

CROSS-CORRELATION BETWEEN TRACE N AND CROSS-CORRELATION BETWEEN TRACE N AND


TRACE N+1 TRACE N+1

NEAR TRACES
POWER SPECTRUM

FAR TRACES
POWER SPECTRUM

10 - Seismic inversion course, UPPA, 2008


Preliminary QC’s of seismic substacks

Resolution near around 13ms Average envelope correlation near/mid : 72%

Well 2 Well 2

Well 1 Well 1

Resolution map Envelope correlation map


Average phase shift: 0° Average time shift : 0ms

Well 2 Well 2

Well 1 Well 1

Phase shift map Time shift map

11 - Seismic inversion course, UPPA, 2008


NMO correction quality
Correct velocity picking is necessary!
Alignment between substacks is often required

Under-correction Over-correction

12 - Seismic inversion course, UPPA, 2008


Time alignment between sub-stacks

Mandatory: alignment of events in prestack gathers is never perfect


over an entire seismic volume

Based on cross-correlation around key horizons

Stretch & squeeze of the seismic can be performed

Smoothing of the time shift cubes is required


ƒ Absolute values should remain small

13 - Seismic inversion course, UPPA, 2008


Time alignment between substacks

14 - Seismic inversion course, UPPA, 2008


Cross correlation between the stacks before and after time
alignment.

15 - Seismic inversion course, UPPA, 2008


Prestack inversion

Data QC
Wavelet extraction
A priori model building
Inversion parameterization and QC of
results

16 - Seismic inversion course, UPPA, 2008


Well tie for each angle stack

Near substack well tie


Far substack well tie

Mid substack well tie

17 - Seismic inversion course, UPPA, 2008


Wavelet for each angle stack
Near wavelet

Mid wavelet

Far wavelet

Consistency of
wavelets across
substacks

No physical reason
for the phase to
change between
substacks

18 - Seismic inversion course, UPPA, 2008


Prestack inversion

Data QC
Wavelet extraction
A priori model building
Inversion parameterization and QC of
results

19 - Seismic inversion course, UPPA, 2008


Low Frequency (a-priori) model building
Objectives
ƒ Filling in the frequency part absent from the seismic data (0 to 10Hz)
ƒ Initial solution for some algorithms

Uses well data, horizons, seismic velocities, geology

Strong impact: Absolute impedance = Relative impedance (from


sesimic) + LF model

LF model

Well HR Seismic
Velocities Interp. inversion

20 - Seismic inversion course, UPPA, 2008 3Hz 10Hz


Frequency
Relationships between model parameters

3 low-frequency models needed: IP, IS or PR, RHO

When seismic velocities are used in the a-priori model building,


correlations have to exist between the P-wave velocities and the other
two parameters

Vp
Vp

Vp
P Impedance Poisson ratio Density

21 - Seismic inversion course, UPPA, 2008


Final low frequency models
Ip low frequency model PR low frequency model

Density low frequency model

22 - Seismic inversion course, UPPA, 2008


Prestack inversion

Data QC
Wavelet extraction
A priori model building
Inversion parameterization and QC of
results

23 - Seismic inversion course, UPPA, 2008


P impedance QC plot IP
IP Inv. Trace 5 seismic
Inversion Log Near Input
Near Input + SS LF IP model Section LF trace Residual traces

24 - Seismic inversion course, UPPA, 2008


Poisson ratio QC plot PR
PR Inv. Trace 5 seismic
Inversion Log Far Input
Far Input + SS LF PR mode section LF trace Residual traces

25 - Seismic inversion course, UPPA, 2008


Pre-stack inversion QC’s
Analysis of the cross-correlation maps between sub-stacks AFTER residual
alignment

Absolute inverted traces compared to log at well locations for IP, IS and PR

Relative inverted traces compared to relative log at well locations

These have to be checked while keeping in mind the initial seismic to well tie

Cross-plotting absolute and relative inverted traces vs relevant logs

Cross-plotting IP/PR logs and IP/PR inverted traces in absolute and relative mode

Analysis of inverted IP and PR sections: consistency between the two parameters,


ringing, interpretability, continuity

Analysis of residuals per substack


ƒ RMS amplitude maps over zone of interest
ƒ Residuals from a pre-stack inversion are always higher than for a post-stack one

Comparison of the spectrum of P imp. reflectvities to the spectrum of Near input


data and the spectrum of PR reflectivities to the spectrum of Far input data

26 - Seismic inversion course, UPPA, 2008


P impedance volume

27 - Seismic inversion course, UPPA, 2008


S impedance volume

28 - Seismic inversion course, UPPA, 2008


Poisson’s ratio volume

29 - Seismic inversion course, UPPA, 2008


Comparison of residuals for each substack

Mid residuals

Far residuals Near residuals

30 - Seismic inversion course, UPPA, 2008


Seismic energy in a reservoir interval
Near

Input seismic Synthetic seismic Residual seismic

Far

31 - Seismic inversion course, UPPA, 2008


Ip/Is Crossplots
Water
Sand

Cross-plot at seismic scale


IP

Oil
Sand
Cross-plot on inverted traces
IP

IS
Cross-plot at log scale

IP
IS
Water sand (GamaRay<40) + Shale (GamaRay>40)

Oil sand (GamaRay<40)

IS

32 - Seismic inversion course, UPPA, 2008


Lithological classification
1
2
2.5 2 2: Shale Facies regions defined in

Lithologic facies
3
4 a cross-plot
5
6
3 3: Hydrocarbon sands
8
14
15 1 1: Water sands
2
16

33 - Seismic inversion course, UPPA, 2008


Facies classification

IS
Sandstone
Shales 4000

(levees)
Shales
(L. or H.)
Hemipel. 3000
Debris
Flows
2000
4000 6000 8000 IP

p(Ci/IP,IS)
1
1

0
0 20 60 70 80 20 60 70 80

34 - Seismic inversion course, UPPA, 2008


Data limitations for prestack inversion

Strong acquisition footprint (as for post-stack inversion)

Poor amplitude preserved processing (between sub-stacks)

Low signal/noise ratio (low fold per sub-stack) => Difficult well tie
for some substacks

Near angle stack affected by multiples

Far angle stack affected by complex raypaths

Poor alignment between sub-stacks

Large differences in frequency content between sub-stacks

35 - Seismic inversion course, UPPA, 2008


Summary – Quizz 3

What are the seismic input data for a prestack inversion?


Seismic angle stacks (=substacks)

What are initial seismic data QCs in a prestack inversion workflow?


Correlation between substacks, phase shift, time shift, frequency
content across substacks

In which way does the wavelet extraction in a prestack inversion differ


from a poststack one?
A wavelet has to be extracted for each substack.

Which petrophysical parameter is the main interest of a prestack


inversion?
Poisson’s ratio

36 - Seismic inversion course, UPPA, 2008

Vous aimerez peut-être aussi